Keywords: Biomechanics and rehabilitation, Activity recognition and health monitoring, Exoskeletons and prostheses - design
Abstract: Welwalk WW-1000 is a gait exercise robotic assist system that allows subjects to walk on treadmill by attaching a knee-ankle-foot robot to a paralyzed limb. Abnormal gait patterns during exercise using Welwalk WW-1000 are evaluated by gait observation or marker-based motion analysis systems. However, gait observation is a subjective and ordinal measure, and marker-based motion analysis systems are challenging to implement due to the complexity of preparing equipment and attaching markers to subjects. In this study, we propose the Welwalk WW-2000 system, which incorporated a marker-less motion analysis system that detects abnormal gait patterns during exercise using the robotic system. Using this system, it is expected that a gait exercise program can be planned from easily obtainable, objective information. This system detects the features of abnormal gait patterns using the body position coordinates of the subject obtained from three-dimensional, inertial, knee angle, and load sensors. The purpose of this study was to validate the marker-less motion analysis system against marker-based motion analysis systems. One healthy male simulated the seven abnormal gait patterns which occur frequently in stroke patients, with four grades of severity. Spearman's rank correlation coefficients were calculated for the relationship between the abnormal gait pattern parameters calculated by each motion analysis system. The correlations between the two systems ranged from 0.81 to 0.95. Therefore, it was confirmed that the marker-less motion analysis system of the Welwalk WW-2000 was valid.

Keywords: Biomechanics and rehabilitation, Biologically inspired systems - control, Human-centered design
Abstract: Bipedal gait is the natural means of human locomotion. Nonetheless, it is still unclear how the central nervous system coordinates the whole-body segments for gait generation. We address this question based on the well-known hypothesis that the human motion is the result of an optimization process. We consider a reduced set of criteria taken from the observation of human walking and the study of the related literature, which seem to be optimized during the human gait. Differential Dynamic Programming is applied on these criteria with a 3D whole-body skeletal model involving 42 degrees of freedom to generate walking motions. Nine different skeletal models and gaits reconstructed from motion capture data are used to this end. The simulated walking motions are then analyzed and compared to the human reference to show the quality of the gait generation process. The interest of this optimization approach for human-like motion generation is finally discussed.

Keywords: Surgical robotics - design, Surgical robotics - control, Novel mechanisms and actuation
Abstract: In spinal surgery, the surgeon needs superb skills to determine the extent of penetration of cutting tool for preventing the damage of nerves or organs. Thus, a cutting system with autonomous penetration detection would drastically improve the safety and efficiency of surgery. In this study, a hand-hold bone cutting system for laminectomy with on-line autonomous penetration perception of bone is presented. Since the penetration during operation process is invisible, a practical surgeon uses force feedback and sound to perceive cutting state. Inspired by that, the proposed system was designed to recognize different cutting states by measured cutting force and sound signals. A radial basis function neural network was employed to classify different bone cutting states. Results of cross validation prove that the accuracy of perception system reaches up to 95%. The actuation of the cutting tool can be automatically stopped as the penetration happens. In addition, the proposed system can also be applied to other bone cutting operations like craniotomy or orthognathic.

Keywords: Soft robotics, Biologically inspired systems - design, Novel mechanisms and actuation
Abstract: Gastrointestinal (GI) tract related diseases are common and deadly. To avoid the diseases developing into an advanced stage, early inspection and intervention are crucial. However, existing inspection methods are either uncomfortable or lack of active steering. Therefore, it is necessary to investigate a locomotion method that is friendly to the GI tract. This paper proposed a novel soft snail robot targeted for active locomotion inside the GI tract. By studying the locomotion mechanism of gastropods, a snail robot propelled with longitudinal travelling wave was designed. The wave was generated by deflating and inflating a series of chambers. The snail robot could adhere to different surfaces, including the ex vivo porcine intestine, by the mucus underneath. Experiments were designed to study the performance of the proposed snail robot. Results show that, the soft snail robot could crawl over different substrates including hard substrate, soft substrate and biological tissue. The speed of the snail is affected by multiple factors, including the surface condition, wave frequency and the mucus concentration.

Keywords: Human-centered design, Biomechanics and rehabilitation, Novel mechanisms and actuation
Abstract: A major desire of individuals with a spinal cord injury (SCI) is the ability to functionally move the trunk while at the same time maintaining a stable posture. The trunk movements can be actuated by Functional Neuromuscular Stimulation (FNS), which applies low-level current to the peripheral nerves to activate the paralyzed muscle , and thus generate forces and joint moments sufficient to move the limbs. Ensuring the precision and stability of the movement requires powerful control methods as the musculoskeletal dynamics of the trunk are highly complex. In this paper, an integrated control loop, which combines a joint moment (JMT) controller and a muscle (MUS) controller, is developed to control trunk motions. The JMT controller which is comprised of a prior computed feed-forward torque and a model predictive controller (MPC), determines the torque vector required to achieve the motion. However, the computed torque cannot be directly applied to an actual trunk system. This limitation motivates the MUS design, where a feed-forward activation map (FAM) determines the muscle activation that can generate the desired torque vector while a neural network (NN) based correction mechanism is adjusted by an anatomy-based updating law. Therefore, when the desired motion is determined, the corresponding activation can be found and applied to the muscles of the trunk. The integrated control system is validated in simulation, in which the trunk is moved to a variety of reference postures. Simulation performance suggests that the method has a high potential for trunk control applications in individuals with SCI and other impairments.

Keywords: Biologically inspired systems - design, Soft robotics, Human-centered design
Abstract: Flexure graspers have been identified as a bioinspired evolution in robotics to adapt to unpredictable disturbances and uncertainties in unstructured environments. However, these soft robots with infinite degrees of freedom (DOFs) have brought in many technical challenges, among which is the sensing of spatially-distributed and time-varying force and displacement fields for feedback control with a finite number of nodal measurements. This paper proposes a simultaneous motion and force sensing method based on discrete strain data with an illustrative application to a flexure finger that can serve as an actuation and sensing unit in a robotic hand. The method is rigorously formulated by introducing the strain mode shapes and correlating continuous displacement and force/moment distributions with curvatures of deformation. Design analysis of the flexure robotic finger is presented and the compliant composite joints are fabricated via shape deposition with embedded strain gauges. The proposed sensing method is numerically verified with finite element analysis and experimentally validated on a prototype of the flexure finger.

Keywords: Human-robot interaction, Biomechanics and rehabilitation, Technology assessment in human subjects/outcomes
Abstract: Combining functional magnetic resonance imaging (fMRI) with models of neuromotor adaptation is useful for identifying the function of different neuromotor control centers in the brain. Current models of neuromotor adaptation have been studied primarily in whole-arm reaching tasks that are ill-suited for MRI. We have previously developed the MR-SoftWrist, an fMRI-compatible wrist robot, to study motor control during wrist adaptation. Because the wrist joint has intrinsic dynamics dominated by stiffness, it is unclear if these models will apply to the wrist.

Here, we characterize adaptation of the wrist to lateral forces to determine if established adaptation models are valid for wrist pointing. We recruited thirteen subjects to perform wrist pointing with the MR-SoftWrist during lateral perturbations. Our task included a clockwise (CW) - counterclockwise (CCW) - error clamp schedule and an alternating CW-CCW force field schedule. To determine applicability of previous models, we fit three candidate models - a single-state, two-state, and context dependent multi-state model - to behavioral data.

Our results indicate that features of sensorimotor adaptation reported in the literature are present in the wrist, including spontaneous recovery, and anterograde and retrograde interference between the learning of two oppositely directed force fields. A two-state model best fit our behavioral data. Under this model, adaptation was dominated by a fast learning state with minor engagement of a slow learning state. Finally, all adaptation models tested showed a consistent over-estimation of performance error, suggesting that the control of the wrist relies not only on internal models but likely other mechanisms, like impedance control, to reject perturbations.

Keywords: Exoskeletons and prostheses - control, Exoskeletons and prostheses - design, Technology assessment in human subjects/outcomes
Abstract: Clinical prescription and tuning of prosthetic legs relies heavily on a trial-and-error paradigm. Patient preference, a formalized representation of “what patients want”, provides an efficient method for identifying appropriate prosthesis mechanics. Unfortunately, a lack of the scientific and technological tools necessary to formally measure and rigorously understand preference has led to limitations in ascertaining its role in the prescription process. In this study, our objective was to understand how preference changes with walking speed and how preference corresponds to clinical metrics of gait performance. We characterized preferred stiffness of the Variable Stiffness Prosthetic Ankle Foot (VSPA Foot) across speeds in persons with below knee amputation. We then assessed metabolic expenditure and performance in the ten-meter walk test at several stiffness values around the preferred stiffness. Preferred stiffness did not correlate with body mass, and was lower at the self-selected treadmill walking speed than at either the fast or slow speeds. There was a significant increase in self-selected overground speed in the ten meter walk test at stiffness values at or above the preferred stiffness. Changes to prosthetic ankle stiffness within +/- 30% of the preferred stiffness did not significantly affect metabolic expenditure. These results indicate that preferred stiffness may be associated with significant changes in self-selected overground walking speed (p < 0.001), and may not be driven by metabolic optimality (p > 0.93). This outcome motivates future work to identify other potential drivers of preference, including biomechanical correlates of gait, and highlights the potential value of formally incorporating patient preference into the clinical prescription process.

Keywords: Human-robot interaction, Human-machine interfaces, Surgical robotics - design
Abstract: Foot-based teleoperation can provide the user with hands-free control of a robotic assistive arm without interrupting the ongoing task in various potential industrial and medical applications. In this paper, a foot interface (FI) was adapted to control a robotic camera holder in laparoscopic surgery, where surgeons are in direct contact with patients, and their hands are occupied. A novel four degrees of freedom FI with intuitive two-level rate-based control and vibrotactile feedback is proposed. A comparative study was conducted between the proposed FI and a conventional FI with binary switches on a planar surface. The experiment includes a follow-the-command test and a peg-and-holes test. Results showed that our FI was less mentally tiring with shorter task completion time and fewer visual checks. A suitably designed FI is a good candidate for controlling a robotic assistive arm in surgical applications.

Keywords: Biological signal processing and identification, Exoskeletons and prostheses - control, Biomechanics and rehabilitation
Abstract: Acquisition of surface electromyography (sEMG) from a person with an amputated lower extremity (LE) during prosthesis-assisted walking remains a significant challenge due to the dynamic nature of the gait cycle. Current solutions to sEMG-based neural control of active LE prostheses involve a combination of customized electrodes, prosthetic sockets, and liners. These technologies are generally: (i) incompatible with a subject’s existing prosthetic socket and liners; (ii) uncomfortable to use; and (iii) expensive. This paper presents a flexible dry electrode design for sEMG acquisition within LE prosthetic sockets which seeks to address these issues. Design criteria and corresponding design decisions are explained and a proposed flexible electrode prototype is presented. Performances of the proposed electrode and commercial Ag/AgCl electrodes are compared in seated subjects without amputations. Quantitative analyses suggest comparable signal qualities for the proposed novel electrode and commercial electrodes. The proposed electrode is demonstrated in a subject with a unilateral transtibial amputation wearing her own liner, socket, and the portable sEMG processing platform in a preliminary standing and level ground walking study. Qualitative analyses suggest the feasibility of real-time sEMG data collection from load-bearing, ambulatory subjects.

Keywords: Human-robot interaction, Haptics
Abstract: In this paper we derive the control parameter gains required to guarantee a stable human-robot interaction (HRI). One goal in HRI research field is to reduce the robot mechanical impedance and enable humans to easily manipulate the robot. Force feedback is an effective way to reduce robot inertia and friction, but stability is paramount especially when interacting with humans. This paper discusses gains and stability boundaries. The phase response of the open loop transfer function that represents the human robot interaction is used to derive the gains. The environment is modeled as a second-order spring-mass-damper system. The stability boundaries correspond to the region when the phase response of the transfer function becomes greater than -180 degree at all frequencies. This enables the system to interact stably regardless of the environment second-order system parameters. One of the interesting results is that the virtual spring used to generate the impedance field has nothing to do with the stability condition. The experimental results show the validity of our method.

Keywords: Human-robot interaction, Human-machine interfaces, Activity recognition and health monitoring
Abstract: In this paper, we propose a method for improving the human operator's arm posture during bilateral teleoperation. The method is based on a musculoskeletal model that considers human operator's arm dynamics and the feedback force from the haptic interface (master), which is used to control a robotic arm (slave) in a remote environment. We perform an online optimisation to find the optimal configuration that has the longest endurance time with respect to muscle fatigue. Next, a trajectory is generated on the haptic interface in order to guide the human arm into the optimal configuration. The teleoperation is temporarily suspended by decoupling the master from the slave robot when the haptic device is being reconfigured. Afterwards, the loop is coupled again and the slave robot is controlled from the position where it stopped after the haptic interface guided the operator's arm to the optimised configuration. The main advantage of the proposed method is that the human operator can perform the task with less effort, which increases the endurance time. To validate our approach, we performed proof-of-concept experiments on a teleoperation system composed of two Franka Emika robots, where one was serving as master and the other as slave.

Keywords: Biomechanics and rehabilitation, Activity recognition and health monitoring, Biological signal processing and identification
Abstract: This paper proposes an identification method of a human’s standing stabilization control scheme. The movement of the center of mass (COM) in the motion is focused on rather than that of a couple of joints employed in well-acknowledged ankle/hip strategies in order to understand the human’s behaviors in more general situations. It is mathematically represented as a piecewise-affine dynamical system, in which the state space is divided into some regions described by different equations of motions. The representation is based on the authors’ previous finding that the COM-ZMP regulator, which was originally designed for humanoid robots to stabilize COM by manipulating the zero-moment point (ZMP), qualitatively models the humans’ control scheme. A technical difficulty in the identification of such a piecewise system is that it is a chicken-and-egg problem since the equation of description has to be provided in order to identify system parameters, while the system parameters are required in order to choose the equation of description. The proposed method utilizes K-means method and/or EM algorithm and was applied to motion loci of a human subject in lateral direction measured in the previous study. The result of the identification quantitatively supported the above hypothesis. The differences of the actual human’s behavior from the model are additionally discussed.

Keywords: Exoskeletons and prostheses - control, Exoskeletons and prostheses - design, Human-centered design
Abstract: In recent years, while the number of elderly lower limb amputees has been increasing, the use rate of lower limb prosthesis among the elderly has been low due to the risk of falling. Knee joints in the transfemoral prosthesis play an important role in regaining the gait ability of amputees. Particularly, the electronically controlled knee joint enables full prevention, by a system that automatically adjusts the optimal knee flexion resistance force based on the estimation of the gait state. However, it is difficult for the user to predict how much load is applied to the knee joint, which is one of the references for adjusting the bending resistance. This reduces the intuitiveness of the control system given to the user. In this study, we developed a robotic knee prosthesis that prevents falling with a proposed gait model to control the locking mechanism of knee flexion. Finally, we conducted experiments with the developed knee joint and showed the effectiveness of fall prevention.

Keywords: Biomechanics and rehabilitation
Abstract: Hand rehabilitation requires intensive training with various gross and fine movements. Thus, rehabilitation robots for the hand that accommodate for the different degrees of freedom are often complex and expensive. This work presents the design of a portable plug-and-train robot (PLUTO), which tackles the problem by having a single actuator that can be coupled with different passive mechanisms for training wrist flexion-extension, ulnar-radial deviation, hand opening-closing, and forearm pronation-supination. The robot is capable of providing training in active and assisted regimes. Training is through performance adaptive computer games to provide feedback to the patients and to motivate them during training. The usability was evaluated in patients, caregivers, and clinicians with standardized questionnaires: System Usability Scale (SUS) and User Experience Questionnaire (UEQ). Patients and caregivers were administered the questionnaire after two training sessions. Clinicians, on the other hand, had a single session demo after which their feedback was obtained. In this paper, we present the initial results of 5 clinicians, 5 caregivers, and 5 patients. All groups found the system to be highly usable (>80 scores on the system usability scale). Furthermore, the scores from UEQ feedback were all positive, and all groups found the system attractive. The patients and the clinicians rated the system positively in both pragmatic and hedonic scales. We believe that a simple approach proposed here can result in a compact tool with a high benefit-to-cost ratio for both in-clinic and home-based hand rehabilitation.

Keywords: Exoskeletons and prostheses - design, Biomechanics and rehabilitation, Novel mechanisms and actuation
Abstract: Effective rehabilitation treatment is crucial for children with cerebral palsy to maintain their mobility while growing up. However, conventional interventions like passive orthosis do not sufficiently relieve pathological gait in most patients. In recent years, motorized exoskeletons showed great success in the rehabilitation of stroke and paralyzed patient and they also give a promising opportunity to the treatment of cerebral palsy. However, the rapid growth rate and morphological variation among children with cerebral palsy bring a great challenge to the structure design. In this paper, the mechanical structure of an adjustable lower-limb exoskeleton is designed for children with cerebral palsy. In order to have general applicability for users with different sizes, this exoskeleton can change its frame lengths to adapt to the height of children from eight to twelve years old. Besides, for having the kinematic compatibility, it also has a joint-aligning mechanism that lets the exoskeleton's hip joint comply with the natural hip internal/external rotation. Simulation is conducted for verifying the kinematics of the exoskeleton and comparison is made between the one with/without the joint-aligning mechanism for studying its effectiveness. Results show that the extra torque due to the hip joint misalignment is eliminated.

Keywords: Exoskeletons and prostheses - design, Human-robot interaction, Biomechanics and rehabilitation
Abstract: This paper developed a cable-driven 3-DOF (degree of freedom) wrist rehabilitation robot with optimized human-robot interaction performance. The workspace and cable tension efficiency of the robot are increased by adding the dynamic adaptive structure of cable attachment points, which improves the safety and comfort of human-robot interaction. In addition, the DASA (distributed active semi-active) system based on MRs(magnetorheological) is adopted, which reduces the effects of elastic elements (e.g., cables or Bowden cables) in the system, improves the force-feedback performance. The robot is compact and can be used independently or transplanted to any upper limb rehabilitation robot without wrist rehabilitation function. The performance of the system and control algorithm is verified by several experiments on healthy subjects. The results show that the system can meet the needs of rehabilitation training for workspace and force-feedback. This system also has the potential to be a force-feedback device for the healthy person.

Keywords: Exoskeletons and prostheses - design, Biomechanics and rehabilitation, Wearable technologies
Abstract: Stroke often results in hemiplegia, which greatly affects the walking ability of the patients. We propose a multi-functional portable ankle exoskeleton for use in preventing foot-drops, assisting propulsion, and stabilizing inversion/eversion during walking to help gait rehabilitation of stroke patients. The portable ankle exoskeleton was fabricated by 3D printing a soft/rigid hybrid structure. The device was able to prevent foot-drop and assist propulsion with a bi-directional cable-driven actuation system. It also showed a capability of stabilizing inversion/eversion motions using a counter-electromotive force of two small, lightweight gear motors. The device was controlled by a microcontroller based on real-time feedback from one inertial measurement unit and a customized force sensitive resistor. The device is fully untethered with all the components integrated on-board, with a total weight of less than 1 kg. Five healthy subjects performed over-ground walking tests with the proposed ankle exoskeleton for three different walking situations (normal walking, walking with simulated foot-drop, and walking on an uneven terrain) and three walking conditions (without the exoskeleton, with the exoskeleton powered off, and with the exoskeleton powered on). From the test results, we confirmed the feasibility of the proposed ankle exoskeleton for foot-drop prevention, propulsion assistance, and inversion/eversion stabilization. The ankle exoskeleton showed a potential for wearable gait rehabilitation for stroke patients with high mobility and portability.

Keywords: Biologically inspired systems - control, Biologically inspired systems - design, Human-robot interaction
Abstract: The head mechanism is a critical part of a humanoid healthcare robot for potential healthcare or elderly care. The head structure of a humanoid robot should provide both motion and perception capabilities to mimic head function (e.g. object tracking) for Human-robot interaction (HRI) at the head level. In this paper, an innovative hybrid head mechanism consisting of one rotational servo motor and two linear servos has been proposed. Structural design, kinematic analysis, and automation have been studied to advance locomotive capability as well as functionality. Additionally, a visual servoing object tracking system was implemented as a perception unit to achieve real-time object tracking function. For the structural design, a parallel configuration of two linear actuators was selected to simplify the kinematic analysis, which enables precise control of joint angles. For the perception part, a vision-based real-time ArUco maker tracking system with proportional–integral–derivative (PID) control was set as a test scenario to mimic human head function with joint control of 3 DOFs (yaw, pitch, and roll). The functionality of this head mechanism has been validated by tests on potential use case (visual servoing object tracking). This head mechanism can servo as a valuable platform for broader HRI applications with humanoid healthcare robots.

