Keywords: Sensors and Signal Processing, Control Theory and Applications, Civil and Urban Control Systems
Abstract: Distributed state estimation (DSE) has emerged as a scalable alternative to centralized state estimation (CSE) in large-scale power systems, where a single control center faces computational and communication bottlenecks. However, existing fully distributed state estimation (FDSE) approaches often suffer from reduced accuracy compared to centralized methods, particularly in multiarea power grids. This paper presents a non-gradient-based FDSE algorithm, referred to as DSE to CSE (DSE2CSE), which achieves centralized-level estimation accuracy while preserving a fully distributed framework. Inspired by the Gauss-Seidel method, the proposed method allows each local estimator to refine its estimate by iteratively exchanging information on unshared states with neighboring areas. Through the iterative process, the distributed state estimates converge to the global optimal solution achieved by CSE. Simulation results on the IEEE 14-bus system demonstrate that DSE2CSE precisely recovers the centralized estimates across all tested scenarios, highlighting its effectiveness for high-accuracy FDSE in large-scale power networks.

Keywords: Sensors and Signal Processing, Artificial Intelligence Systems
Abstract: Accurate classification of brushing regions remains a critical challenge in the development of intelligent oral hygiene monitoring systems. This paper introduces a novel approach that integrates an inertial measurement unit (IMU) and an orange ball marker affixed to a toothbrush, captured by an external RGB-D camera. By fusing data from the IMU and camera, the orientation and position of the toothbrush are accurately estimated. A brushing direction vector is then computed from the relative positions of the toothbrush head and the user’s mouth. This direction vector, combined with acceleration and orientation data, serves as input to a transformer encoder-based model that classifies brushing activity across 16 distinct regions. Experimental results demonstrate that the proposed method outperforms existing approaches, achieving an F1-score of 95.23% in the all-subjects evaluation and 80.34% in the leave-one-subject-out scenario.

Keywords: Control Theory and Applications
Abstract: This paper addresses the leader-following consensus problem for linear multi-agent systems and proposes a novel dynamic event-triggered mechanism that balances communication efficiency and control performance. The proposed control strategy incorporates internal variables that evolve adaptively based on the local states of agents and system dynamics, which are then used to determine the timing of control updates. These internal variables help reduce unnecessary communications while maintaining consensus performance. The stability analysis in this work adopts a piecewise framework, allowing for effective theoretical validation under more practical conditions. A key contribution of the proposed method is the design of two distinct triggering conditions that are selectively applied depending on whether the estimation error is zero or nonzero, providing an event-triggering strategy that is context-aware and more responsive to individual agent behavior. Although some parameters in the triggering condition require knowledge of global network properties, the execution of the control law is decentralized. Rigorous theoretical analysis ensures asymptotic convergence of follower agents to the leader’s trajectory, and simulation results demonstrate the effectiveness, robustness, and communication-saving benefits of the proposed strategy.

Keywords: Control Theory and Applications
Abstract: This paper presents a design of supervisory control(SC) for Heating, Ventilation, and Air Conditioning (HVAC) systems in general buildings. The SC design aims to optimize the operating cost and includes the interaction between the dehumidifier input and power consumption of the heat exchanger. Disturbances are derived from both historical data and estimations, and are integrated into the mathematical model of the HVAC system to determine the optimal set-points for room temperature and humidity. The supervisory control (SC) strategy is formulated as a quadratic programming problem which can be efficiently solved. The results suggest that estimating power consumption in supervisory control layer does not require the humidity signal to be included, and could be treated as a fixed cost for efficiency for optimization.

Keywords: Control Theory and Applications, Process Control Systems
Abstract: The Membrane Bioreactor (MBR) is a wastewater treatment system that combines membrane filtration with biological degradation, allowing key process variables to be regulated through automated control systems. Conventional controllers such as PID and linear Model Predictive Control (MPC) have been applied in previous studies, but they face limitations in handling the nonlinear behavior of the system. This paper proposes the design of a Nonlinear Model Predictive Control (NMPC) strategy to enhance reference tracking performance and maintain ammonia concentration within environmental limits. The NMPC is developed based on a nonlinear dynamic model and formulated as quadratic program involving the nonlinear characteristics. We apply a Sequential Quadratic Programming to solve nonlinear optimization problem to determine the control input. Numerical results demonstrate that NMPC provides smooth and accurate tracking responses, whereas the designed PID controller, although faster in some cases, tends to cause higher overshoot and degraded performance under certain conditions.

Keywords: Biomedical Instruments and Systems, Human-Robot Interaction
Abstract: Human-robot interaction (HRI) requires robotic systems to adapt to human movement. Human arm impedance, a key factor in this tuning process, is typically estimated using either perturbation methods or electromyography (EMG)-driven models. The latter allows continuous estimation by calculating muscle forces based on sensor signals, but only computes stiffness. However, damping estimation is equally crucial for stability