Keywords: Biologically inspired systems - design, Novel mechanisms and actuation, Biologically inspired systems - control
Abstract: Animals often use their external appendages (such as tails, limbs) to achieve spectacular maneuverability, energy efficient locomotion, and robust stabilization to large perturbations, which may not be easily attained in the existing legged robots. Their appendages, particularly, the tails are very compact, light, highly dexterous with a large range of motion (RoM). Animals can also curl and straighten up their tails in less than one-tenth of a second to facilitate rapid adjustment for the moment of inertia (IoM) and the center of mass (CoM). Most of the existing robotic tail designs still lack dexterity, output force, dynamic response as compared to their biological counterparts. The objective of this research is to create a viable and compact solution to adjust the tail IoM rapidly and effectively for the enhancement of robot agility. We design and build a novel 3-DoF morphable inertial tail based on a spherical linkage mechanism, incorporated with a spring-load telescopic tube to realize a wider range of inertial adjustment ability. We also develop a dynamic model and propose motion controllers to study the tail-inspired dynamic locomotion. Simulation and initial experimental results demonstrate the effectiveness of the proposed mechanical design and controllers to enhance more agile locomotion using the swinging tail.

Keywords: Soft robotics, Biologically inspired systems - design, Novel mechanisms and actuation
Abstract: — Liver cancer is the second leading cause of cancer deaths worldwide. Without treatment, the five-year survival rate is less than 5%. Liver tumors are often detected using MRI. MRI is preferable over other imaging modalities include absence of ionizing radiation, superior soft tissue contrast resolution, and multiplanar imaging capabilities. Needle interventions are then needed for further diagnosis (biopsy) and treatment (ablation). Respiratory motion is the main cause of inaccurate needle placement leading to misdiagnosis and insufficient treatment. In order to train clinicians to compensate for the liver respiratory motion during intervention, and also test needle insertion robotics technologies, a realistic liver motion simulator is of great interest which can be used inside MRI scanner. In this study, an MR-compatible soft robotic platform is developed to simulate liver respiratory motion. Pneumatically actuated soft robotic actuators were developed and validated using finite element simulations. A feedforward controller was developed to control the input air pressures of the pneumatic actuators to simulate the respiratory motion. The evaluative experiments indicated that the developed soft robotics diaphragm could acceptably simulate liver respiratory motion in the two directions of motion: superior-inferior and anterior-posterior directions.

Keywords: Biological signal processing and identification, Biologically inspired systems - design, Human-machine interfaces
Abstract: Functional electrical stimulation (FES) is a widely used method to bypass spinal chord damages and bring the neurological input to muscles. FES methods that allow the activation of both superficial and deep muscles were proposed in the literature. However, even if many studies using surface electromyography (sEMG) for the estimation of muscles' activation were presented, it still remains impossible to access information of deep muscles. Our group proposed a solution that exploits the morphological information of magnetic resonance imaging (MRI) and the sEMG for the estimation of the average muscles' activation pattern during isometric muscle contractions of forearm muscles.

With this work we introduce the time dimension to study how the activation patterns changes while performing 3 simple wrist movements. The study involved data collected from a single row of electrodes wrapped around the forearm of a single healthy participant. Morphological information are extracted from the MRI to build a model that is then used to estimate the muscles' activation pattern.

The results show that the method is able to provide an estimation of the muscles' activation patterns that explain more than 93% of the input sEMG information and that have a corresponding in anatomical information from the literature.

Keywords: Wearable technologies, Soft robotics
Abstract: In this paper, we present a biomimetic 2-DOF SMA-actuated robotic arm that can be controlled by a wearable sleeve in real-time. The designed lightweight robotic arm is intended to be an alternative to the existing heavy and bulky systems, used in different areas such as rehabilitation, haptics and, surgical robotics, etc. which are actuated by the regular hydraulic/pneumatic pistons and brushed/brushless motors. The robotic arm weighs 59g with a wide controllable range of motion (119˚ and 123˚ for the 1st and the 2nd joints, respectively). To enable closed-loop control of the joint angular positions, a PID controller was implemented, and its performance was evaluated. Then, the robot payload was evaluated by finding the maximum torque for each joint. The comparison between the existing commercial DC motors and the designed SMA-actuated rotary joints shows that the performance of the designed SMA-actuated rotary joints are acceptable as they outperform well-known commercial motor-based rotary joints in terms of power consumption, nominal voltage, nominal torque, and mass. Next, an End Effector displacement analysis was conducted to assess the robot positioning. Finally, the entire designed teleoperation system comprising the robotic arm and the wearable measurement sleeve was assessed by performing 20 flexion-extension trials. An average RMSE of 13.1mm was achieved for EE displacement.

Keywords: Biomechanics and rehabilitation, Activity recognition and health monitoring, Biological signal processing and identification
Abstract: The goal of this work is to find a plausible mathematical model of a human’s step-and-brake control scheme as the minimum motion unit of skillful locomotive behaviors. A controller designed for biped robots based on the reduced dynamics and minimal conditions to stabilize robots, namely, the terminal capturability was picked up as the primary candidate of the model. Motion loci of the center of mass (COM) and the zero-moment point (ZMP) of a human subject were measured, and control parameters were identified from them. The identification result showed that the model well suited the human’s behavior in spite of its minimal property only with a slight modification taking an asymmetry of the natural falling speed depending on the direction and the inertial torque about COM into account. A contribution of this work over the previous researches which focused on the capturability is that a model of the feedback controller that reproduces the overall dynamic stepping motion process was identified in a mathematically explicit form in it.

Keywords: Exoskeletons and prostheses - control, Activity recognition and health monitoring, Brain-machine interfaces
Abstract: Brain-computer interfaces (BCI) in combination with assistive robotic devices, such as wearable robotics, has the potential of augmenting the capabilities of disabled people to carry out activities of daily living with success. To improve applicability of such systems, workload and stress should be reduced to a minimal level. In this paper, the degradation of the performance of EEG hand-exoskeleton control with the exhaustive use of the interfaces is analysed through the monitoring of user's physiological reactions. Eleven BCI-naive volunteers participated in the study. The participants performed several open/close hand motor imagery trials for 6 minutes. After completing the task, both the NASA TLX questionnaire and self-assessment manikin (SAM) were submitted to the user. The results broadly suggest that there are significant differences (p-value<0.05) in heart rate variability (HRV) changes between subjects that showed good and poor performance using the BNCI. In addition, these objectives results are corroborate with the results of subject's workload perception and emotional responses assessed through NASA-TLX questionnaires and Self-Assessment Manikin (SAM) respectively. Our main finding is that the subjects' performance using a BCI-controlled hand exoskeleton produce physiological reactions in that subjects.

Keywords: Wearable technologies, Algorithms and machine learning, Activity recognition and health monitoring
Abstract: Training induced fatigue and recovery are deemed to drive adaptations leading to performance enhancements in exercise and sport. As a result, training loads should be accurately controlled and regulated to promote physiological and biomechanical adaptations. In this work, we investigated the sensitivity of inertial sensors to detect fatigue induced changes in gait kinematics in well trained team sports athletes. Thirteen young healthy subjects volunteered to perform a walking trial before and after an exhausting field test while wearing inertial sensors fit on their lower limbs, pelvis and trunk. Stride time (ST), ST variability (STV), stride length (SL), SL variability (SLV), gait speed (GS), symmetry index (SI), knee range of motion (ROM) and shank angular velocity (AV) were computed. These features were used to feed a support vector machines (SVM) classifier to distinguish non-fatigued and fatigued walking trials. Results showed significant (p < 0.05) pre-to-post changes in ST, STV, GS, SI and AV with the SVM classifier reporting an 84.62% accuracy. Thus, classification using gait features collected through inertial sensors could be promising for in-field fatigue detection.

Keywords: Activity recognition and health monitoring, Technology assessment in human subjects/outcomes, Biomechanics and rehabilitation
Abstract: The spontaneous kicking patterns of infants can provide markers that may predict the trajectory of their future development. Atypical kicking patterns may forecast the possibility of developmental disorders like Cerebral Palsy (CP). Early intervention and physical therapy that encourages the practice of proper kicking motions can help to improve the outcomes in these scenarios. We introduce a system that utilizes computer vision and a robotic infant mobile to detect spontaneous kicking patterns and activate different stimuli provided by the mobile. The hypothesis is that the combination of stimuli will encourage the continuation of kicking actions. This would help with studies in infant motor development and identify at-risk populations. Additionally, the system is designed for in-home use based on its space-invariant approach to changes in the environment. In this paper, we detail the design of the robotic system, discuss the preliminary results of deploying the system in an infant’s home environment and the effect it has on the frequency of kicking actions.

Keywords: Biomechanics and rehabilitation, Activity recognition and health monitoring, Biological signal processing and identification
Abstract: A method to identify a human model as a rigid kinematic chain and marker arrangement, which is referred as a human skeleton-marker model in this paper, from the data collected by an optical motion capture system is proposed. The model is utilized for convenient and accurate motion analyses based on robotics computations. A chicken-and-egg problem to find the model to estimate the whole-body posture and the whole-body posture to fit the model simultaneously is resolved by a dual-phase nonlinear least square error minimization. The computation cost mainly due to the numerical computation of the gradient of the cost function with the inverse kinematics is significantly reduced by applying the stochastic gradient descent (SGD) method. Despite it only uses partial samplings for the estimation of the global gradient, it does not sacrifice the accuracy since it stochastically reflects information of the overall motion sequence through the iteration.

Keywords: Activity recognition and health monitoring, Biological signal processing and identification, Algorithms and machine learning
Abstract: Human activity recognition is an essential field to study for healthcare services and smooth human-machine interaction. For human motion data collection, wearable sensors are the common devices utilized. In this paper, a new double-channel motion data structuring method for classification is proposed for the data collected from the inertial sensor of a smartphone. It was shaped as a virtual image in a way to extract deep features of the temporal and spatial dependencies among motion signals. The time-series raw data underwent Fourier and wavelet transformations as alternative forms of input data. The virtual images from the three types of representations were made to fit a Convolutional Neural Network model for classification. The proposed model was evaluated using our dataset and other public datasets, where it performed well with all datasets. The trained model showed excellent results when tested on a computer and smartphone for real-time recognition. An exclusive iOS application with data handling and real-time recognition functions was developed for the smartphone. The model has the attributes of lower computational cost and better accuracy, which makes it a sound model for a practical purpose.

Keywords: Biologically inspired systems - design, Pathological assessment/diagnosis
Abstract: Current reconstructive hand tendon surgeries are based on clinical experience and cadaver studies. New biomechanical hand models and anatomically correct robotic testbeds provide a potential alternative to surgical trial-and-error or cumbersome cadaver studies. We introduce a new index finger biomechanical model and anatomically correct robotic testbed to explore this potential application. The model and testbed are used in developing a novel methodology for simulating the active range of motion outcomes of flexor digitorum profundus repair surgery based on shortening of the tendon that can occur during reconstruction. Simulated and experimental results demonstrate the benefits of using both models and robotic testbeds. Results also support the clinical recommendation of a 10 mm shortening limiting to maintain functionality.

Keywords: Pathological assessment/diagnosis, Brain-machine interfaces, Wearable technologies
Abstract: A new neuroimaging device has been developed to diagnose and monitor brain disorders using functional near-infrared spectroscopy (fNIRS). The system has enhanced spatial as well as temporal resolution, and it is configured in a tightly arranged mesh, which consists of a single photodiode (PD) and 128 dual-wavelength LEDs. The developed device can span a square-shaped area of nearly 49 cm2. Multiple source-detector separations are used that enable us to reach multiple depths inside the brain from 2 cm to 3.5 cm. The system also provides the superficial layer information by measuring the short-separation channels. The short-separation channels allow the removal of noise and enhancement of signals. The drive circuit of LEDs is carefully designed to switch the light with appropriate intensity, which provides a stable reception for each channel. Low-side switching ICs based on MOSFET technology are used for high-speed fNIRS signal recording. The system can display the acquired HbO and HbR signals as well as activation maps in real-time on a lab-developed Windows-based software. The hardware connects to the software using Wi-Fi. Phantom model with known optical properties and a human subject were used for testing the functionality and efficacy of the device. A complete 128 channel fNIRS sample was recorded in 25 ms. The phantom results showed reduced signal intensity when the channel separation was increased that provides the HbO and HbR. The activation was seen using HbO in the human subject while performing hand tapping task.

Keywords: Surgical navigation and localization, Computer vision in surgery
Abstract: Laparoscopic surgery can realize minimal invasive surgery. However, it’s difficult for surgeon to recognize the depth during suturing. Binocular endoscope helps surgeons to recognize the depth, but surgeons do not understand the circumstance by equipping with head mounted display. Since shadow helps surgeon to recognize the depth, in this paper, we developed the virtual shadow drawing system. The system shows shadow like actual by the estimation of the forceps position, surface shape and shadow’s position. We tested the accuracy of the system by evaluating the estimated distance and angle between the forceps and the surface. The error and delay were enough small to draw shadow like actual. Furthermore, participants performed the suturing task while looking at the shadow. The experiment was carried out in a variety of the shadow’s direction. As result, the suturing error’s mean and variance value was the least at the 270 deg. In conclusion, the appropriate shadow would be vertical to the wounds.

Keywords: Biomechanics and rehabilitation, Activity recognition and health monitoring, Biological signal processing and identification
Abstract: Two approaches to on-line fitting of the human kinematics model and marker arrangement are proposed for accurate pose estimation in motion tracking. A ‘flexible’ skeletonmarker model that can adjust the lengths of the body segments and relative locations of markers in addition to the whole joint angles is employed. In order to avoid the ill-posedness due to the huge degrees of freedom and the overfitting, the model and the whole joint angles are updated frame-wise. The particle filter and a frame-wise gradient descent method were examined for the model update. The former was applied based on an expectation that a stochastic technique can help to avoid the overfitting of the model. The latter had a success in the authors’ another work to estimate the global gradient to reduce the estimation error only from one sample. While the both techniques worked from the viewpoint of accuracy, it was found that the particle filter had a drawback in the aspect of computation cost, so that the frame-wise gradient descent method is more preferable although it still has a problem in a high-rate implementation.

Keywords: Biomechanics and rehabilitation, Human-machine interfaces, Pathological assessment/diagnosis
Abstract: Acute sleep deprivation affects negatively cognitive and motor tasks such as postural control, but less results have been presented about the effects of sleep conditions on other motor tasks, such as the bimanual guidance of a handlebar. The design of an instrumented robotic handlebar is presented along with a motor control test aimed at evaluating sleep effects on task performance. The actuated handlebar is able to register the forces and the movement synchronized with a display that shows targets to be reached with the handlebar. Prior to tests, ten health subjects answered to questionnaires in order to access their sleep parameters. The participant's performance data (spatial accuracy, reaction time and time to reach the target) was compared in terms of the sleep assessment outcomes. We observed an association between the total sleep time and the time of reach, which is very similar to the observed in Psychomotor Vigilance Tests in subjects who not slept enough. In this way, this handlebar test appears to be a useful tool to detect impairments resulting from sleep disturbances. However, authors point out to the need of further studies.

Keywords: Surgical robotics - design, Novel mechanisms and actuation
Abstract: In natural orifice transluminal endoscopic surgery (NOTES), the manipulator design is a great challenge because it needs to meet different requirements for stiffness and size during the whole process. In our recent work, a braid-based manipulator has been proposed, which behaves as a stand-alone braid to fold and deploy in normal state, but turns to be rigid when a negative pressure is applied. It exhibits a stiffness ratio of 6.85 and diameter variability, but the control of the tunable diameter was not declared. In this paper, a bi-directional tunable-diameter mechanism for the braided skeleton is proposed, which uses shape memory alloy (SMA) braided fibers to deploy to a large profile state by electric heating, and rubber bands to force the skeleton back to the slim state when cooled. The mechanism is validated with a physical model, which can achieve a ratio of 1.46 between the maximum and minimum diameters. A theoretical model is also established based on the open-coiled theory, and parametric analysis is conducted to further detail the mechanism. The results show the promise of the mechanism in braided manipulator.

Keywords: Technology assessment in human subjects/outcomes, Exoskeletons and prostheses - control
Abstract: Stroke survivors usually present asymmetric gait and mobility impairments as consequence of hemiparesis. In order to alleviate this unilateral loss of mobility and correct this asymmetry, we developed an active knee-orthosis to assist the movement of the impaired leg. The control strategies of the orthosis that we have implemented are based on the kinematics of the healthy leg, by replicating the movement of the healthy one, or by using this movement to synchronize the application of a normalized healthy pattern over the impaired leg. Our main aim is to understand how stroke patient’s gait evolve to adapt to the assistance provided, as a preliminary study, in this paper we describe the effects of the assistance provided in twelve (12) healthy subjects who wore the robotic exoskeleton while they performed overground walking trials. Results indicated that the robot improved the gait symmetry of the users, reducing the spatial and temporal differences between the movements of both legs, without significantly disturbing the stance/swing ratio of the users. In addition, results showed that users tended to respond to this assistance by generating torque in the same direction of the robot action instead of taking advantage of the assistance provided by the exoskeleton. Muscular response of the assisted leg also shows variations according to this torque response, increasing the muscular activity in the 22.7% of cases against the 4% of cases where it was decreased. Finally, we detected changes not only in the assisted leg, but also in the unassisted one, whose muscular activity increased in the 28% of cases against the 1% of cases where it was decreased, showing the effect that the assistance on one leg have over the whole gait dynamics.

Keywords: Biomechanics and rehabilitation, Algorithms and machine learning, Technology assessment in human subjects/outcomes
Abstract: This paper summarizes our efforts on organizing a data structure for the analysis of human generated kinematic and kinetic data. We introduce a framework to perform data analysis in a straightforward manner, from the raw data to the refined results of a particular study. We applied this framework for data collected with the MIT MANUS gym. The proposed data structure carries a summary of the raw data, in terms of mean-aggregated robotic assay, also preserving the metrics of each movement separately for further study of movements at different granularity levels. Our framework allows us to export the structure of preprocessed data in a portable format, so that data can be accessed or edited with other statistical packages. Because the framework structure is modular and scalable, requiring no prior installation or setup, it simplifies the expansion to novel metrics and novel robotic devices. Hence, it is a convenient tool to be used in studies involving the analysis of human kinematic and kinetic data collected by any robotic device.

Keywords: Human-centered design, Human-machine interfaces, Human-robot interaction
Abstract: In designing a human-computer interface, it is vital to understand the principle behind motion generation in humans. Previous research has mostly focused on kinematics and not kinetics. In this paper, we present the characteristics of human behavior while performing force tracking tasks. Here we present a preliminary study in which three type of measurements were conducted with three healthy young adult male subjects to characterize force modulation (discrete, rhythmic, and target transition). Subjects were instructed to control the force exerted to the load cell during a force pursuit tracking task. The visually-guided discrete attempt investigated human behavior when the target changed discretely (every 5 seconds). The rhythmic attempt investigated human behavior when the target changed continuously and periodically (every 1 second). The target transition required subjects to alter the behavior of the input force. The results suggest the following: the differential value of the force generated by a human is sufficiently large, while the response speed is based on visual information; humans can predict future reference to a small degree and attempt to improve or correct their force; and the rule of superimposition comes into effect with regard to the differential value of force.

Keywords: Exoskeletons and prostheses - design, Exoskeletons and prostheses - control, Technology assessment in human subjects/outcomes
Abstract: Repetitive tasks as those performed by automotive workers are the main factor that contributes to work-related joint injuries. In the automotive industry, overhead screwing tasks are extremely common, and they seem to be the main predisposing factor to shoulder injuries. In order to mitigate this scenario, several strategies have been proposed. One of those strategies is the usage of passive exoskeleton. With the fourth industrial revolution, several companies created different exoskeletons models in order to offer a commercially available solution. Thus, our purpose was to conduct a pilot study to compare shoulder muscles activity between three commercial upper limbs exoskeletons and to assess the feasibility of a study with a large scope. We assessed two automotive workers on their workplace performing three everyday tasks with a light and a heavy pneumatic screwdriver tool, using no exoskeleton, and using ShoulderX by SuitX, Mate by Comau and Paexo by Ottobock. Surface electromyographic activity of the shoulder muscles (Deltoideus Anterior and Medialis) was assessed. The preliminary results suggest that the conditions in which the workers used the exoskeletons significantly reduced shoulder muscles activation. We also found what seems to be a significant effect of the tool’s weight. Finally, small differences were detected between the three exoskeletons on reducing shoulder EMG activity. Nevertheless, a larger sample is needed in order to confirm its significance. These preliminary results are valuable before implementing the passive exoskeleton solution in a workstation that needs to reduce joint stress.

Keywords: Biomechanics and rehabilitation, Technology assessment in human subjects/outcomes
Abstract: The technical development of the virtual reality platform provides multiple levels to understand human behaviors in simulated environments and to develop interventions for functional rehabilitation. In this study, a dual-task paradigm in a virtual environment is designed where both tasks demand motor skills. Three healthy adults (mean age: 24.3 years) participated in this study. The experiment involved two conditions of overground walking in virtual reality - normal walking and catch and throw a ball while walking. In this work, we investigated the dual-task gait characteristics and the strategy adopted at the lower limb to perform better in the secondary task of throwing the ball. Results show that more balls were thrown during the terminal stance phase of the dominant leg. Thus, the participants utilized the forward momentum built during the foot-to-foot transition by the lagging dominant foot while throwing. This study provides a new and engaging paradigm to analyze dual-motor-task in a virtual reality environment. It can be used as a powerful tool to characterize gait and cognitive performance measures in individuals.

Keywords: Biomechanics and rehabilitation, Biological signal processing and identification
Abstract: The ankle is the critical joint between the leg and foot that helps control physical interactions for propulsion, shock absorption, and balance of the body. This paper describes the effect of modulation of ankle impedance via functional electrical stimulation (FES) on balance control in humans while they are in quiet and perturbed stances. Subjects were instructed to stand on a computer-controlled treadmill while their kinematics and foot pressure were measured by an optical motion capture system and a force plate, respectively. The computer-controlled treadmill produced sinusoidal movement, which provided a mechanical perturbation to the subjects. While in quiet and perturbed stances, transcutaneous electrical stimulation with interferential current activated the agonist and antagonist muscles of the tibialis anterior and soleus muscles, which resulted in the co-contraction of the antagonistic ankle muscles. The experimental results show that simultaneous stimulation of the ankle antagonist muscle group using the interferential electrical stimulation (1) increases the root mean square (RMS) of the center of pressure (CoP) while in the quiet stance, and (2) decreases the RMS of the CoP and increases knee movement while in the perturbed stance. These findings highlight the different effects of FES-induced ankle impedance while in quiet and perturbed stances, suggesting that a simple modulation of ankle impedance via FES may not stabilize the postural sway in the quiet stance but rather may disturb it.

Keywords: Novel mechanisms and actuation
Abstract: Although general purpose robotic manipulators are becoming more capable at manipulating various objects, their ability to manipulate millimeter-scale objects are usually limited. On the other hand, ultrasonic levitation devices have been shown to levitate a large range of small objects, from polystyrene balls to living organisms. By controlling the acoustic force fields, ultrasonic levitation devices can compensate for robot manipulator positioning uncertainty and control the grasping force exerted on the target object. The material agnostic nature of acoustic levitation devices and their ability to dexterously manipulate millimeter-scale objects make them appealing as a grasping mode for general purpose robots. In this work, we present an ultrasonic, contact-less manipulation device that can be attached to or picked up by any general purpose robotic arm, enabling millimeter-scale manipulation with little to no modification to the robot itself. This device is capable of performing the very first phase-controlled picking action on acoustically reflective surfaces. With the manipulator placed around the target object, the manipulator can grasp objects smaller in size than the robot's positioning uncertainty, trap the object to resist air currents during robot movement, and dexterously hold a small and fragile object, like a flower bud. Due to the contact-less nature of the ultrasound-based gripper, a camera positioned to look into the cylinder can inspect the object without occlusion, facilitating accurate visual feature extraction.

Keywords: Human-machine interfaces, Novel mechanisms and actuation, Surgical robotics - design
Abstract: In traditional minimally invasive surgery, a camera holding assistant is required to provide a view of the operation site for the surgeon. The surgeon's performance relies on the quality of this video footage, which is sensitive to surgeon-assistant communication errors and the assistant's fatigue or inattention. Thus, we propose a novel foot interface to give the surgeon direct control over a robotic camera holder to replace the human camera holding assistant. The foot interface should allow the surgeon to control the camera while their hands are occupied with the primary surgical task. It can control 4 degrees of freedom of the camera's pose at 2 speeds. A preliminary experiment consisting of representative endoscope control tasks was conducted to evaluate and compare the performance of the novel foot interface and a voice interface using Google Cloud Speech. The foot interface performed better than the voice system in average completion time and error rate, with lower cognitive burden. Participants found the foot interface intuitive with a natural mapping between the foot and the robotic arm movements. The command actuation process is simplified and optimized to reduce the physical strain to the same low level as the voice system based on the subjective assessment after each task. Subjective assessment showed that 95% of the participants thought the foot interface provided better control of the camera movement and 90% preferred using the foot interface over the voice interface.

Keywords: Novel mechanisms and actuation, Exoskeletons and prostheses - design, Wearable technologies
Abstract: In this paper, we present the modelling of a hyper-redundant mechanism for assisting fingers in achieving desired configurations. The device is composed of serially linked rotating rigid elements and is actuated by a cable-driven mechanism that can provide both flexion and extension. The modelling is based on the analysis of the static equilibrium configurations that the structure acquires, which are dependent on the design parameters of each segment and on the ratio of the tensions applied on the cables. We show how it is possible to obtain different configurations by properly shaping the geometry of each element. Such a relationship was theoretically investigated and experimentally validated to compare the actual configuration with the one predicted by the model. The prototype was designed so that its static equilibrium configuration approximates the geometry of a finger during flexion-extension. It was observed that the customised structure achieves a configuration more closely resembling the desired one than a non-customised actuator. By adapting the design of each element and properly adjusting the ratio of tensions, one is able to compute the final configuration achieved by the actuator so that it can approximate the desired posture.

Keywords: Novel mechanisms and actuation, Biomechanics and rehabilitation, Wearable technologies
Abstract: Diagonal spring elements can render torque to any orthotic joint, and here we describe the theoretical framework for an ExoNET device that utilizes stacked spring elements as torque generators. Stacked spring elements act mathematically as basis functions, which can be simultaneously tuned to deliver any torque-angle relation. Here we outline the theory, demonstrate our initial developments in several example applications, and then describe the design considerations necessary to develop a functional prototype. We show several exemplary solutions: replicating the torque-angle profile of a single muscle (brachioradialis), two-joint gravity compensation for arm weight, error augmentation and limit push fields capable of providing forces for rehabilitation, and attractor torque fields that collectively pull the arm towards a desired position. This ExoNET system has the potential to be quickly and inexpensively constructed and easily configured by the end user or clinician for specific needs. It shifts control intelligence from the software to physical hardware, which is an efficient solution for neurorehabilitation, military, manual labor, and performance enhancement.

Keywords: Human-robot interaction, Novel mechanisms and actuation, Biomechanics and rehabilitation
Abstract: Assisted therapy in activities of daily life is a method in which through repeated movements, patients are able to recover gradually their range of motion. We have focused on re-learning feeding motor skills. Since patients with hemiparesis have some degree of trouble moving and present weakness on one side of their bodies, they need assistance to carry out rehabilitation tasks. In that sense, an intention-based sEMG-controlled 3 DOF upper limb exoskeleton was designed to assist them during therapy process and evaluate the progress of patients.

Keywords: Exoskeletons and prostheses - control, Biologically inspired systems - control, Wearable technologies
Abstract: Gait phase estimation plays an important role in the control of wearable robot. This paper focuses on the continuous gait phase estimation based on capacitive sensing signals for one transtibial amputee wearing robotic prosthesis. First, the capacitive sensing signals are used to record muscles' contraction and relaxation to detect two gait events (heel strike and toe off) as the start points (reset 0 rad) of one gait cycle. Secondly, continuous gait phase estimation is conducted based on the detected gait events (heel strike and toe off, respectively) with adaptive oscillators. In this study, the heel strike and toe off in each gait cycle can be detected with 100% accuracy based on linear discriminant analysis algorithm. The delay ratios for heel strike and toe off are less than 1% within one gait cycle. The root-mean-square errors of continuous gait phase estimation based on the different reset timings (heel strike and toe off) are 0.15 rad and 0.19 rad. The results show the feasibility to estimate the gait phase for amputee based on the residual muscles' capacitive signals.

Keywords: Brain-machine interfaces, Human-machine interfaces, Algorithms and machine learning
Abstract: This paper presents a mechatronic P300-based brain computer interface (BCI) for wheelchair control applications. A translucent visual stimulus panel (TVSP) is set up in front of the wheelchair to provide an intuitive P300 visual stimulus operation as well as to realize the see-through scene during operating wheelchairs. In this research, a micro projector is utilized to produce flickering visual stimuli on the display board which is 35cm away from the user. To improve the information transfer rate (ITR), a spatial filter based on Canonical Correlation Analysis (CCA) and Support Vector Machine (SVM) were also applied to this work to improve the performance of BCI classification. The result of experiments showed that the proposed BCI is with 88.2% in accuracy and 22.97 bits/min information transfer rate in average received from ten subjects. In ground truth experiments of practical trajectory tracking, the root mean squared error (RMSE) of P300 BCI are 12.11cm in “U” trajectory test.

Keywords: Algorithms and machine learning, Activity recognition and health monitoring, Wearable technologies
Abstract: This paper presents a novel algorithm using Lie group representation of position and orientation alongside a constrained extended Kalman filter (CEKF) to accurately estimate pelvis, thigh, and shank kinematics during walking using only three wearable inertial sensors. The algorithm iterates through the prediction update (kinematic equation), measurement update (pelvis height, zero velocity update, flat-floor assumption, and covariance limiter), and constraint update (formulation of hinged knee joints and ball-and-socket hip joints). The paper also describes a novel Lie group formulation of the assumptions implemented in the said measurement and constraint updates. Evaluation of the algorithm on nine healthy subjects who walked freely within a 4 times 4 m^2 room shows that the knee and hip joint angle root-mean-square errors (RMSEs) in the sagittal plane for free walking were 10.5 pm 2.8^circ and 9.7 pm 3.3^circ, respectively, while the correlation coefficients (CCs) were 0.89 pm 0.06 and 0.78 pm 0.09, respectively. The evaluation demonstrates a promising application of Lie group representation to inertial motion capture under reduced-sensor-count configuration, improving the estimates (ie{} joint angle RMSEs and CCs) for dynamic motion, and enabling better convergence for our non-linear biomechanical constraints. To further improve performance, additional information relating the pelvis and ankle kinematics is needed.

Keywords: Micro/nano robotics, Novel sensors, Human-robot interaction
Abstract: This paper proposes a novel real-time three dimensional (3D) position recognition method for an untethered magnetic capsule endoscope (CE) that is manipulated by an electromagnetic actuation (EMA) system. The developed localization system employs one receiving coil that is embedded in the CE and three external transmitting coils surrounding the spherical region of interesting (ROI) with diameter of 200mm. Based on the assumption that the orientation of the CE can be precisely controlled by the EMA system, the developed 3D localization could successfully recognize the position of the CE. By taking the advantage of the excellent penetration characteristics in human body, the effect of electromagnetic induction could track the CE that moves in the gastrointestinal (GI) tract. The resultant averaging position error is 2.1815mm with 4Hz of localization frequency. The proposed method will be further used clinically for both of disease location determination and position feedback control of CE in the future.

Keywords: Novel sensors, Soft robotics, Haptics
Abstract: Many mammals use whiskers to sense their surrounding environments. By whisking over an object to transduce tactile signals (forces or moments) to mechanoreceptors at the snout, then transmitting this data to the somatosensory cortex in the brain, they can extract features such as contact distance, texture and shape of the object. In this study, we propose a morphological computation method to localize the contact position/locate the contacted object by investigating the induced strain, measured by a strain gauge representing sensory nerves, along the length of a whisker. To accomplish this task, an artificial tapered whisker sensor was made from a soft material (silicon rubber) to provide flexibility, adaptability and more importantly sensitivity to a strain gauge. The first part of this paper introduces an analytical model of the proposed whisker based on elastic linear beam theory, and describes the unique correlations among the strain, contact distance, and whisker movements. The second part of this paper analyzes differences between the analytical model and a practical model to ascertain the optimum sensing conditions and any similarities in tactile behavior among the proposed whisker and a biological one. Finally, underpinned by experiment results, we argue that the proposed whisker can localize the contact point with optimum precision as long as its angular displacement is relatively small.

Keywords: Brain-machine interfaces, Algorithms and machine learning, Human-robot interaction
Abstract: Functional near-infrared spectroscopy (fNIRS) can help to diagnose specific diseases or distinguish motions for brain-computer interface (BCI). Also, repeating the same experiment can be uncomfortable for participants. It is difficult for researchers to obtain enough measurements to train the classification model sufficiently, which results in unstable classification accuracy. In this study, we investigated how to expand fNIRS data using the deep convolutional generative adversarial network (DCGAN) to improve classification accuracy and training stability. The data were measured using fNIRS during the finger tapping tasks, and then the proposed data augmentation method was used for generating artificial fNIRS datasets. These data were used to train a convolutional neural network (CNN), CNN model gave final classification accuracies for two classes (i. e. thumb finger tapping and little finger tapping). The AlexNet model, which is one famous model of the well-made model structure, was used as a CNN classifier model, the acquired accuracy for distinguishing the two types of tasks has been improved compared with the obtained accuracy using original data. This result suggests that the proposed deep learning model as a data generator is useful for improving classification performance.

Keywords: Novel sensors, Exoskeletons and prostheses - design, Novel mechanisms and actuation
Abstract: Scalable, high-density electronic skins (e-skins) are a desirable goal of tactile sensing. However, a realization of this goal has been elusive due to the trade-off between spatial and temporal resolution that current tactile sensors suffer from. Additionally, as tactile sensing grids become large, wiring becomes unmanageable, and there is a need for a wireless approach. In this work, a scalable, event-based, passive tactile sensing system is proposed that is based on radio-frequency identification (RFID) technology. An RFID-based tactile sensing hand is developed with 19 pressure sensing taxels. The taxels are read wirelessly using a single ‘hand-shaped’ RFID antenna. Each RFID tag is transformed into a pressure sensor by disconnecting the RFID chip from its antenna and embedding the chip and antenna into soft elastomer with an air gap introduced between the RFID chip and its antenna. When a pressure event occurs, the RFID chip contacts its antenna and receives power and communicates with the RFID reader. Thus, the sensor is transformed into a biomimetic event-based sensor, whose response is activated only when used. Further, this work demonstrates the feasibility of constructing event-based, passive sensing grids that can be read wirelessly. Future tactile sensing e-skins can utilize this approach to become scalable and dense, while retaining high temporal resolution. Moreover, this approach can be applied beyond tactile sensing, for the development of scalable and high-density sensors of any modality.

Keywords: Exoskeletons and prostheses - control, Human-robot interaction, Soft robotics
Abstract: Soft wearable robots, or exosuits, are a promising technology to assist the upper limb during daily life activities. So far, several exosuits have been proposed, some of which were successfully tested in open-loop control. However, though simple and robust, open-loop control unintuitive for use in daily life. Here, we closed the control loop on the human-robot interface of the Myoshirt. The Myoshirt is an upper limb exosuit that supports the shoulder during arm elevation. A direct force controller (DF) as well as an indirect force controller (IF) were implemented on the Myoshirt to assess their suitability for autonomously tracking human movement. In a preceding analysis, a DF controller with friction compensation (DFF) could be excluded, as linearly compensating friction aggravated the force tracking error in the ramp response (RMSE mean|sd: 32.75|10.95 N) in comparison to the DF response (27.61|9.38 N), while the IF controller showed substantially better tracking performance (17.12|0.99 N). In the subsequent movement tracking analysis including five participants (one female), the position tracking error and smoothness (med(RMSE); med(SPARC)) were similar with the DF (3.9°; -4.3) and IF (3.4°; -4.1) controllers and in an unpowered condition (3.7°; -4.2). However, the force tracking error and smoothness were substantially better when the IF controller (3.4 N; -4.5) was active than with the DF controller (10.4 N; -6.6). The Bode magnitude response indicated that both controllers obstructed the movement at higher frequencies, however with 0.78 Hz, the IF controller satisfied the bandwidth requirement for daily life assistance, while the DF controller (0.63 Hz) did not. It can be concluded that the IF controller is most suitable for daily life assistance with the Myoshirt.

Keywords: Soft robotics, Wearable technologies
Abstract: Pneumatic bending soft actuators have been used in various innovative applications. However, there is a need for efficient design principles and methods that can improve the control accuracy of soft robots. In this paper, we present a mathematical model of soft bending actuators based on optimization and design a cylindrical soft actuator with asymmetric inner cavity that will perform bending motion upon pressurization. We introduce neutral plane as a reference layer with constant length instead of using a strain-limiting layer on the soft actuator. Preliminary experimental results verify the modelling approach and demonstrate the effectiveness of the proposed method.

Keywords: Wearable technologies, Human-robot interaction
Abstract: Assisting a human motion with an orthosis requires the modeling of the human-orthosis system to estimate online human joint torque. The current paper proposes an assessment framework for: (i) the analysis of the sources of inaccuracies in joint torque estimation and (ii) the selection of the minimal dynamic model required to estimate joint torque. The proposed framework was validated with eight-healthy subjects wearing a knee joint orthosis and performing standardized sitting knee flexion/extension motions. As a result, the misalignment between the orthosis and wearer knee joint axis was quantified, as well as the influence of the kinematic and dynamic measurements in the knee joint torque estimation. Based on these results, a minimal dynamic model, of only two parameters, estimating 97.5% of the reference knee joint torque was selected.

Keywords: Haptics, Pathological assessment/diagnosis
Abstract: Access to primary health care is severely lacking in rural India. There are already some efforts where health practitioners interact with patients in rural areas through mobile applications that provide audio and visual communication. It could be useful to augment this interaction with some measure of tactile communication between the doctor and patient. One application of such tactile interaction is perceiving tissue stiffness increase due to ailments such as liver cirrhosis that is prevalent in rural India. The goal of this paper is to perform a preliminary computer simulation study to ascertain whether a pair of simple planar two-link manipulators forming a teleoperated master-slave system over mobile phone networks (these are ubiquitous even in rural India) are capable of "qualitatively" discerning liver stiffness change due to cirrhosis. It is shown that the error in stiffness estimation is reasonably small compared to the increase in liver stiffness due to cirrhosis and is acceptable to obtain a "qualitative" sense of increase in liver stiffness due to disease.

Keywords: Haptics, Human-machine interfaces, Biomechanics and rehabilitation
Abstract: Task-oriented training is crucial in the rehabilitation of patients with paresis to maximize the training effect of target tasks. To train the stroke patients with the closest form of activities in daily living, a new type of rehabilitation strategy is suggested that can train rotational tasks similar to active daily living tasks. The device is structured with a 3-RRR parallel manipulator to create three-dimensional rotations in a large workspace and interact with the user as an external object. The design parameter is optimized in two steps using the conventional global conditioning index and the global minimum active load to ensure controllability and force transmission through the device. The kinematic structure is verified with the motion capture system and encoders on motors. We envision using this compact table-top device emulating physical characteristics of daily living tasks to enhance the motor performance of stroke patients and their quality of lives.

Keywords: Soft robotics, Exoskeletons and prostheses - control, Human-robot interaction
Abstract: This paper presents quasi-static model-based control algorithms for controlling motion and force applied to the human hand by a soft robotic exoskeleton for physical assist and rehabilitation. Although soft robotics has shown promising for safe interaction with human body and reducing complexity in mechanisms for wearable medical and industrial applications; dynamics and control of physical interactions between the user and a soft wearable robot is still a challenge. A quasi-static model and the corresponding forward and inverse formulations were developed for modeling the physical interaction between the human finger and wearable soft robotic digits. A controller was derived for the joint position and interaction force control that determines the actuation pressure at each soft joint section of the robotic digit. The analytical model and controller were examined through simulation for tracking the desired joint trajectories and applying the interaction forces. The results showed the effectiveness of the model based control approach.

Keywords: Human-machine interfaces, Biomechanics and rehabilitation, Exoskeletons and prostheses - control
Abstract: Back-support (BS) exoskeletons aim at preventing or minimizing low-back pain in workers within occupational environments. Currently, there is no consensus on the optimal controller for BS exoskeletons. We propose a controller based on electromyography (EMG)-informed musculoskeletal modeling that estimates back muscle-tendon forces and moments. In this study, we validate an EMG-driven trunk model to estimate flexion-extension moments at the lumbar L5/S1 joint, during symmetric lifting tasks. In a first experimental session, ground reaction forces, subject kinematics and bipolar EMG activity from abdominal and lumbar muscles were recorded to estimate L5/S1 moments using both, inverse dynamics (ID) and EMG-driven modeling approaches. One subject performed squatting and stooping lifting tasks with three weight conditions (0, 5 and 15 kg). Correlation coefficients, R^2, between reference moments (from ID) and corresponding EMG-driven estimates ranged between 0.94 and 0.98, with root mean squared errors between 10.23 and 20.30 Nm. In a second experimental session, 4 high-density EMG (HDEMG) grids (256 channels) were used to generate high-fidelity topographical activation maps of thoracolumbar muscles during lifting tasks. These maps revealed that lifting objects using the squatting technique, underlay a shift of activation from caudal muscle trunk regions to cranial areas while lowering the weights. Muscle forces derived from EMG-driven modeling altogether with HDEMG activation maps are here proposed as a new framework to understand trunk neuromechanics during complex lifting tasks.

Keywords: Exoskeletons and prostheses - control, Activity recognition and health monitoring, Biomechanics and rehabilitation
Abstract: Existing strategies for controlling lower-limb robotic exoskeletons place different emphasis on the user’s intentions considered at various resolutions, from high-level goals (increase speed) to mid-level actions (increase stride length) to low-level joint behaviors (increase hip flexion). While sensors onboard the exoskeleton sense the human only indirectly, via the human-robot interface, they offer advantages over more direct methods in terms of the time required to don the device. In this study, exoskeleton users, both able-bodied and having spinal cord injury, were asked to perform changes in their intended gait speed. Onboard sensor measurements were used offline to test an intent identification algorithm based on the Mahalanobis distance. The algorithm’s goal is to identify an intent change and correctly classify its type, but not to realize that change via the exoskeleton. The algorithm correctly identified instances in which the user desired to walk faster or slower than the nominal speed in the device. For able-bodied subjects, the average delay between the known intent change and correct identification by the algorithm was 0.63 s. This delay for non-able-bodied subjects was 0.93 s on average. These proof-of-concept results show that intent identification based on the Mahalanobis distance is possible, while analysis of the approach suggests areas for further improvement.

Keywords: Biologically inspired systems - control, Biomechanics and rehabilitation, Biologically inspired systems - design
Abstract: The human hand is capable of successful, stable completion of amazingly complex and dexterous in-hand manipulation tasks through modulation of musculotendon stiffness. Although several studies have evaluated biomechanical stiffness for grasping and manipulation, no prior works have evaluated the effect of anatomical stiffness parameters on complex in-hand manipulation performance. In this work, we analyze the passive stiffness boundaries of a biomechanically accurate, tendon-driven human-like index finger to quantify the effect of stiffness parameter modulation on stability within the Cartesian workspace. The passive stiffness model shows that the greatest stiffness ellipsoid volume is aligned with efficient opposition of the anatomical thumb and bounds the conservatively stable region of the 3D workspace. Based on this model, we developed a biomechanically informed stiffness controller which increases the stable manipulation region and trajectory tracking performance within the reachable workspace. The result of this work is a method to quantify the stable manipulation region for tendon-driven systems that operate in a 3D space, enabling biomechanically informed mechanical and control design for stable, dexterous in-hand manipulation.

Keywords: Biologically inspired systems - control, Biologically inspired systems - design, Algorithms and machine learning
Abstract: Robotic vision introduces requirements for real-time processing of fast-varying, noisy information in a continuously changing environment. In a real-world environment, convenient assumptions, such as static camera systems and deep learning algorithms devouring high volumes of ideally slightly-varying data are hard to survive. Leveraging on recent studies on the neural connectome associated with eye movements, we designed a neuromorphic oculomotor controller and placed it at the heart of our in-house biomimetic robotic head prototype. The controller is unique in the sense that (1) all data are encoded and processed by a spiking neural network (SNN), and (2) by mimicking the associated brain areas' topology, the SNN required no training to operate. A biologically-constrained Hebbian learning further improved the SNN performance in tracking a moving target. Here, we report the tracking performance of the robotic head and show that the robotic eye kinematics are similar to those reported in human eye studies. This work contributes to our ongoing effort to develop energy-efficient neuromorphic SNN and harness their emerging intelligence to control biomimetic robots with versatility and robustness.

Keywords: Human-machine interfaces, Wearable technologies, Biologically inspired systems - control
Abstract: Accurate user intent recognition is vital to the success of achieving volitional control of rehabilitation robotics. Real-time ultrasound (US) imaging of skeletal muscle, or sonomyography, is an alternative noninvasive sensing mechanism for device control. The objective of this study was to evaluate sonomyography for continuous estimation of hip, knee and ankle joint moments during multiple ambulation tasks. Ten able-bodied subjects completed level, incline and decline walking while equipped with a portable US transducer on their anterior thigh. Multiple time-intensity features were extracted from US images of the knee extensor muscles collected during the three ambulation tasks. Hip, knee and ankle moments were continuously estimated by Gaussian process regression models in both fully subject-dependent and partially subject-independent frameworks. A two-way analysis of variance was completed to assess the effect of subject independence as well as joint level (hip/knee/ankle) on the moment estimation. Subject-dependent regression models resulted in the lowest error for estimation of hip, knee and ankle moment during all three ambulation tasks in comparison to partially subject-independent regression models (p<0.01). Remarkably, within the subject-dependent regression models there was no significant difference in the mean error of moment estimation when comparing across the three joints, with mean percent errors as low as 0.74%, 0.68%, and 3.02% for the hip, knee, and ankle, respectively. Despite only capturing sonomyographic features from the anterior thigh, this high-dimensional sensing data can be used to accurately estimate changes in both proximal and distal joint kinetics during varying ambulation tasks.

Keywords: Human-machine interfaces, Algorithms and machine learning, Wearable technologies
Abstract: People with high-level cervical spinal cord injury can have significant impairments in their ability to control their environment, including challenges operating a smartphone or navigating a power wheelchair. Smartphones are often controlled using a mouth stick and mobility is controlled using either a head array or sip-and-puff control system. A wearable system that allows continuous, multi-dimensional control of both smartphones and mobility based on the intuitive movement of cervically-innervated muscles with intact volitional activation could provide an improvement in quality of life for this population. Here I present a number of steps towards this including 1) a Bluetooth connected 8-channel, wearable electromyography sensor, 2) a neural network running on a smartphone that allows continuous two-dimensional control, and 3) rapid training of the neural network by calibrating to self-selected movements. This system was validated on two participants with a cervical spinal cord injury.

Keywords: Biological signal processing and identification, Algorithms and machine learning, Exoskeletons and prostheses - design
Abstract: Robotic rehabilitation has been proposed as a promising alternative in recovery after stroke, which still presents many challenges. We present here an initial approach to a progressive robot-assisted hand-motion therapy. Firstly, our system identifies finger motion patterns from electromyo- graphic (EMG) signals of 20 control volunteers during 5 hand exercises commonly used in rehabilitation. Secondly, the system characterizes 3 muscular condition levels, using muscular contraction strength, co-activation level and muscular activation level measurements. We compared the performance of Artificial Neural Networks (ANN), Support Vector Machines (SVM), Linear Discriminant Analysis Classifier (LDA) and k- Nearest Neighbor (k-NN) algorithms to classify the 5 gestures and 3 levels. Thirdly, each identified gesture and level was mapped into a spatial trajectory of an exoskeleton model, using a generalization of joint trajectories from subjects and a posterior interpolation. The statistical analysis between 36 different classifier architectures showed that a SVM classifier (cubic kernel) had the best performance to identify the 15 classes (F-score of 0.8 on average). Furthermore, the average correlation between the generated spatial trajectories and the tracked hand-motion was 0.89. In the future, the trajectories controlled by EMG signals could drive the exoskeleton for rehabilitation patients.

Keywords: Exoskeletons and prostheses - design, Technology assessment in human subjects/outcomes, Biomechanics and rehabilitation
Abstract: Simulators are often used in prosthesis research to evaluate new devices or characterize aspects of prosthesis use, so as to recruit participants without amputations. Simulators, in general, must locate the prosthesis somewhere other than where the intact biological limb exists. In this study, we compared two configurations of simulators for hand prostheses to determine which leads to more natural elbow and shoulder kinematics, and in turn, which is the more valid simulator. One configuration located the prosthesis in-line with the forearm, beyond the biological hand; the other located it beside the hand. We measured the kinematics of 12 non-amputee participants during three clinical tests of hand-arm dexterity, which were completed 1) using each simulator configuration with a body-powered Hosmer 5X hand prosthesis and 2) using the biological hand with a wrist brace. The beside-the-hand configuration resulted in kinematics that were more similar to those measured with the biological hand, particularly during the Box and Blocks Test, which involved navigating the prosthesis around obstacles. Therefore, we concluded that simulators with the beside-the-hand configuration are likely to better emulate the use of hand prostheses for activities involving a wide variety of arm movement. We suggest using this configuration in general, except when arm movement is of secondary importance and when this configuration would be obstructive, visually or otherwise.

Keywords: Wearable technologies, Soft robotics, Exoskeletons and prostheses - design
Abstract: Hand function rehabilitation is one of the most important aspects to improve the quality of life and independence for stroke survivors. Access to therapy is a key factor in one’s ability to perform rehabilitation exercises and recover motor function. The aim of this research is to show the viability of using a soft robotic glove to perform home-based rehabilitation of the hand function of stroke survivors. Soft wearable robotic devices are a promising approach to hand rehabilitation due to their lightweight, compliant, and low-cost design. The goal of this wearable technology is to reduce hyperexcitability of the flexor muscles through cyclic stretching of the fingers, active-assisted exercise, and task-oriented training. In this pilot study, four chronic stroke survivors were provided 300 repetitions of opening and closing their hand with the soft robotic glove over 6 training sessions. This high-intensity, high-repetition protocol was found to be safe and well-tolerated by all subjects. Results demonstrated average increases of 18.5° and 17.3° in the relaxed metacarpophalangeal and proximal interphalangeal joint angles, respectively, of the index finger and an average increase of 33 Newtons in grip strength. This work motivates further development of this home-based soft robotic rehabilitation device and lays the groundwork for a longer-term study.

Keywords: Soft robotics, Exoskeletons and prostheses - design, Biomechanics and rehabilitation
Abstract: Recovery from an upper limb musculoskeletal injury can take months or years, and often requires continuous visits to a therapist for specialized training and evaluation. Research has found that active assistance using mechatronic devices can significantly improve the quality and speed of recovery, especially when used for periods beyond a typical clinical appointment. However, current devices commonly use conventional actuation methods that are too heavy and rigid to be worn as a portable system outside of clinical settings. A recently discovered actuator made from twisting and coiling nylon thread (TCA) has the potential to improve wearable mechatronic designs due to its high power density, large strain, and inherent compliance. TCAs require heat to contract, which limits the actuation bandwidth due to the slow cooling rate, making it difficult to implement in wearable devices that must respond to voluntary motion. This study presents an active cooling method for TCAs to improve their frequency response, and evaluates their feasibility in a wrist orthosis based on real anatomical constraints. The frequency response was evaluated using a square wave input from an electrical power source during heating and an air pressure source when cooling, while the displacement was measured as the output at a constant load equivalent to the weight of an average wrist. The results show that the TCAs were able to respond to frequencies of up to 6.5 Hz, and also achieve full wrist range of motion at 0.55 Hz. This indicates that the proposed wrist orthosis could effectively assist the wrist throughout its entire range of motion and at the speeds required for voluntary motion using parallel TCAs.

Keywords: Exoskeletons and prostheses - design, Human-robot interaction, Wearable technologies
Abstract: Although the wrist plays a crucial role in performing interactive tasks, the wrist impedance has not been deeply studied so far. Moreover, all studies that aimed at estimating wrist impedance cannot be carried out in unstructured environments, thus limiting the gamut of possible research in this field. The major reasons underlying such limitations are the encumbrance and the non-portability of the robots used for this kind of applications. Within this work we presented the validation of a novel portable wrist exoskeleton to estimate the passive impedance of the wrist in Flexion-Extension (FE). To this aim, we enrolled 9 subjects whose wrist was passively moved around FE of textpm 20 deg by the device, characterized by 1 active degree of freedom. During the experimental sessions, we measured the joint torque and joint displacement, and we derived the angular velocities through numerical differentiation. Considering the wrist as a linear time-invariant second order mechanical system, we estimated impedance evaluating stiffness and damping coefficients, neglecting the contribution of the inertia due to the low speed of the movement applied by the robot. To this aim, we run a linear regression which allowed us to estimate stiffness and damping values. Passive stiffness estimated was equal to 1.794 textpm 0.514 Nm/rad for extension and 1.418 textpm 0.445 Nm/rad for flexion, confirming also the well-known difference between extension and flexion (p = 0.027). Passive damping was found equal to 2.054 textpm 1.202 Nms/rad for extension and 1.403 textpm 1.196 Nms/rad for flexion. These results are consistent with previous studies reported in literature, demonstrating the efficacy of our fully portable wearable wrist robot in estimating the wrist impedance around FE.

Keywords: Exoskeletons and prostheses - control, Exoskeletons and prostheses - design, Biologically inspired systems - design
Abstract: Lower limb prostheses have undergone significant developments in the last decades. However, there are several areas that have a scope for improvement through simplifications in the mechatronic design as well as in the control architecture. This paper focuses on the mechatronic design of a powered transtibial prosthesis and on the implementation of a control architecture, which is based on an adaptive frequency oscillator method that makes use of one inertial measurement unit. The control is capable of providing a positive push-off power to the prosthesis during level-ground walking and of adapting the response of the prosthesis to different walking speeds. The control architecture has been implemented and validated on a 3D printed prototype of a transtibial prosthesis. The experimental results show that the ankle joint can mimic the angle of a healthy subject with a root mean square error of 2.9 degrees and that the gait transitions are tracked within two gait cycles.

Keywords: Exoskeletons and prostheses - control, Biologically inspired systems - control
Abstract: This paper reports the development and preliminary validation of a bio-inspired controller using an adaptive oscillator and artificial primitives for a powered ankle-knee prosthesis, namely the CYBERLEGsPlusPlus Gamma-prosthesis. The proposed controller combines a feed-forward reference torque component with a feedback impedance-based torque component for the prosthetic knee and ankle joints. A treadmill walking experiment was conducted with a transfemoral amputee to assess the relevance of the feed-forward torque component, compared to a pure feedback control approach. Results showed that adding a feed-forward torque component to both prosthetic joints allowed a decrease of the feedback gains, and thus a more compliant behavior of the prosthesis. This combined condition was well-perceived by the pilot transfemoral amputee.

Keywords: Exoskeletons and prostheses - design, Biologically inspired systems - design, Wearable technologies
Abstract: Minimizing the mechanical motor output work or power required to create the desired actuator behavior in a quasi-static manner is a widely used, but incomplete method for actuator design. Here we use a dynamic electric and mechanical motor model combined with constrained output kinematics to find spring parameters for an active knee prosthesis using a series elastic actuator. This simulation was used to examine the landscape of motor torque tracking ability, motor electrical consumption, and the mechanical output of the system to determine spring stiffness ranges that should be further examined in the real device. Even though the task is negative net work, the actuator seems to benefit from having a elastic design, with a range of stiffness that improve energy consumption and torque tracking compared to a stiff device.

Keywords: Exoskeletons and prostheses - control, Biologically inspired systems - control, Exoskeletons and prostheses - design
Abstract: A main challenge in the development of active prosthetic knees is how to determine (estimate) the required motion of the missing joint/limb in line with the motion of the remaining biological ones. To do so, a motion planner is required. This study proposes a motion planner for active prosthetic knees. Two inputs (thigh angular velocities and angles obtained from IMU) are used to estimate the corresponding knee joint positions for walking at 0.6, 0.9, 1.2, 1.4 and 1.6 m/s. The motion planner does not require speed estimation, gait percent identification, or switching rules and estimates the outputs (knee joint positions) continuously. This is achieved through a functional (y=f(x)) approach. The strengths and limitations of the proposed motion planner are evaluated at different scenarios.

Keywords: Exoskeletons and prostheses - control, Biological signal processing and identification, Algorithms and machine learning
Abstract: Poly-articulated, myoelectric hand prostheses reproduce complex multi-degree of freedom movements, which are fundamental to effectively assist upper limb amputees in the execution of daily life activities. In this scenario, the control system consists in a pattern recognition algorithm translating the recorded electromyographic (EMG) activity into joint movements. However, the low decoding performance typically reached by the control system results in poor stability of the prosthetic device. In order to solve this issue, here we tested several state-of-the-art classifiers for decoding multi-joint hand movements from electromyographic recordings of arm muscles, collected from healthy subjects. Specifically, we tested: Non-Linear Logistic Regression (NLR), Regularized Least-Square, Artificial Neural Network, Support Vector Machine, and Linear Discriminant Analysis. We aimed at minimizing number of EMG electrodes (6 maximum) by optimizing each classifier in terms of the F1Score, and then compared the performance of the classifiers. We found that the NLR algorithm achieved the best results with only 3 EMG electrodes. The optimized algorithms were then adopted on three right arm amputees controlling a virtual hand. We obtained that algorithm’s performance was comparable with that obtained from healthy subjects. In particular, the NLR classifier achieved 99% correct classification for all the patients, indicating its potential effective use in prosthetic applications.

Keywords: Exoskeletons and prostheses - design, Human-centered design, Biologically inspired systems - design
Abstract: Passive ankle prostheses are the most common prosthetic device used by people with lower limb amputation. They use an energy recycling mechanism that mimics the behavior of the Achilles tendon and gastrocnemius muscle. The passive ankle prosthetic foot, which is usually made of carbon fiber, similarly stores energy during early to mid-stance by bending under the body’s weight. This energy is returned to the gait as the weight of the body is shifted off the device during terminal stance. However, most current standard of care passive ankle prostheses return the stored energy too early, dissipating most of the energy pushing the limb upward rather than propelling body forward due to the absence of any muscle control. In this study, we developed an ankle prosthetic timing module that can control the energy return timing of an ankle prostheses, replicating the function of the gastrocnemius muscle. Our ankle prosthetic timing module is designed to be installed onto most current standard of care ankle prostheses without major modifications. The timing module controls the energy release timing of the carbon fiber foot by holding it in deformation until later in the gait cycle, increasing the propulsion generated by the passive prosthetic device. This mechanism can potentially reduce the effort in and increase the function of walking for people with lower limb amputation.

Keywords: Technology assessment in human subjects/outcomes, Exoskeletons and prostheses - control, Biomechanics and rehabilitation
Abstract: The reduction of metabolic rate has become a pillar for quantifying the success of exoskeletons for performance augmentation, but how well can humans perceive these metabolic benefits? Measuring human perceptual ability in this context presents a unique challenge, since the stimulus of metabolic rate is not directly controllable. In this paper we introduce and compare two methods for addressing indirect stimuli: conventional method of constant stimuli (MOCS) with binning, and single presentation (SP). Both methods are based on a maximum likelihood estimation for a sigmoidal psychometric curve, under the assumption of a constant Weber fraction. Using Monte Carlo simulations, we conclude that while conventional MOCS with binning performs better for low actuation noise levels, single presentation strategies are better otherwise. We additionally present a pilot result for one subject using the single presentation strategy, which quantified the Just Noticeable Difference (JND) of changes to metabolic rate of 25.2 +/- 23.2.

Keywords: Technology assessment in human subjects/outcomes, Biomechanics and rehabilitation, Exoskeletons and prostheses - control
Abstract: Stroke is the main neurological condition causing disability worldwide. Physical therapy and robotic devices have been used in rehabilitation to recover lost locomotor functions. Despite the advantages of using robots in rehabilitation scenarios, some joints remain with alterations after therapy processes (e.g., the ankle joint). This paper presents a single case study of a patient with chronic stroke who participated in 18 sessions to assess the effects of T-FLEX in lower limb kinematics, spatiotemporal parameters, and muscular activity. To this end, each session consisted of two modalities: (1) 90-degree knee flexion, and (2) complete knee extension. The results showed improvement in the participant's spatiotemporal and kinematic parameters, as well as in the foot clearance during the swing phase. Regarding the muscular activity, the first sessions showed considerable increases related to the patient's inactivity. However, as the experiment proceeded, this value decreased as a consequence of the adaptation to the device. Regarding the electrical activity measured during each session, both muscles (i.e., gastrocnemius and tibialis anterior) tended to increase at the end-stage.

Keywords: Technology assessment in human subjects/outcomes, Biomechanics and rehabilitation
Abstract: Spinal Cord Injury (SCI) affects trunk control and determines altered or absent neuromuscular activity and sensory feedback below the lesioned spinal segment. The practice of sport or of any physical activity are key elements for improving the health and quality of life of people with SCI. Paralympic athletes overcome limits related to their injuries, achieving high neuromuscular control and coordination. Among the sports that have been adapted for people with disabilities, sit-skiing is a sport that requires good trunk control. However, there is a lack of instruments and protocols for its quantitative assessment. In this work we describe a robot-based protocol designed to assess trunk control and tested with two expert sit-skiers and eight unimpaired subjects. We used the robotic platform hunova® to evaluate both the active and the reactive component of trunk control through two different exercises. We investigated the strategy adopted by subjects to perform these exercises and the changes due to their repetition. All unimpaired subjects successfully completed the proposed protocol. The repetition of both exercises induced a learning process leading to differences in motor performance. Similar results could be observed also in the two athletes, whose performance was characterized by differences due to the severity of their lesion and their skiing skills.

Keywords: Algorithms and machine learning, Technology assessment in human subjects/outcomes, Biomechanics and rehabilitation
Abstract: The World Health Organization estimates that 15 million infants are born preterm every year. This is of concern because these infants have a significant chance of having neuromotor or cognitive developmental delays due to cerebral palsy or other developmental issues. Our long-term goal is to determine the roles emotion and movement play in the diagnosis of atypical infants. In this paper, we examine how automated emotion assessment can classify typically and atypically developing infants. We compare a custom supervised machine learning algorithm that utilizes individual and grouped facial features for infant emotion classification with a state-of-the-art neural network. Our results show that only three concavity features are needed for the concavity algorithm, and the custom algorithm performed with relatively similar accuracy to the neural network. Automatic sentiment labels used in tandem with infant movement kinematics would be further investigated to determine if emotion and movement are interdependent and predictive of an infant’s neurodevelopmental delay in disorders such as cerebral palsy.

Keywords: Biomechanics and rehabilitation, Biological signal processing and identification, Exoskeletons and prostheses - control
Abstract: Robotic-assisted systems have been playing a key role in improving and speeding up motor recovery during stroke rehabilitation therapies. This paper presents an approach to determine velocity patterns based on the analysis of the EMG muscular condition of the hand. To this purpose, we conducted an experimental protocol with 18 subjects participating as vol- unteers, with the aim of acquiring EMG signals for three levels of the muscular condition: non-fatigue, transition-to-fatigue, and fatigue. Artificial Neural Networks (ANN) were trained to identify the aforementioned muscular condition levels, while a Sugeno-Type Fuzzy Inference system was used to determine the velocity based on the output of the ANN classifiers. Results indicate the proposed approach can be used for the accurate modulation of pinch-grip therapies according to the muscular condition. These are promising results towards the development of EMG-driven robotic-assistance rehabilitation therapies for stroke patients.

Keywords: Biological signal processing and identification, Biomechanics and rehabilitation, Algorithms and machine learning
Abstract: This paper investigates an ultrasound (US) imaging-based methodology to assess the contraction levels of plantar flexors quantitatively. Echogenicity derived from US imaging at different anatomical depths, including both lateral gastrocnemius (LGS) and soleus (SOL) muscles, is used for the prediction of the volitional isometric plantar flexion moment. Synchronous measurements, including a plantar flexion torque signal, a surface electromyography (sEMG) signal, and US imaging of both LGS and SOL muscles, are collected. Four feature sets, including sole sEMG, sole LGS echogenicity, sole SOL echogenicity, and their fusion, are used to train a Gaussian process regression (GPR) model and predict plantar flexion torque. The experimental results on four non-disabled participants show that the torque prediction accuracy is improved significantly by using the LGS or SOL echogenicity signal than using the sEMG signal. However, there is no significant improvement by using the fused feature compared to sole LGS or SOL echogenicity. The findings imply that using US imaging-derived signals improves the accuracy of predicting volitional effort on human plantar flexors. Potentially, US imaging can be used as a new sensing modality to measure or predict human lower limb motion intent in clinical rehabilitation devices.

Keywords: Biomechanics and rehabilitation, Algorithms and machine learning, Biological signal processing and identification
Abstract: Although cognitive engagement (CE) is crucial for motor learning, it remains underutilized in rehabilitation robots, partly because its assessment currently relies on subjective and gross measurements taken intermittently. Here, we propose an end-to-end computational framework that assesses CE in near real-time, using electroencephalography (EEG) signals as objective measurements. The framework consists of i) a deep convolutional neural network (CNN) that extracts task-discriminative spatiotemporal EEG features to predict the level of CE for two classes- cognitively engaged vs. disengaged; and ii) a novel sliding window method that predicts continuous levels of CE in short time intervals. We evaluated our framework on 8 healthy subjects using an in-house Go/No-Go experiment that adapted its gameplay parameters to induce cognitive fatigue. The proposed CNN had an average leave-one-subject-out accuracy of 88.19%. The CE prediction correlated well with a commonly used behavioral metric based on self-reports taken every 5 minutes (ρ=0.93). Our results objectify CE measurement in near real-time and pave the way for using CE as a rehabilitation parameter for tailoring robotic therapy to each patient's needs and skills.

Keywords: Biomechanics and rehabilitation, Algorithms and machine learning, Biological signal processing and identification
Abstract: Mounting evidence suggests that adaptation is a crucial mechanism for rehabilitation robots in promoting motor learning. Yet, it is commonly based on robot-derived movement kinematics, which is a rather subjective measurement of performance, especially in the presence of a sensorimotor impairment. Here, we propose a deep convolutional neural network (CNN) that uses electroencephalography (EEG) as an objective measurement of two kinematics components that are typically used to assess motor learning and thereby adaptation: i) the intent to initiate a goal-directed movement, and ii) the reaction time (RT) for that movement. We evaluated our CNN on data acquired from an in-house experiment where 12 healthy subjects moved a rehabilitation robotic arm in four directions on a plane, in response to visual stimuli. Our CNN achieved average test accuracies of 80.08% and 79.82% in a binary classification of the intent (intent vs. no intent) and RT (slow vs. fast), respectively. Our results demonstrate how individual movement components implicated in distinct types of motor learning can be predicted from synchronized EEG data acquired before the start of the movement. Our approach can, therefore, inform robotic adaptation in real-time and has the potential to further improve one's ability to perform the rehabilitation task.

Keywords: Biological signal processing and identification, Biomechanics and rehabilitation, Algorithms and machine learning
Abstract: This paper investigates the use of ultrasound imaging to characterize tremor during a grasping motion. Ultrasound images were collected from three human participants including an able-bodied participant, a patient with Parkinson’s disease, and a patient with essential tremor. Each human participant was instructed to grasp and hold objects with three different masses in a vertical upright position with an ultrasound probe strapped to their forearm while seated. The images were processed using an ultrasound speckle tracking algorithm to measure muscle strain during the grasping and holding motion. Analysis of the computed strain values showed marked differences in the strain peaks and frequencies between able-bodied participant and the patients with tremor. The detected frequencies depict how the strain measurement changes during the grasping and holding motion. The frequency for tremor participants fall within accepted frequency ranges for Parkinson’s Disease and Essential Tremor, and thus can be representative of the actual tremor frequency.

Keywords: Pathological assessment/diagnosis, Technology assessment in human subjects/outcomes, Biomechanics and rehabilitation
Abstract: Spasticity is a serious problem that affects a large percentage of patients affected by neurological disorders including stroke, multiple sclerosis, cerebral palsy, traumatic injury. Unfortunately, no reliable assessment method of spasticity is currently available. This is a critical missing tool for a rational, assessment-based approach to rehabilitation of patients with neurological impairment. In this paper we propose a robot-based assessment method, focused on the wrist joint, that could be readily integrated with robot treatment protocols. The method consists of the delivery of small and quick rotational disturbances to one of the Degrees of Freedom (DoF) of the wrist (Flexion/Extension DoF). With the aim of validating its consistency and reliability, the method was preliminary tested on a population of healthy subjects, providing a promising initial operational basis.

Keywords: Biologically inspired systems - control, Wearable technologies, Human-machine interfaces
Abstract: People coordinate their eyes and head during typical activities of daily life. Subjects with poor head-eye coordination find it difficult to make eye contact with others, limiting their social interaction. Inspired by the natural vestibulo-ocular reflex (VOR) during head-eye coordination, we propose a gaze control interface to drive a robotic neck brace. This system will enable those with poor head control to perform head movements using their eyes. The control interface of this system is triggered by the user's pupil motion in a virtual field of view (FOV) to produce head movements. The results indicate that the control interface is capable of rotating the head towards the target within 5degree of difference of the desired head angle. This study shows that the proposed bio-inspired control promotes natural eye-head coordination, which may be potentially useful for individuals with poor or limited neck control.

Keywords: Human-robot interaction, Biologically inspired systems - control
Abstract: Ten right-handed male subjects turned a crank (radius 10.29 cm) in two directions at three constant instructed speeds (fast, medium, very slow) with visual speed feedback. They completed 23 trials at each speed. With the hand constrained to move in a circle, non-zero forces against the constraint were measured. Assuming a plausible mathematical model of interactive dynamics, the peripheral neuromechanics could be ‘subtracted’, revealing an underlying motion that reflected neural control. We called this data-driven construct the zero-force trajectory. The observed zero-force trajectory was approximately elliptical. Its major axis, estimated by the principal eigenvector of the covariance matrix, differed significantly for the two movement directions. As peripheral neuromuscular compliance (i.e. low mechanical impedance) mitigates the consequences of imperfect execution, the required precision of motion commands is reduced. An oscillatory zero-force trajectory that leads hand motion suffices to produce circular hand motions. Due to non-isotropic peripheral dynamics, that lead differs between degrees of freedom, resulting in an elliptical zero-force trajectory. The ellipses’ orientations differ with direction of rotation, as observed in the experimental data. As elliptical motion is generated by two non-colinear sinusoids with non-zero phase difference, these results support the hypothesis that humans simplify this constrained-motion task by exploiting primitive dynamic actions, oscillations and impedance.

Keywords: Biologically inspired systems - control, Human-machine interfaces, Wearable technologies
Abstract: Objective: Current limitations in Electromyography (EMG)-driven Neuromusculoskeletal (NMS) modeling for control of wearable robotics are the requirement of both Motion Capture for both an indoor system and numerous EMG electrodes. These limitations make the technology unsuitable for amputees with only proximal muscles, who need optimal prosthetic device control during everyday activities. Therefore, we developed a novel Machine Learning (ML)-driven NMS model able to predict lower limb joint torque only from wearable sensors than can be embedded in a prosthetic device. Methods: After the NMS model calibration of a single healthy subject (OpenSim-CEINMS software), an additional ML layer was added to the model to replace the MoCap-derived dependent variables with estimations obtained only from wearable sensors. An on-line open-loop Forward Dynamic (FD) simulation of the knee joint is computed and torque trajectories are compared to experimental ones. Results: Estimations of the novel ML-driven MS model were comparable with experimental knee joint torque during typical locomotion tasks. Accuracy results were comparable to standard EMG-driven MS models and errors are below the threshold of NRMSD < 0.30 recognized in literature. Conclusions: We developed the first concept of completely wearable and subject-specific EMG-driven NMS model control for lower limb prostheses. The possibility to use this NMS model for FD simulations and the estimation of torque reference control avoids the use of current heuristic and overly complex standard controllers for lower limb prostheses. This research, in fact, represents a key step for the definition of a novel human-machine interface able to create a seamless interconnection between human native control and future wearable robotics.

Keywords: Exoskeletons and prostheses - control, Biological signal processing and identification, Biologically inspired systems - control
Abstract: Spinal Cord Injury (SCI) affects thousands of peo- ple worldwide every year. SCI patients have disrupted muscle recruitment and are more predisposed to other complications. To recover or enhance lower limbs functions, conventional rehabilitation programs are typically used. More recently, conventional programs have been combined with robot-assisted training. Electromyography (EMG) activity is generally used to record the electrical activity of the muscles, which in turn can be used to control robotic assistive devices as orthoses, prostheses and exoskeletons. In this sense, surface EMG can be used as input to myoelectric control but presents some limitations such as myoelectric crosstalk, as well as the influence of motion artefacts, and electromagnetic noise. EMG can also be recorded using intramuscular detection systems, which allows the detection of electric potentials closer to the muscle fibres and the recording of EMG activity from deeper muscles. This paper evaluates the quality of intramuscular EMG recordings com- pared to surface EMG signals, as a preliminary step to control EMG-driven exoskeletons. Seven healthy subjects performed submaximal knee and ankle flexion/extension movements with and without the use of a lower limb exoskeleton. Intramuscular recordings presented early muscle activation detecting times, which is a very important feature in real-time control, and good signal-to-noise ratio values, showing the potential of these biosignals as reliable input measures to control exoskeletons for rehabilitation purposes.

Keywords: Biomechanics and rehabilitation, Exoskeletons and prostheses - design
Abstract: Running-specific prosthesis (RSP) have been around for over 20 years and have had a huge impact on the capabilities of para athletics. Measuring the performance of the RSP and the amputee in the real-world has not been performed extensively. This paper describes a system using a camera and an inertial measurement unit that, when attached to a RSP, estimates the deflection of the running blade and its ground reaction force (GRF). The independent system was attached to the RSP of one amputee during jumping tests and the results from the system were verified with a force plate and a high speed camera at the side. Although the performance of the system can still be optimized, the system shows great potential to become a training tool to monitor RSP. The deflection was calculated with a mean error (SD) of 1.2 mm (0.7 mm) and the GRF with 15 N (263 N). The system is small-sized, light-weight, cheap and fully adaptable to an amputee’s prosthetic setup with little adjustments.

Keywords: Wearable technologies
Abstract: The main objective of the study included in this paper is to analyze the possible advantages of the incorporation of the upper-limb exoskeleton developed within the ExIF project in the performance of a repetitive task of industrial scope. For this purpose, the times used to carry out the movements inherent to the task have been studied, as well as the changes in oxygen consumption when performing the activity under two conditions (with and without the robotic device). After analyzing the results obtained, it can be stated that carrying out the activity with the help of the exoskeleton means a reduction in oxygen consumption by the user, which implies that there is a decrease in the level of intensity of the task to be performed.

Keywords: Exoskeletons and prostheses - control, Biologically inspired systems - control, Wearable technologies
Abstract: Environment recognition systems can facilitate the predictive control of lower-limb exoskeletons and prostheses by recognizing the oncoming walking environment prior to physical interactions. While many environment recognition systems have been developed using different wearable technology and classification algorithms, their relative operational performances have not been evaluated. Motivated to determine the state-of-the-science and propose future directions for research innovation, we conducted an extensive comparative analysis of the wearable technology, training datasets, and classification algorithms used for vision-based environment recognition. The advantages and drawbacks of different wearable cameras and training datasets were reviewed. Environment recognition systems using pattern recognition, machine learning, and convolutional neural networks for image classification were compared. We evaluated the performances of different deep learning networks using a novel balanced metric called “NetScore”, which considers the image classification accuracy, and computational and memory storage requirements. Based on our analysis, future research in environment recognition systems for lower-limb exoskeletons and prostheses should consider developing 1) efficient deep convolutional neural networks for onboard classification, and 2) large-scale open-source datasets for training and comparing image classification algorithms from different researchers.

Keywords: Wearable technologies, Algorithms and machine learning
Abstract: The security of private information is becoming the bedrock of an increasingly digitized society. While the users are flooded with passwords and PINs, these gold-standard explicit authentications are becoming less popular and valuable. Recent biometric-based authentication methods, such as facial or finger recognition, are getting popular due to their higher accuracy. However, these hard-biometric-based systems require dedicated devices with powerful sensors and authentication models, which are often limited to most of the market wearables. Still, market wearables are collecting various private information of a user and are becoming an integral part of life: accessing cars, bank accounts, etc. Therefore, time demands a burden-free implicit authentication mechanism for wearables using the less-informative soft-biometric data that are easily obtainable from modern market wearables. In this work, we present a context-dependent soft-biometric-based authentication system for wearables devices using heart rate, gait, and breathing audio signals. From our detailed analysis using the ``leave-one-out'' validation, we find that a lighter k-Nearest Neighbor (k-NN) model with k = 2 can obtain an average accuracy of 0.93 pm 0.06, F_1 score 0.93 pm 0.03, and {em false positive rate} (FPR) below 0.08 at 50% level of confidence, which shows the promise of this work.

Keywords: Biomechanics and rehabilitation, Algorithms and machine learning, Wearable technologies
Abstract: Fatigue deteriorates temporary motor functions in individuals which often leads to performance drop of occupational workers, poor postural control of patients, and falls in elderly persons. Fatigue management and prevention of its adverse effects significantly depend on timely detection of fatigue. Advent of novel wearable sensor technologies enabled real time data collection and gait monitoring. Using IMU data, we propose a new method to detect fatigue with sole acceleration data from ankle. This method uses computationally-light Statistical Process Control (SPC) which does not require big data to set the algorithm and is also robust to noise. Instead of using simple gait parameters that represent intermittent gait data, we used the acceleration profiles of the whole gait cycles to detect fatigue. Workers were recruited to perform walking, loading, and un-loading tasks and their baseline and fatigued gait patterns were recorded. We explored cumulative and non-cumulative statistical process control methods for online monitoring of fatigue using the recorded data. Results from the non-cumulative method showed dominant changes in the gait pattern after participants were fatigued. We envision this method can be used to detect fatigue in real time in occupational workers, patients with ambulatory disorders, and elderly population.

Keywords: Biomechanics and rehabilitation, Human-centered design, Exoskeletons and prostheses - design
Abstract: Falls are a major cause of injury among people over 65, being the hip fracture the most common consequence. The morbidities and even the high mortality associated lead to considerable social and economic costs surrounding the rehabilitation process. Hip fracture rehabilitation is long and complex and related to advanced age. The relentless growth of the older population in developed countries has made technology pursues new challenges to improve the quality of life and independence of this population. One of the newest applications has been the development of robotic platforms to rehabilitate musculoskeletal diseases. Within this context, this paper describes the design and development of a robotic platform, called SWalker, for hip fracture rehabilitation. The main goal of the platform is to promote early weight-bearing and mobilization. The paper describes the technical and functional validation of the system with a limited number of elderly patients. The positive results coming from the users suggest that the system is reliable, and it is ready for future clinical validation.

Keywords: Human-centered design, Wearable technologies, Exoskeletons and prostheses - design
Abstract: Human-centered design of assistive technology aims to achieve high functional benefits while considering user opinions to increase device usability and promote acceptance. Considering the needs and opinions of all real-world user groups, including not only the target end-users with disabilities but also their caregivers (e.g. family members, friends, or paid helpers) can improve the design of assistive devices. In this study, the perspectives and performances of 15 participants (3 clinical, 12 non-clinical) placed in a caregiver role to prepare and don a wearable robotic hand orthosis on a wheelchair-bound mock end-user were investigated. Through the use of eye tracking combined with a retrospective think-aloud protocol, usability issues of the orthosis were identified and addressed in a design iteration. These design changes positively influenced the device usability by means of fewer problems occurring during the unsupervised setup procedure, as well as a significantly lower total procedure time (p = 0.02) in a second evaluation round. The caregiver perspective proved useful to improve the robotic hand orthosis design and usability, paving the way for unsupervised administration and use of the assistive device.

Keywords: Wearable technologies, Novel mechanisms and actuation, Human-centered design
Abstract: This paper presents the design and preliminary testing of an electronically-controlled cold-gas-thruster intended to potentially be used as a backpack-worn device for fall prevention for individuals at fall risk. The device is comprised of a high-pressure air tank, a pilot-operated valve, and a nozzle, which are employed together to create a thrust intended to correct a state of imbalance. The authors first employed modeling and simulation of the thruster-based system to assess feasibility of such a device, and based on these results designed and constructed a device prototype. The prototype was then tested on a rocking block that was used as an experimental model of a standing human subject. Experiments were conducted to assess the angular response of the rocking block starting from various initial tilt angles, both with and without the use of the cold gas thruster. For the experiments with the thruster, the valve was energized at the moment of release, thereby activating the cold-gas thruster, the thrust of which was directed to oppose the “fall.” A comparison between the two cases indicates that the fall prevention prototype increased the region of block stability by more than a factor of three (from a tilt angle of approximately 7 to 26 deg).

Keywords: Exoskeletons and prostheses - design, Biomechanics and rehabilitation, Human-centered design
Abstract: Simulations of walking biomechanics offer a tool for optimizing prosthesis performance while including estimates of the effects of the prosthesis on the rest of the body. We have previously used this technique to optimize the output behaviors of a generalized prosthesis model in the sagittal plane. In this paper, we present the design of a prototype prosthesis testbed for validating generalized prosthesis model predictive simulation results with experimental feedback. Design specifications are generated from simulated prosthesis dynamics and comparison with existing powered prostheses. A complete mechatronic system design based on these specifications is presented. The design consists of two sub-systems: the ankle-foot prosthesis and a wearable off-board actuation and control system. The overall system is designed to function as a validation tool for prosthesis simulation experiments generally, and to provide experimental feedback to the simulation-based prosthesis design loop.

Keywords: Exoskeletons and prostheses - control, Human-centered design, Wearable technologies
Abstract: The daily use of advanced wearable robotic devices for the assistance of people with locomotor disabilities is still facing clear limitations in usability and acceptance (e.g. cost, complexity, and inability to maintain balance). In most devices, the correct selection and initiation of pre-defined functions and activities (e.g. walking and stair ascent-descent) rely on the user's input and constant interpretation of the environment, which results in a substantial cognitive workload. In this study, a novel environment recognition and parameterization system that uses depth camera images is proposed as a potential assistant in the control of powered lower-limb exoskeletons. The feasibility of an online shared-control approach between the user and the system was assessed in two specific use-cases of lower-limb exoskeletons: Mode selection assistance and dynamic step adaptation. In a sequence of realistic daily life tasks, the assistance provided by the proposed system achieved an error below 10% with a loop frequency up to 400 Hz in terms of parameterizing the environment, and reduced the mean overall workload, measured with the Raw Task Load Index, by 19% in a group of seven neurologically intact subjects. In conclusion, an assistive environment recognition and parameterization system shows potential to reduce the cognitive workload on the user, and thereby positively influence device usability.

Keywords: Human-robot interaction, Human-machine interfaces, Exoskeletons and prostheses - control
Abstract: Robot-assisted therapy is a promising field in terms of motor rehabilitation for stroke victims. However, the interaction between the user and the robot must be done in a way to ensure patient safety and treatment effectiveness. Several human-robot interaction controls have been proposed in order to meet these requirements. However, testing and validating these controls remain a challenge, requiring physical contact between a user and a robot, which can put the user at risk as well as it can require a reasonable amount of time and resources for preparing such tests. This work proposes the development of a human-robot interaction model based on the Leg6Dof9Musc neuromusculoskeletal model from OpenSim and a simulation environment developed in MATLAB, in order to test the interaction controls developed for the rehabilitation of lower limbs. An impedance control was tested for the swing phase during gait movement and the results proved that the system is feasible and useful, with flexibility, and savings in time and resources, in addition to preventing users and robots from being put at risk during tests of interaction controls.

Keywords: Brain-machine interfaces, Human-machine interfaces, Human-robot interaction
Abstract: In this paper, the most effective stimulation durations for the purpose of brain-computer interface (BCI) with functional near-infrared spectroscopy (fNIRS) is investigated. To do so, this study investigates the influence of varying stimulation duration on the hemodynamic response (HR) signal in the primary visual cortex of the human brain. fNIRS is used for the measurement of HRs. For brain stimulation, a flickering checkerboard task is utilized, and the HR signals are acquired. Three different stimulation durations, i.e., 1, 3, and 5 secs were used in this study, and a total of 5 subjects participated in the experiment. From the results of the study, it is concluded that among the tested stimulation durations, the HR signal rises with the increase of the stimulation duration. Also, it was found that even the smallest stimulation duration, i.e., 1 sec, can generate a detectable activation in the visual cortex of the brain, which can well serve the purpose of BCI.

Keywords: Exoskeletons and prostheses - control, Human-machine interfaces, Human-robot interaction
Abstract: Robot prosthetic hands intend to replicate one's lost abilities through intuitive control. So far, control methods that rely heavily on the human input such as Electromyographic (EMG) and Electroneurographic (ENG) signals have been predominantly studied. However, these methods face issues such as lack of robustness resulting in abandonment of this technology by the users. There is a need for a paradigm shift in the robot prosthetic hand control methods. With this regard, we propose an end-to-end learning of control policy for a robot prosthetic hand through reinforcement learning. Imitation learning has been fostered to help with the sparse reward setting in the hard-to-explore state-space of the problem. The results in simulation show the feasibility of successfully learning an end-to-end policy for grasping objects by robot prosthetic hands, potentially increasing robustness for grasp control of future robot prosthetic hands.

Keywords: Human-machine interfaces, Exoskeletons and prostheses - design, Human-robot interaction
Abstract: Sensory feedback is highly desirable in upper limb prostheses as well as in human robot interaction and other human machine interfaces. Bone conduction as sensory feedback interface is a recently studied approach showing promising properties. A combination of different feedback information is often necessary for prosthetic grasping, thus multiple feedback channels are required for effective sensory feedback. The use of multiple bone conduction stimulation sites simultaneously has not yet been studied. In this paper, the psychometric evaluation of multiple stimulation sites on the physiologically given bony landmarks on the elbow is investigated. The proposed approach is evaluated on humansubject experiments with six able-bodied subjects and one subject with transradial amputation. Vibrotactile transducers are placed on the bony landmarks of the elbow to determine the identification rate of each stimulation point separately as well as the identification rate of the number of active stimulation points for different frequencies. The outcomes show high identification rates for a frequency range from 100 to 750 Hz whilst performance deteriorates to at chance level at higher frequencies. A decreasing performance in identifying the number of active stimulation sites for an increasing number of simultaneous active transducers was observed. The obtained good performance in location identification suggests that information can be encoded via the location of the stimulation.

Keywords: Exoskeletons and prostheses - control, Human-machine interfaces, Algorithms and machine learning
Abstract: Limb position and load are factors that negatively affect myoelectric pattern recognition. While these effects have been studied separately, it remains unclear how limb position affects pattern recognition when limb load changes, and vice versa. Understanding this relationship will aid the development of algorithms and training protocols that allow prosthesis users to reliably grab and manipulate objects in various limb positions. We evaluated the effects of limb position and external load on offline pattern recognition accuracy in fourteen intact limb subjects and five below-elbow amputee subjects. Three clinically viable training methods were used to determine if training protocols can mitigate the negative effects of limb position and load. We found that limb position and load effects are independent in intact limb subjects, but are dependent in amputee subjects based on how the controller is trained. Although the advanced training protocols did not eliminate the limb position and load effects in most cases, they were able to reduce the effects and improve performance without increasing training time.

Keywords: Exoskeletons and prostheses - design, Wearable technologies, Novel mechanisms and actuation
Abstract: Exoskeletons for industrial use are emerging to mitigate the risk of occurrence of work-related musculoskeletal diseases, such as low-back pain due to repetitive lifting activities. Yet, several factors are limiting the practical use of these devices in real-case scenarios, such as their weight and comfort. In this work, we describe iWear, an underactuated multi-joint exoskeleton for lifting applications. The key design feature of the device is the single actuation unit based on a Series-Elastic Actuation architecture, which drives three output joints, i.e. the left and right hips and the trunk. Moreover, passive degrees of freedom are designed to ensure the self-alignment of the robot with the user’s trunk. Along with the presentation of the device, we report the performance of the actuation unit in torque control, which demonstrates to be fast and for the target application. The underactuated design of iWear holds promise for the design of wearable robots.

Keywords: Exoskeletons and prostheses - control, Algorithms and machine learning, Biological signal processing and identification
Abstract: Powered prostheses and orthoses are expected to significantly improve quality of life of individuals with motor impairment or limb amputation. Despite the advances in the mechatronic design of powered prostheses and orthoses, their use is not yet widespread because of limitations in their control schemes. Methods that use Electromyography (EMG) measurements to estimate a physical signal related to joint kinematics or kinetics to be used as a control variable are promising as these methods may enable robust control schemes for high-dimensional powered prostheses and orthoses.

In this study, we propose a method to estimate joint torques about two degrees of freedom of the wrist joint. The developed method uses regression ensemble of bagged trees, and is applied to isometric tasks under multiple wrist postures. The proposed regression model estimated Flexion/Extension and Radial/Ulnar Deviation torques with coefficient of determination equal to 0.94 and 0.88, respectively, when training and testing at the neutral wrist posture.

The analysis showed that even using only one wrist posture for training, good results can be obtained in estimating joint torque under different postures. The outcome of this study suggests that a single-posture calibration combined with the proposed estimation method can be used as a paradigm to derive a linear multi-dimensional control signal that could be suitable for EMG control of prosthetic or exoskeletal devices.

Keywords: Biologically inspired systems - control, Algorithms and machine learning, Human-robot interaction
Abstract: Conventional methods for motion control and path planning in robots are nowhere near as reactive and flexible as in nature. Brains solve navigation using place cells – neurons that provide a cognitive representation of a specific environment. Neural techniques for path planning in 2D have been developed for years, however, to allow their application for robotic tasks beyond locomotion, a transmission to 3D is required. We present an implementation for path planning via a propagating wavefront on 3D environments. The algorithm operates on a Spiking Neural Network of excitatory place cells structured as a grid. A wavefront travelling through the network is initiated by activating the goal place cell. The wave strengthens synapses in the direction of the propagation using STDP, as a synaptic learning rule. By interpreting the synaptic weights as a vector field, a path can be derived from any place cell, reached by the wave, to the destination. We demonstrate, using a neural simulator, that our algorithm works well on maps with multiple obstacles. Our method allows fast simulation and query times and we expect to considerably improve the network creation time by using dedicated hardware, allowing massive parallelism. Our algorithm applies bio-inspired techniques and is especially interesting for Human-robot interaction, which requires reactive flexible motion planning.

Keywords: Surgical robotics - design, Surgical robotics - control, Surgical navigation and localization
Abstract: Steerable needle technology has the promise of improving outcomes by enhancing the accuracy of different therapies and biopsies, as they can be steered to a target location around obstacles. Achieving small radius of curvature and being able to control both radius of curvature and tip travel are of paramount importance in steerable needles to accomplish the increase in efficacy of the medical procedures. In this paper, we present a new class of the steerable needles, which we call waterjet-directed steerable needles, where the underlying principle is to first control the direction of tissue fracture with waterjet, after which the needle will follow during subsequent insertion. In this paper, the direction of the tissue fracture is controlled by an angled waterjet nozzle and control of the water velocity, and then the flexible Nitinol needle follows. It is shown that by changing the velocity of waterjet and thus depth of cut, radius of curvature can be controlled. A discrete-step kinematic model is used to model the motion of the waterjet steerable needle. This model consist of two parts: (1) the mechanics-based model predicts the cut-depth of waterjet in soft tissue based on soft tissue properties, waterjet diameter, and water exit velocity, and (2) a discrete-step kinematic unicycle model of the steerable needle travel. Path planning is accomplished through a genetic algorithm, and the efficacy of waterjet steerable needle is tested for different paths. The key finding of the paper is that the radius of curvature of the waterjet steerable needle can be controlled by a fixed waterjet tip angle and varying water exit velocity to control the depth of cut.

Keywords: Haptics, Algorithms and machine learning, Wearable technologies
Abstract: There is a need to gather rich, real-time tactile information to enhance prosthetic hand performance during object manipulation. To that end, a highly stretchable liquid metal tactile sensor was designed, manufactured, and integrated into the fingertip of an i-limb Ultra prosthetic hand. With this novel tactile sensor, the feasibility of real-time slip detection and prevention of a grasped object was demonstrated. Furthermore, this liquid metal tactile sensor was used to distinguish between five different surface patterns with high accuracy using three different classification algorithms: K Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forest (RF). The K-nearest neighbors (KNN) classifier produced the highest classification accuracy of 98% to distinguish between five different surface textures. Taken together, this novel prosthetic fingertip tactile sensor has the potential to improve grasp control and object manipulation operations for upper limb amputees. Additionally, this paper documents the first time that a liquid metal tactile sensor has been used to distinguish between different surface features, to the best knowledge of the authors.

Keywords: Biologically inspired systems - control, Haptics, Soft robotics
Abstract: The ability of animals to interact with complex dynamics is unmatched in robots. Especially important to the interaction performances is the online adaptation of body dynamics, which can be modelled as an impedance behaviour. However, variable impedance control still continues to be a challenge in the current control frameworks due to the difficulties of retaining stability when adapting the controller gains. The fractal impedance controller has recently been proposed to solve this issue. However, it still has limitations such as sudden jumps in force when it starts to converge to the desired position and the lack of a force feedback loop. In this manuscript, two improvements are made to the control framework to solve these limitations. The force discontinuity has been addressed introducing a modulation of the impedance via a virtual antagonist that modulates the output force. The force tracking has been modelled after the parallel force/position controller architecture. In contrast to traditional methods, the fractal impedance controller enables the implementation of a search algorithm on the force feedback to adapt its behaviour to the external environment instead of on relying on a priori knowledge of the external dynamics. Preliminary simulation results presented in this paper show the feasibility of the proposed approach, and it allows to evaluate the trade-off that needs to be made when relying on the proposed controller for interaction. In conclusion, the proposed method mimics the behaviour of an agonist/antagonist system adapting to unknown external dynamics, and it may find application in computational neuroscience, haptics, and interaction control.

Keywords: Biological signal processing and identification, Biomechanics and rehabilitation
Abstract: Despite the utility of musculoskeletal dynamics modeling, there exists no safe, noninvasive method of measuring in vivo muscle output force in real time. In this paper, we demonstrate that muscle deformation constitutes a promising, yet unexplored class of signals from which to infer such forces. Through a preliminary case study of the elbow joint, in which we examine simultaneous flexion force, surface electromyography (sEMG), and ultrasound imaging data during isometric contraction, we provide evidence that even simple measures of deformation (including cross-sectional area and thickness variation in the brachioradialis muscle) correlate well with elbow output torque to an extent comparable with standard sEMG activation measures. We then show that these simple signals, as well as the overall brachioradialis contour, can be tracked over time via optical flow techniques, enabling the use of these signals in real-time applications (e.g., assistive device control), as well as larger-scale study of deformation signals without necessitating manual annotation. To enable such future work, all modeling and tracking software described in this paper, as well as all raw and processed data, have been made available on SimTK as part of the OpenArm project (https://simtk.org/projects/openarm) for general research use.

Keywords: Biomechanics and rehabilitation
Abstract: This paper focuses on the development and validation of a generic musculoskeletal model of an osseointegrated transfemoral amputee. The model has been developed using OpenSim with the final goal of obtaining a competent tool to study and better understand the biomechanics of osseointegrated transfemoral amputees. The model has been validated on the experimental data obtained on one osseointegrated transfemoral amputee during level ground walking.

Keywords: Biologically inspired systems - control, Algorithms and machine learning
Abstract: The brain outperforms computer architectures in aspects of energy efficiency, robustness and adaptivity. Brain computations are modeled in silico with spiking neural networks and neuromorphic hardware. Recently, three-factor synaptic plasticity rules approximating backpropagation have been derived. Suited to neuromorphic hardware, these rules can learn online with asynchronous updates. In this paper, we present Continuous Random Backpropagation (cRBP), a continuous version of Event-Driven Random Backpropagation. This learning rule performs comparably to state-of-the-art rules on the DvsGesture dataset. We additionally show that the accuracy can be significantly increased with a simple attention mechanism. This mechanism provides translation invariance at low computational cost compared to convolutions by exploiting event stream sparsity. Subsequently, we integrate cRBP in a real robotic setup, where a gripper grasps objects according to the detected visual affordances. In this setup, visual information is actively sensed by a Dynamic Vision Sensor (DVS) mounted on a robotic head performing microsaccadic eye movements. Our results suggest that advances in neuromorphic technology and plasticity rules enable the development of learning robots operating at high speed and low power.

Keywords: Biological signal processing and identification
Abstract: Movements are associated with a beta (15-30 Hz) power decrease during movement preparation and a rebound after its termination. Motor learning and practice are characterized by an increase of the rebound. However, the introduction of a sensorimotor perturbation negatively affects such measure. In this preliminary study, we investigated whether two learning processes affect beta electroencephalographic activity: (a) learning to perform a center-out reaching task by moving a passive robotic arm, and (b) learning to reach the same targets in the presence of a velocity-dependent force field. Seven subjects were exposed to the (a), then to the (b) condition. Results revealed that beta power rebound and its latency increased with practice with the passive manipulandum. When the force field was applied, beta power rebounded and its latency immediately dropped, but they started to increase again with the prolonged exposure to the predictable perturbations. These results suggest that increased rebound in beta power reflects cortical activity elicited by the correct prediction of movement outcomes.

Keywords: Haptics, Soft robotics, Surgical robotics - control
Abstract: Recent advances in soft robotics are utilized to solve challenges in endoscopy, such as maneuverability, flexibility, and the structural stiffness required to deliver enough force during endoscopic surgical procedure. Other major challenge is the lack of haptic feedback from the tool end-effector to the surgeon. Current clinical practice in minimally invasive intervention requires an assistant to control the camera since the surgeon is preoccupied with task at hand, creating a indirect control procedure for maneuvering the endoscope. For the soft robotic endoscope, we implemented a haptic feedback interface along with a novel control method to concurrently tackle these challenges. The user of the developed system can visualize the planned 2D insertion path and steer the endoscope module accordingly using an inertial measurement unit mounted on a head-band. Furthermore, five different haptic feedback methods (three kinesthetic and two vibrotactile) were compared in term of user accuracy while steering the endoscope along a planned path. The results show that the user's accuracy using kinesthetic and vibrotactile feedback were comparable, however, participants of this study find vibrotactile feedback approach more preferable for its intuitiveness and comfort.

Keywords: Biomechanics and rehabilitation, Exoskeletons and prostheses - control, Haptics
Abstract: Research on neurorehabilitation has emphasized that somatosensory information about the interaction with the environment during physical training is crucial to provoke brain plasticity. Despite this, only a small number of robotic devices provide haptic rendering of the virtual environment during neurorehabilitation exercises, the majority with simple structures. However, to provide realistic haptic rendering while supporting neurological patients to perform motor tasks, a transparent robot with several degrees of freedom is needed. In this study, we employed Disturbance Observers to achieve high transparency and fine haptic capabilities on the six DoF exoskeleton ARMin. We incorporated arm weight compensation to reduce the excessive physical effort required to move against gravity, promoting movement performance and directing the participants' effort to the interaction with the environment. The effect of haptic rendering and its interaction with arm weight compensation were evaluated with six healthy participants. The task consisted of inverting a virtual pendulum and keeping it inverted. We found that haptic rendering of the pendulum dynamics affects the movement strategy the participants follow, i.e., they covered a significantly larger workspace with the end-effector at a significantly higher speed, and required moderate physical effort. The inclusion of arm weight support increased task performance and reduced participants' effort, while it did not change the movement strategy. Our results suggest that haptic rendering, together with arm weight support, are potential interventions to enhance neurorehabilitation due to the added somatosensory information during motor training.

Keywords: Exoskeletons and prostheses - control, Biomechanics and rehabilitation, Human-robot interaction
Abstract: Vibrations can be used to convey positional or sensory information to prosthetic users. However, for the feedback to convey information consistently, fine grained adjustments on a daily basis is required. This paper investigates whether vibration dissipation through the muscle can be tracked using EMG with the aim of providing reliable long term sensory feedback. The results of this study showed that the magnitude of vibration artefacts can be measured using EMG and used to create a dissipation trend. This trend varies between participants but shows consistency when measured across multiple days. This novel way of measuring vibration dissipation can be used as a basis of adaptive sensory control in future prosthesis studies.

Keywords: Exoskeletons and prostheses - control, Algorithms and machine learning, Biological signal processing and identification
Abstract: Hand amputations can dramatically affect the capabilities of a person. Machine learning is often applied to Surface Electromyography (sEMG) to control dexterous prosthetic hands. However, it can be affected by low robustness in real life conditions, mainly due to data variability depending on various factors (such as the position of the limb, of the electrodes or the characteristics of the subject). This paper aims at improving the understanding of sEMG for prosthesis control introducing the type of hand movement as a variable that influences classification performance in both intact subjects and hand amputees. Five hand amputees and five matched intact subjects were selected from the publicly available NinaPro database. The subjects were recorded while repeating 40 hand movements. Movement classification was performed on the sEMG data with a window-based approach (concatenating several signal features) and a Random Forest classifier. The results show that some hand movements are classified significantly better than others (p<0.001) and there is a correspondence in how well the same hand movements are classified in intact subjects and hand amputees. This work leads to advancements in the domain, highlighting the importance of the acquisition protocol for sEMG studies and suggesting that specific movements can lead to better performance for the control of prosthetic hands.

Keywords: Biomechanics and rehabilitation, Biological signal processing and identification
Abstract: Mirror therapy, the rehabilitative technique that uses a mirroring illusion to induce cognitive responses, could benefit by performing the task in a virtual world. This research is aimed to build an effective virtual reality environment with gesture-level hand tracking and determine its viability when compared to traditional mirror therapy. A total of 24 healthy participants were recruited to perform a block stacking task in both the physical and virtual environments, with and without mirroring, using their dominant arms. Surface electromyography (sEMG) was used to measure activation of the flexor carpi radialis (FCR), extensor digitorum (ED), biceps brachii (BB) and triceps brachii (TB) in both arms. The virtual environment was shown to decrease task completion speed due to the added difficulty of control in that environment. Significant differences in muscle activity between conditions for the non-dominant arm were not achieved; however, there were significant differences in muscle activity between conditions for the dominant arm. Difficulty and weight discrepancies between the physical and virtual tasks may have contributed to this significance.

Keywords: Biomechanics and rehabilitation, Activity recognition and health monitoring, Wearable technologies
Abstract: Wearable sensing technology is proving useful for promoting health and fitness for the general public and athletes, however few are tailored to people with movement impairments. For devices targeting home-based rehabilitation, it is crucial to have robust and non-obtrusive sensors capable of measuring activity for long periods of time outside of a laboratory environment. Studies focusing on continuous monitoring of arm activity during daily life over weeks or months only use IMU sensors or accelerometers at the wrist, and do not capture multi-segment kinematics. In this paper we present the development of the ArmTracker, a fully, portable non-obtrusive and wearable IMU-based motion capture system that can measure arm and torso kinematics for long-periods of time during daily life. We also present the results of a preliminary evaluation of the prototype carried out with one unimpaired subject and one subject with Becker muscular dystrophy (BMD). Both subjects were asked to wear the ArmTracker device during daily life for 7 hours. We carried out an exploratory graphical analysis with the measured data using three types of movement quality metrics: 1) Range of Motion, 2) Functional Workspace Distribution and, 3) Accelerometry. Results provided an insightful view on the motor function capabilities and limitations of the two subjects. Arm activity of the subject with BMD showed low variability in terms of joint angles and hand positions over the workspace with a clear preference of using his hands in front and below the shoulder height. Arm movements were slower compared with the unimpaired subject and with a slight preference of using the dominant arm. We plan to extend this study with measurements over several days and weeks to capture representative data on arm activity.

Keywords: Exoskeletons and prostheses - control, Technology assessment in human subjects/outcomes, Human-machine interfaces
Abstract: Controlling a complex upper limb prosthesis, akin to a healthy arm, is still an open challenge due to the inadequate number of inputs available to amputees. Designs have therefore largely focused on a limited number of controllable degrees of freedom, developing a complex hand and grasp functionality rather than the wrist. We introduce a novel 3 degree of freedom wrist trajectory control which takes advantage of joint coordination that aims to vastly simplify its use in a prosthetic device. We demonstrate its efficacy through a series of tasks performed by participants in a virtual environment. Trajectory control enables users to complete tasks faster with a more intuitive interface without additional body compensation, while featuring better cosmesis when compared to sequential and simultaneous control.

Keywords: Exoskeletons and prostheses - design, Biomechanics and rehabilitation, Wearable technologies
Abstract: Nowadays, exoskeletons are widely used as assistive tools during gait rehabilitation, implementing strategies such as gravity compensation and metabolic cost reduction. However, these devices often require the use of motors and controllers. Our research is driven by the need for a simple, inexpensive and customizable device, a passive exotendon able to provide a desired torque profile to multiple joints. In a simulation model study, we asked whether it was possible to use diagonal tension elements, which act mathematically as basis functions, in order to reproduce the torque field generated by a healthy person during walking. This would allow us to create a wearable exotendon system made of passive elastic elements able to deliver this torque field to the lower extremity of the patient during gait rehabilitation. Our results showed that it is indeed possible to create a passive torque field from elastic elements that approximates the muscular torque demand for the walking cycle. This represents a starting point for the design of a passive exotendon capable of providing assistive torques to patients with motor deficits, thus reducing the metabolic cost of walking.

Keywords: Biomechanics and rehabilitation, Wearable technologies, Exoskeletons and prostheses - design
Abstract: Models of the human arm may help to estimate design parameters like peak torque and power of wearable assistive devices by predicting required forces to compensate for motor control impairments. This work focuses on the idea of compensating hypermetria (overshoot)—a motor control deficit that may occur in neurodegenerative diseases—by a simple assistive device. As musculoskeletal dynamics play an important role in the interaction between an assistive device and the neuro-musculoskeletal system, we hypothesized that their consideration in the model might influence the predicted design parameters. To test this, we simulated two-degree-of-freedom point-to-point arm movements. By introducing inconsistent neuronal control parameters, we induced hypermetria. We im- plemented mechanical and low-level assistive torque strategies in simulation which lead to a reduction of hypermetria. We found that–depending on the type of assistance–the predicted torques and powers can differ by more than a factor of 10 between musculoskeletal and torque-driven arm models. We conclude that the magnitude of torque and power required to reduce hypermetria by simple wearable assistive devices may be significantly underestimated if muscle-tendon characteristics are not considered.

Keywords: Biomechanics and rehabilitation, Exoskeletons and prostheses - control, Algorithms and machine learning
Abstract: The primary goal of an assist-as-needed (AAN) controller is to maximize subjects’ active participation during motor training tasks while allowing moderate tracking errors to encourage human learning of a target movement. Impedance control is typically employed by AAN controllers to create a compliant force-field around the desired motion trajectory. To accommodate different individuals with varying motor abilities, most of the existing AAN controllers require extensive manual tuning of the control parameters, resulting in a tedious and time-consuming process. In this paper, we propose a reinforcement learning AAN controller that can autonomously reshape the force-field in real-time based on subjects’ training performances. The use of action-dependent heuristic dynamic programming enables a model-free implementation of the proposed controller. To experimentally validate the controller, a group of healthy individuals participated in a gait training session wherein they were asked to learn a modified gait pattern with the help of a powered ankle-foot orthosis. Results indicated the potential of the proposed control strategy for robot-assisted gait training.

Keywords: Algorithms and machine learning, Exoskeletons and prostheses - control, Biomechanics and rehabilitation
Abstract: Machine learning (ML) algorithms present an opportunity to estimate joint kinetics using a limited set of mechanical sensors. These estimates could be used as a continuous reference signal for exoskeleton control, able to modulate exoskeleton assistance in real-world environments. In this study, sagittal plane biological hip torque during level ground, incline and decline walking was calculated using inverse dynamics of human subject data. Subsequently, this torque was estimated using neural network (NN) and XGBoost ML models. Model inputs consisted solely of mechanical sensor data onboard a robotic hip exoskeleton. These results were compared to a baseline method of estimating hip torque as the mean torque profile during ambulation. On average across conditions, the NN and XGBoost models estimated biological hip torque with an RMSE of 0.116±0.015 and 0.108±0.011 Nm/kg, respectively, which was significantly less than the baseline estimation that had an RMSE of 0.300±0.145 Nm/kg (p<0.05). Fitting the baseline method to ambulation mode specific data significantly reduced overall RMSE by 59.3%; however, the ML models were still significantly better than the baseline method (p<0.05). These results show that machine learning algorithms can estimate biological hip torque using only mechanical sensors onboard a hip exoskeleton better than simply using an average torque profile. This suggests that these estimation models could be suitable for modulating exoskeleton assistance. Additionally, no evidence suggested the need to train separate ML models for each ambulation mode as estimation RMSE was not significantly different across unified and separated ML models.

Keywords: Wearable technologies, Biomechanics and rehabilitation, Exoskeletons and prostheses - design
Abstract: The inability to stand and walk is one of the major consequences of spinal cord injury (SCI). Wearable lower limb exoskeletons are emerging as a promising solution to restore mobility after SCI. The ABLE exoskeleton is a powered-knee exoskeleton intended for patients with low SCI, who preserve motor function at the hip. This paper presents a comparative study to investigate the biomechanical effects of adding a hip structure that connects two individual powered-knee exoskeletons, and constrains hip abduction-adduction and internal-external rotation. We carried out a series of tests with an adult man with a motor complete SCI (injury level T11) walking with two independent knee-exoskeletons one in each leg (ABLE Knee) and with the knee-exoskeletons connected to the hip structure (ABLE Hip). The participant was asked to walk with both configurations while we recorded body motion, actuator current, and foot contact pressure. We observed significantly larger deviations in terms of hip abduction-adduction, internal-external rotation and pelvic obliquity with the ABLE Knee compared to the ABLE Hip. In addition, walking with the ABLE Hip significantly increased the range of motion of knee and hip flexion-extension. However, these improvements in joint kinematics did not translate to significant differences in step length or walking speed, when comparing the two configurations. In conclusion, these results suggest that adding a hip structure that limits undesirable hip rotations can improve walking performance for people with SCI.

Keywords: Biomechanics and rehabilitation, Biologically inspired systems - control, Activity recognition and health monitoring
Abstract: Stepping is a vital strategy for legged systems to recover balance while they interact with their environment. This work presents normalized criteria for the analysis of legged system stability based on balanced and steppable regions. These criteria are applied to both a comprehensive region-based analysis of the contact transitions during a normal human step cycle and a comparative analysis of humanoid and human systems. In this work, the steppable, single support balanced, and double support balanced regions are evaluated for a human model with full-order system dynamics and joint actuation limits based on biomechanical models of maximum voluntary joint torques. The normalized regions are also compared with the capturability of an equivalent reduced-order model in order to demonstrate the role of steppability as a general extension of and complementary concept to capturability. In the comparative analysis of the humanoid and human systems, the normalized regions are used to directly compare the capability for balance recovery in both systems. The proposed approaches can be generalized beyond sagittal planar walking to analyze balance stability in any multi-contact scenario where stepping or step-like contact transitions can occur.

Keywords: Technology assessment in human subjects/outcomes, Exoskeletons and prostheses - control
Abstract: Exoskeletons could compete with active prostheses as effective aids to reduce the increased metabolic demands faced by lower-limb amputees during locomotion. However, little evidence of their efficacy with amputees has been provided so far. In this paper, a portable hip exoskeleton has been tested with seven healthy subjects and two transfemoral amputees, with the final goal to verify whether a hip flexion-extension assistance could be effective in reducing the metabolic cost of walking. The metabolic power of the participants was estimated through indirect calorimetry during alternated repetitions of three treadmill-based walking conditions: without the exoskeleton (NoExo), with the exoskeleton in zero-torque mode (ExoTM) and with the exoskeleton providing hip flexion-extension assistance (ExoAM). The results showed that the exoskeleton reduced the net metabolic power of the two amputees in ExoAM with respect to NoExo, by 5.0% and 3.4%. With healthy subjects, a 5.5±3.1% average reduction in the metabolic power was observed during ExoAM compared to ExoTM (differences were not statistically significant), whereas ExoAM required 3.9±3.0% higher metabolic power than NoExo (differences were not statistically significant). These results provide initial evidence of the potential of exoskeletal technologies for assisting lower-limb amputees, thereby paving the way for further experimentations.

Keywords: Exoskeletons and prostheses - control
Abstract: Locomotion is essential for a person's ability to function in society. When an individual has a condition that limits locomotion, such as lower limb amputation, the performance of a prosthetic often determines the quality of life an individual regains. In recent years, powered prosthetic devices have shown nearly identical replication for human leg motion on non-compliant terrains. However, they still face numerous functional deficits such as increased metabolic cost and instability for walking on surfaces of varying compliance and complexity. This paper proposes joint angles of the biological leg are uniquely altered by surface compliance regardless of a subject's individual walking pattern. These differences are then displayed and quantified as a way to better characterize able-bodied walking compensation typical with three common terrains: sand, grass, and gravel. This study also collects data outdoors using IMU sensors and is not limited by lab setup and conditions. These results are important since a better understanding of joint angle kinematics on varying terrains could enable the formulation of advanced controllers for current prosthetic devices allowing them to anticipate surface changes and adapt accordingly.

Keywords: Human-machine interfaces, Biomechanics and rehabilitation, Exoskeletons and prostheses - control
Abstract: Individuals with neuromuscular injuries may fully benefit from wearable robots if a new class of wearable technologies is devised to assist complex movements seamlessly in everyday tasks. Among the most important tasks are locomotion activities. Current human-machine interfaces (HMI) are challenged in enabling assistance across wide ranges of locomoting tasks. Electromyography (EMG) and computational modelling can be used to establish an interface with the neuromuscular system. We propose an HMI based on EMG-driven musculoskeletal modelling that estimates biological joint torques in real-time and uses a percentage of these to dynamically control exoskeleton-generated torques in a task-independent manner, i.e. no need to classify locomotion modes. Proof of principle results on one subject showed that this approach could reduce EMGs during exoskeleton-assisted even ground locomotion compared to transparent mode (i.e. zero impedance). Importantly, results showed that a substantial portion of the biological ankle joint torque needed to walk was transferred from the human to the exoskeleton. That is, while the total human-exoskeleton ankle joint was always similar between assisted and zero-impedance modes, the ratio between exoskeleton-generated and human-generated torque varied substantially, with human-generated torques being dynamically compensated by the exoskeleton during assisted mode. This is a first step towards natural, continuous assistance in a large variety of movements.

Keywords: Algorithms and machine learning, Exoskeletons and prostheses - control, Biomechanics and rehabilitation
Abstract: Osteoarthritis knee braces require large brace-leg interface forces to stabilize and unload the joint during weight bearing. Actively removing support while the user is in a non-weight-bearing state could improve the comfort of the brace but requires the timing of weight-bearing states to be known. This study presents two artificial neural networks (ANNs) for predicting time-to-next heelstrike during walking using only data from two accelerometers placed on the thigh and shank. One ANN used teacher forcing and the other did not. Walking data were collected from 10 subjects and leave-one-subject-out cross-validation was used to evaluate the performance of the two models. Input features for the ANNs included tibial and femoral accelerations, concatenated into one array. The teacher forcing ANN and the non-teacher forcing ANN performed equally well (RMSE = 0.23 +/- 0.13s for the non-teacher forcing ANN, RMSE = 0.27 +/- 0.08s for the teacher forcing ANN). The performances of the models were worse than those of previously published studies that predicted heelstrike events. Accelerations were insufficient for an ANN to predict time-to-next heelstrike during walking.

Keywords: Exoskeletons and prostheses - control, Human-machine interfaces
Abstract: For robotic ankle exoskeletons, the choice of controller and the many associated controller parameters (e.g., magnitude or timing of assistance) directly affect the user's performance, comfort, and biomechanics. Some studies have proposed tuning such controllers for individual users using optimization techniques to minimize (or maximize) a physiological objective, such as metabolic energy expenditure. One drawback to this approach is that it necessitates having one predefined, measurable objective function that is relevant in all situations. In reality, people may prioritize many different less-quantifiable metrics, such as stability, comfort, pain, or perceived effort, in any given situation. In this paper, we demonstrate a method of exoskeleton controller customization that is based on subject preference, which likely encodes many of these subjective cost functions at once. In this pilot study, two subjects self-tuned their own controller parameters in two dimensions, by directly manipulating the magnitude and timing of bilateral ankle exoskeleton assistance using a touch screen tablet interface. Subjects exhibited high repeatability in identifying their preferred exoskeleton parameters at each of three walking speeds (coefficient of variation between 0.9-14.5%). Subjects' preferences differed from each other, which highlights the importance of individual customization. The results from this pilot study will inform future experiments that incorporate rigorous measurement of subject preference into the control of robotic ankle exoskeletons. This research has the potential to provide more complete insights into important elements of exoskeleton control that will be useful outside of the steady-state, laboratory environment.

Keywords: Algorithms and machine learning, Surgical navigation and localization, Technology assessment in human subjects/outcomes
Abstract: Objective: Accurate needle tip placement for biopsy procedures is difficult to achieve with low-fidelity imaging systems. Conventionally, surgeons while performing biopsies rely on ultrasound images and intuitive feeling about needle tissue interaction forces to confirm target location. Currently, robotic assistance for biopsy uses only the position parameter to address localization challenges. In the present work, in addition to a robot’s position sense, we propose to integrate needle-tissue force parameter. This force model presents a new way to built an intelligent robot that can identify tissue properties during needle probing/biopsy. Methods: A standard experiment was setup that consist of a force sensor, a linear stage, a biopsy needle, and synthetic tissues. During the experiment, needle penetrates through synthetic tissues, a set of data (force and distance) was acquired and manually labeled. A recurrent neural network (RNN) based Long-Short Term Memory (LSTM) model was trained with the data to estimate the various classes air, skin/fibrous tissue, puncture, and hard tissue). Result: The trained model is able to distinguish between the three synthetic materials. Intuitively, this model mimics human perceptions of force during a handheld needle penetration. Conclusion: The constrained experimental setup helps us present a proof of concept for using deep learning models for tissue classification. Significance: Tissue classification is the first step towards solving the more difficult problem of developing a robotic device capable of precise event detection of tissue transitions.

Keywords: Pathological assessment/diagnosis, Biomechanics and rehabilitation
Abstract: Two primary mechatronic protocols have been deployed across several studies for measuring the impact of flexion synergy (loss of independent joint control) on reaching function. The measurement of reaching work area and reaching distance were both conducted as part of a previous clinical trial with the latter being the study’s primary outcome. While evidence exists supporting the reliability and validity of reaching work area, it may be time-inefficient and/or too fatiguing to be viable in clinical practice. Reaching distance was therefore employed in the clinical trial as the primary outcome as a shortened method despite established validity. In this retrospective study, pre- and post-intervention data of both metrics from the clinical trial are used to evaluate concurrent validity of the reaching distance metric and to compare its responsiveness to change with that of reaching work area.

Keywords: Surgical robotics - control, Human-machine interfaces, Human-robot interaction
Abstract: Recently, there has been a strong emergence of teleoperated minimally invasive surgery due to the potential benefits for both, the surgeon’s performance and outcomes for the patient. However, due to high acquisition costs, no widespread use has been achieved so far. The use of low-cost components can help to address this issue. In this paper, a commercial handheld virtual reality controller as an unconventional and versatile approach for the input device, is presented and investigated. Assessment of the input device is done by performing an experimental study with 24 participants on two different fine motoric tasks. Additionally, motion scaling is investigated by using different transmission ratios for the user’s movements. The large and unrestricted range of motion of the presented input device showed to be a promising alternative to conventional input devices. However, the tasks showed room for improvement regarding the controller’s ergonomics and usability. For further comparative investigations, conventional input devices and mimetic input devices need to be considered.

Keywords: Surgical robotics - design, Surgical robotics - control
Abstract: This paper introduces a snake-like continuum robot design intended for minimally invasive surgery (MIS) in which a robotic sheath is designed to control an end effector with enhanced dexterity. The structure of the eccentric tube robot (ETR) is a development and extension of the current concentric tube robot (CTR) design. By introducing a non-extensible outer sheath, three pre-curved superelastic nitinol (NiTi) tubes are held together eccentrically inside the sheath. The articulation of the ETR is achieved by rotating the individual tubes at their proximal ends, and the curvatures interact to determine the robot’s final shape. A kinematic model based on Cosserat rod theory is derived considering the eccentric arrangement and the geometrical constraint imposed by the outer sheath. The kinematic model is validated through simulations and experiments for a single-stage ETR. Experimental results show that, in the worst case, the mean tip error corresponds to 5% of the length of the sheath.

Keywords: Pathological assessment/diagnosis, Biomechanics and rehabilitation, Biological signal processing and identification
Abstract: Spasticity, defined as velocity-dependent resistance to passive stretch, is common after stroke and imposes significant therapeutic challenges. It is believed that disinhibition of brainstem nuclei is responsible for spasticity, but there is debate on which individual nuclei within the brainstem, i.e. the lateral vestibular nuclei or pontine reticular formation within the whole reticular formation, are primarily involved. As such, we aimed to localize the activity of these individual brainstem nuclei via 3T functional magnetic resonance imaging (fMRI) in 10 people with post-stroke spasticity and compared the data with 11 age-matched healthy individuals. We used both acoustic and visual stimuli aimed at activating the pontine reticular formation and lateral vestibular nuclei. Visual stimuli were presented in the form of a moving checkerboard evoking illusionary self-motion to activate the vestibular network. Acoustic stimuli used loud (Sound pressure level [SPL] = 100 dB) acoustic bursts, expected to evoke a startle reflex and therefore activate both pontine reticular formation and lateral vestibular nuclei. We expected to find greater asymmetry in activation of brainstem nuclei in post-stroke compared to healthy individuals. We did not observe any difference in lateral symmetry between the two groups. However, we found that chronic stroke individuals exhibited three significant (p < 0.05) effects: 1) there were different delays in response to visual stimuli compared to acoustic stimuli; 2) the level of activation of the vestibular nuclei in post-stroke individuals was correlated to age, time since stroke, and brainstem volume); 3) brainstem volume of chronic stroke participants was smaller than in healthy individuals. These findings suggest that further efforts would be

Keywords: Surgical robotics - control
Abstract: In this paper, we present a synergic learning algorithm to address the task of indirect manipulation of an unknown deformable tissue. Tissue manipulation is a common yet challenging task in various surgical interventions, which makes it a good candidate for robotic automation. We propose using a linear approximate Q-learning method in which human knowledge contributes to selecting useful yet simple features of tissue manipulation while the algorithm learns to take optimal actions and accomplish the task. The algorithm is implemented and evaluated on a simulation using the OpenCV and CHAI3D libraries. Successful simulation results for four different configurations which are based on realistic tissue manipulation scenarios are presented. Results indicate that with a careful selection of relatively simple and intuitive features, the developed Q-learning algorithm can successfully learn an optimal policy without any prior knowledge of tissue dynamics or camera intrinsic/extrinsic calibration parameters.

Keywords: Novel mechanisms and actuation, Surgical robotics - design, Soft robotics
Abstract: Tendon-driven robots are commonly used in minimally invasive surgery (MIS) due to its small size, high dexterity and great controllability. However, there is a general tradeoff between miniaturization and dexterity of a tendon-driven robot, due to the large number of tendons required to actuate a multi-degree-of-freedom (multi-DoF) robotic motion. This work presents a novel design of robot that employs only one pair of tendons to actuate multiple rigidity-adjustable joints via a time-division multiplexing actuation method. The joints can be locked and unlocked by a fast-response and power-saving mini-solenoid clutch mechanism, and the accumulated actuation of the unlocked joints at each time yields a multi-DoF robotic motion. Experiments on a two-DoF robot prototype show that joint motions under the tendon actuation are well decoupled and have high repeatability in a wide range of angles. The results indicate that this method can be used to develop miniature and dexterous surgical robots with high control speed and accuracy.

Keywords: Surgical robotics - design
Abstract: Millions of patients suffering from eye disease cannot receive proper treatment due to the lack of qualified surgeons. Medical robots have the potential to solve this problem and consequently are attracting significant attention in the research community. This paper introduces a new 7-degree-of-freedom (DOF) 2-PRRRRR parallel robot with remote-center-of-motion for eye surgery. Mobility requirement, robot mechanism synthesis, kinematics, singularity, and dimensional optimization for the prescribed workspace, are analyzed and presented. A number of configurations were explored focusing on cataract surgery resulting in one being identified that covers the area completely and has excellent dexterity, but further work will be needed to enhance accuracy.

Keywords: Technology assessment in human subjects/outcomes, Surgical robotics - design, Computer vision in surgery
Abstract: Robotic Trans Abdominal Pre-Peritoneal (R-TAPP) inguinal hernia repair is a leading technique for repair due to lower incidences of postoperative pain, faster recovery and fewer complications. R-TAPP is often performed using a tele-operated surgical robots and it is a technique that novice surgeons have difficulty learning quickly. A cost-effective physical model with embedded electric circuits and software is created to simulate R-TAPP surgery process on living human tissue and anatomy, allowing for repeated practice and automatic assessment on various portions of the surgery using a daVinci Robot. This model was designed to serve as an assessor of performance and a teacher for the R-TAPP procedure. The feedback given was aural, visual and/or tactile. A case study was conducted to qualitatively evaluate the model. We present the results and suggestions for future work.

Keywords: Surgical robotics - control, Surgical robotics - design, Surgical navigation and localization
Abstract: Percutaneous interventions for diagnostic and therapeutic purposes such as thermal ablation, biopsy, and brachytherapy demand precise navigation of surgical needles in soft tissue towards the target. Active needle steering enhances navigation and increases target placement accuracy, and consequently improves clinical outcomes. In this work, a 3D steerable active flexible needle with multiple Shape Memory Alloy (SMA)-wire actuators is introduced. A resistive-based feedback loop control system was designed and tested to control the SMA’s actuation. The needle tip position was controlled through feedback loop measurements of the electrical resistance of the SMA-wire actuators. Self-sensing capabilities of SMAs were used in the control system to realize a 3D manipulation at the needle tip. The control system performance and accuracy of the control algorithm in manipulating the 3D steerable active needle via precision control of the SMA-wire actuators was tested in reference path tracking of the needle tip.

Keywords: Biomechanics and rehabilitation, Technology assessment in human subjects/outcomes
Abstract: The effects of HIV-related cognitive impairment on motor recovery after stroke are poorly understood. A barrier to this is quantitative measures that can be used to assess both cognitive and motor function in a rehabilitation setting. In this paper, we present preliminary data collected from a sample of stroke and HIV-stroke subjects on robot-based metrics that relate to cognitive and motor clinical scores. Additionally, we explore representing the dominant and non-dominant limb performance in a single laterality index metric. Our results show that this is a feasible approach to stratifying stroke and HIV-stroke patients with varying levels of cognitive and motor impairment by functional level to develop more targeted neurorehabilitation strategies in the future.

Keywords: Algorithms and machine learning, Activity recognition and health monitoring, Wearable technologies
Abstract: Human Activity recognition has many potential applications in telemedicine and rehabilitation. The advent and recent improvements in deep learning have encouraged the development of more accurate classifiers that were previously unprecedented. The present work focuses on classifying human activity using small, raw datasets collected with instrumented and customized footwear. Multi-channel time series data were recorded and transmitted wirelessly to a recurrent neural network classifier, which was able to classify 6 activities with an accuracy of up to 87.0pm8.9%. Lower level features were detected by the use of convolution. The present work shows that it is possible to use artificial neural network based software techniques to accurately classify data, even with small datasets and low-cost electronic hardware.

Keywords: Biologically inspired systems - design, Exoskeletons and prostheses - design, Novel sensors
Abstract: The aim of having artificial hands for robots or prosthetics that achieve fluent manipulation requires the efficient management of data from many sensors, and particularly from tactile sensors. Texture recognition is commonly carried out with tactile sensors, and active touch allows the detection of features beyond the limits imposed by the size and spatial resolution of the sensor. This paper assesses the feasibility of two hardware-friendly preprocessing algorithms to detect different textures. The proposed approach is focused on the implementation of simple hardware that can be replicated in the local electronics of a smart sensor to process the data from every force sensing unit or tactel in the tactile array. Experimental results from periodic textures are provided to show the feasibility of the approach. For instance, the processing of the output of a single tactel provides 13.27 mm for a 13 mm wavelength sample, and 1.66 mm for a wavelength of 1.67 mm, when the spatial resolution of the sensor is 3.70 mm. This solution is not enough for non-periodic textures, but provide useful information about the coarseness of the surface.

Keywords: Wearable technologies, Novel sensors, Activity recognition and health monitoring
Abstract: People with autism spectrum disorder (ASD) man- ifest great heterogeneity in their atypical sensory behaviors. It is estimated that 95% of people with ASD have a Sensory Process Disorder (SPD). People with ASD have the need to control what happens in their environment. However, it is inevitable that new situations occur in a person’s daily life. Therefore, it is important to monitor most of the circumstances they face in an attempt to try to predict the appearance of disorders that end up affecting their behavior. This paper presents the first steps towards the development of a system to know the value and effect on the SPD of different biological and environmental parameters. To know those variables, two electronic devices have been designed. The first one is an electronic system for capturing environmental variables, which is easily controllable by artificial intelligence and that is portable and mobile. The second electronic device is a soft wearable wrist to get biological parameters. To know the effect of those variables on the SPD, a complete software platform has been implemented. Both devices upload day-to- day data to a cloud database where the information is stored in timeseries data of different parameters. The system uses the data to learn a personalized model that is designed to manage the SPD of the user

Keywords: Wearable technologies, Novel sensors, Exoskeletons and prostheses - control
Abstract: The walking speed is an important piece of information in many areas of scientific research. In recent studies, multi-camera based optical motion capture systems have largely been used to obtain accurate information on body motion and walking speed. However, such motion capture systems are expensive, complex, immobile, limited to laboratory usage as they require the installation of multiple (6-10) cameras in an enclosed place. Lack of motion ground truth has limited further development of the state of art methods for motion estimation. Motivated by that, this paper proposes to employ a single LIDAR based portable system to capture body motion and generate ground truth for walking speed both indoors and outdoors. In a laboratory test involving three participants walking 10 times toward the LIDAR sensor, the body motion obtained from the LIDAR matched with an 8-camera based motion capture system with an R2 coefficient of 0.997. A novel application area of the LIDAR based motion estimation is also provided in this paper where the LIDAR was employed to obtain a speed profile of the people approaching a staircase. LIDAR derived this speed profile might be helpful in developing assistive technologies (such as prosthesis or orthosis modules) for the amputees or people in need.

Keywords: Wearable technologies, Algorithms and machine learning, Exoskeletons and prostheses - control
Abstract: Data-driven gait prediction can provide a reference trajectory for prosthetic limb joint motion. We have developed a Coordinated Movement (CM) approach that maps movements of the body to movements of target joints, such as ankle and knee. In this paper we apply a variation of the previous approach, by including a history of the target joint angles as inputs to the model. We also apply Attention, allowing dynamic reallocation of importance of the inputs over time. We present performance improvements over our previous CM controller, including for discrete maneuvers in an obstacle crossing task. We observe that Attention can follow important events in gait over time.

Keywords: Biologically inspired systems - control, Algorithms and machine learning
Abstract: Deep Reinforcement Learning (DRL) approaches have shown tremendous success over the last years in different application areas. But control of robots in real world settings and when facing unpredictable environments has still proven to be a difficult task that requires unreasonable long training times. This has sparked new interest in the organization of animal and human motor control systems and how to transfer these insights into such DRL learning architectures. While a hierarchical organization has now been advocated and introduced into a couple of DRL approaches, we propose decentralization as one further effective organizational principle. As an example we are considering insect locomotion (in particular locomotion of stick insects) which is adaptive and robust even when dealing with unpredictability. The underlying control system is assumed to consist of six individual local control modules that each control the action of a single leg. These local controllers only share limited information with neighboring legs which has shown sufficient to produce robust adaptive walking. In this article, we implement such a decentralized architecture of six local control modules and train it using Deep Reinforcement Learning. This architecture shows faster training times compared to a standard centralized approach and a trend towards more adaptive behavior and better performance when facing uncertain environments.

Keywords: Technology assessment in human subjects/outcomes, Human-machine interfaces, Human-robot interaction
Abstract: Numerous assistive devices possess complex ways to operate and interact with the subjects, influencing patients to shed them from their activities of daily living. With the purpose of presenting a better solution to mitigate issues generated by complex or expensive alternatives, a test comparing different user-prosthesis interfaces was elaborated to determine the effects of diverse aspects in their user-friendliness, including that of a version created for this work. A simplistic, anthropomorphic and 3D-printed upper-limb prosthesis was adapted to evaluate all the renditions considered. The chosen design facilitates the modification of its operational mode, facilitating running the tests. Additionally, the selected prosthetic device can easily be adapted to the amputees' lifestyle in a successful way, as shown by experimental results, providing validity to the study. For the interaction process, a wireless third party device was elected to gather the user intent and, in some renditions, to work in tandem with some sort of visual feedback or with a multimodal alternative to verify their impact on the user.

Keywords: Human-robot interaction, Human-machine interfaces, Biological signal processing and identification
Abstract: In this paper, we introduce and explore a concept called coupling effect, which pertains to the influence of force feedback on the commanded stiffness that is voluntarily controlled by the operator through the stiffness interface during bilateral tele-impedance. The degree of coupling effect depends on the type of interface used to control the impedance of the remote robot. In case of muscle activity based stiffness command interfaces, the force feedback can invoke involuntary changes in the commanded stiffness due to human reflexes. These involuntary changes can be either beneficial (e.g., during position tracking) or detrimental (e.g., during force tracking) to the task performance on the remote robot side. To investigate the coupling effect in different types of stiffness command interfaces (i.e., coupled and decoupled), we conduct an experimental study in which participants are asked to perform position and force tracking tasks. The results show that in both position and force tracking tasks a lower tracking error of the reference stiffness is obtained with a decoupled interface (p < 0.001). However, the unexpected force perturbation yields lower absolute position error when using a coupled interface (p = 0.0091), which indicates a specific benefit of the coupling effect. Finally, a lower absolute force error is found in the force tracking task by using the decoupled interface (p < 0.001), which indicates a specific downside of the coupling effect.

Keywords: Exoskeletons and prostheses - control, Biomechanics and rehabilitation, Human-robot interaction
Abstract: Lower-limb exoskeletons have shown great potential to assist human locomotion. However, effective methods for modulating aspects of gait behavior beyond reducing metabolic effort, such as stride time and gait kinematics, are still needed, especially for individuals with impaired gait. Using the Samsung GEMS-H exoskeleton, we studied the effect on healthy gait behavior of applying hip stiffness during overground walking. We found that applying positive stiffness did not affect stride time, but it did reduce hip range of motion. We also found that applying negative stiffness increased both stride time and hip range of motion. Additional analyses showed that the effect of applied hip stiffness on hip range of motion during overground walking was similar to that observed during treadmill walking. The effect of positive stiffness on stride time was similar during overground and treadmill walking, but the increase in stride time in response to the application of negative stiffness was greater during treadmill walking than in overground walking. Lastly, we found no evidence to indicate that neural adaptation or learning occurred when hip stiffness was applied. These results suggest that applying joint stiffness may be a promising approach to restoring healthy gait kinematics.

Keywords: Human-robot interaction, Biologically inspired systems - control, Robotic companions and social robotics
Abstract: Physical interaction of robots with their environment is a challenging problem because of the exchanged forces. Hybrid position/force control schemes often exhibit problems during the contact phase, whereas impedance control appears to be more simple and reliable, especially when impedance is shaped to be energetically passive. Even if recent technologies enable shaping the impedance of a robot, how best to plan impedance parameters for task execution remains an open question. In this paper we present an optimization-based approach to plan not only the robot motion but also its desired end-effector mechanical impedance. We show how our methodology is able to take into account the transition from free motion to a contact condition, typical of physical interaction tasks. Results are presented for planar and three-dimensional open-chain manipulator arms. The compositionality of mechanical impedance is exploited to deal with kinematic redundancy and multi-arm manipulation.

Keywords: Technology assessment in human subjects/outcomes, Soft robotics
Abstract: Proton therapy has a substantial physical advantage over conventional cancer radiation treatment with X-rays. Proton therapy reduces the radiation dose to healthy tissues and therefore the toxicity and side effects to the patients. However, the current high capital cost and required space make proton therapy a very limited resource. In current proton therapy, the patient is fixed on a table and a gantry is used to bend the proton beam for treatment. We propose to change the model by precisely moving a patient relative to a fixed proton beam rather than moving the beam relative to the patient. This requires a robot to move the patient and a strong immobilization device to ensure that the patient’s body position remains accurate during movement. We introduce a solution to enable compact and affordable proton therapy using a parallel robot with real-time surface positioning feedback and an immobilization system made of soft robotic actuators. Immobilization experiments with healthy volunteers demonstrate that our prototype device can position and immobilize healthy volunteers in sitting position to clinical standards and correct slouching through a feedback loop. The soft robotic immobilization device is strong but soft and comfortable as well as adaptive to the body shape. The force that the immobilization device can exert on the body using negative pneumatic pressure was characterized. This new immobilization device and the robotic positioning system have great potential to significantly reduce the cost of proton radiation cancer treatment.

Keywords: Exoskeletons and prostheses - control, Human-machine interfaces, Technology assessment in human subjects/outcomes
Abstract: Stroke often results in chronic motor impairment of the upper-extremity yet neither traditional- nor robotics-based therapy has been able to affect this in a profound way. Supporting the weak affected shoulder against gravity improves reaching distance and minimizes abnormal co-contraction of the elbow, wrist, and fingers after stroke. However, it is necessary to assess the feasibility and efficacy of real-time controllers for this population as technology advances and a wearable shoulder device comes closer to reality. The aim of this study is to test two EMG-based controllers in this regard. A linear discriminant analysis based classifier was trained using extracted time domain and auto-regressive features from electromyographic data acquired during muscle effort required to move a load equivalent to 50 and 100% limb weight (abduction) and 150 and 200% limb weight (adduction). While rigidly connected to a custom lab-based robot, the participant was required to complete a series of lift and reach tasks under two different control paradigms: position-based control and force-based control. The participant successfully controlled the robot under both paradigms as indicated by first moving the robot arm into the proper vertical window and then reaching out as far as possible while remaining within the vertical window. This case study begins to assess the feasibility of using electromyographic data to classify the intended shoulder movement of a participant with stroke during a functional lift and reach type task. Next steps will assess how this type of support affects reaching function.

Keywords: Exoskeletons and prostheses - design, Biomechanics and rehabilitation, Exoskeletons and prostheses - control
Abstract: Abstract—Here we present our gear controller prosthetic leg for helping lower limb amputees drive a motor bike and control the gear pedal of the motor bike through partial motion of muscles in the unsevered part of the legs. This is more intuitive for the driver. We also present our survey of the prosthetic legs with EMG function and describe our system to show how different it is from the existing systems.

Keywords: Novel mechanisms and actuation
Abstract: We developed an electromechanical aortic constriction device and used a mock circulatory loop to demonstrate its ability to alter ventricular and aortic hemodynamics to simulate pressure overload physiology observed in several cardiovascular diseases. Lumped-parameter and FEA computational models were used to validate our benchtop results.

Keywords: Exoskeletons and prostheses - control, Algorithms and machine learning, Human-robot interaction
Abstract: This study proposes an AI therapist that estimates the Borg scale, the rating of perceived exertion, from clinical data of stroke patients and real-time robotic data of a gait training robot, in robot-assisted gait training. This study was performed using the SUBAR, a lower limb exoskeletal robot, approved as a medical device by the Korea Food & Drug Administration (KFDA). Seventy-two stroke patients wore the SUBAR and participated in the robot-assisted gait training at Korean hospitals. The therapist asked and recorded the patient’s perceived training intensity on the Borg scale every minute. 20,790 Borg scale data, 720 clinical data of stroke patients, and 251,327 robotic data were collected from 812 training sessions. The clinical data contain ten variables, such as stroke patient’s age, height, weight, FAC, MBI, MMSE, and BBS. The 16 robotic data consist of step length, walking speed, trunk angle, and lower limb joint angle. The Borg scale represents the patient’s perceived exertion with numbers from 6 to 20. The AI therapist applied a deep neural network (DNN) for learning the collected big data effectively. Input variables are the clinical and robotic data, and the output variable is the stroke patients’ Borg scale. The AI therapist used sixty participants’ datasets, including 16,420 Borg scale data for training. From the test set of twelve participants containing 4,370 Borg scale data, the AI therapist showed 92.1% accuracy for the Borg scale estimation. Currently, the AI therapist is deployed in the SUBAR and provides the stroke patient’s Borg scale in robot-assisted gait training. The estimated Borg scale can be used to control the training intensity and select the robot's gait parameters for effective rehabilitation.

Keywords: Activity recognition and health monitoring, Algorithms and machine learning, Exoskeletons and prostheses - control
Abstract: Several methods were compared for anticipating speed transitions in lower extremity exoskeletons using electromyography on the gastrocnemius muscles. Using a neural network that incorporated temporal relationships had the highest overall accuracy, while a method that incorporated a moving average had the highest transition accuracy. There are opportunities to further refine these methods.

Keywords: Exoskeletons and prostheses - control, Algorithms and machine learning, Biomechanics and rehabilitation
Abstract: Modern control strategies for wearable assistive technologies involve tuning parameters to enable the assistance to be specific to an individual and activity. This approach can be costly in time and resources, and relies on the interaction with a biomechanics expert over many hours or days, which limits the real-world applicability. In this paper, we present a novel approach in the design and control of wearable assistive technologies using user's preference. The objective of this work is to predict the user’s preferred stiffness of a variable-stiffness ankle prosthesis using different machine learning algorithms.

Keywords: Human-robot interaction, Human-machine interfaces, Biomechanics and rehabilitation
Abstract: Physical interaction with an object that has internal dynamics can be challenging, both for humans and robots. An example is carrying a cup of coffee, where the nonlinear dynamics between the cup and the liquid can be chaotic and unpredictable. This study examined how nonlinearity of an object’s dynamics contributed to the difficulty of a task and if linearization of the object dynamics facilitated performance. Human subjects did a task in a virtual set-up with a haptic interface using a robotic manipulandum. The task of transporting a cup of coffee was reduced to a 2D cart-and-pendulum model; subjects moved the cart and felt the dynamics of the pendulum representing the coffee. Performance with the nonlinear system was compared to a linearized version of the system. Subjects (n=16) executed continuous rhythmic, self-paced movements. In the linearized system subjects chose to move at frequencies close to the resonant frequencies and clearly avoided the anti-resonance frequency. In the nonlinear system subjects did not avoid the anti-resonance. To evaluate performance, mutual information quantified predictability between cup and object dynamics. Mutual information was lower in trials when the cup moved close to the anti-resonance frequency in both linear and nonlinear systems. The interaction forces were higher in the linear system, especially at frequencies slightly below the anti-resonance. These results run counter to the expectation that linearization would simplify this task. These findings may be useful in design considerations for robot control and human-robot interaction: if humans interact with robots that exhibit complex dynamics in the frequency range of human actions, linearizing a nonlinear system may potentially disturb intuitive and low-effort cooperation.

Keywords: Wearable technologies, Novel mechanisms and actuation, Biomechanics and rehabilitation
Abstract: Wearable tremor suppression devices have been proposed as a promising alternative to suppress or reduce tremor motion associated with neurological disorders. To fully benefit patients, available tremor suppression devices need to be improved in their design, size, weight, and control. Although tendon-driven transmission systems are able to decrease the size and weight of these devices, they have complex control system requirements due to their substantially nonlinear behavior. For this purpose, this paper aims to develop a precise kinematic model of a wearable tremor suppression glove by considering the configuration of its tendons and sheaths, in order to improve the tendon arrangement, study the kinetic model of the glove, and increase the accuracy of the control system. A novel model is presented to calculate the tendon travel during hand motion. The derived kinematic model of the glove was verified by both simulation and benchtop experiments, and the new model has been validated. The mean correlation coefficient for the kinematic model is 0.90 ± 0.01.

Keywords: Wearable technologies, Algorithms and machine learning, Community mobility and support
Abstract: The number of visually impaired people is constantly increasing. Mobility aids such as white canes are designed to help these people navigate in their everyday life. However, these systems have limitations in certain environments or activities. To meet this challenge, we propose an assistance system to support the navigation of visually impaired people in low structured environments. The system captures information from its environment with a stereo camera. The disparity and color images are processed by an embedded computer which calculates a safe walking direction and accordingly provides vibrotactile feedback to the visually impaired person. This paper presents an approach to segment the traversable path in the captured color image of low structured environments using artificial neural networks. Therefore, a dataset is generated and several encoder-decoder architectures are evaluated and optimized to achieve a sufficient accuracy and real-time framerate of the binary segmentation on a mobile computer. The optimized architecture manages to segment the path with a test Intersection over Union value of 0.9378 and a segmentation time of 96 ms per image in various low structured environments.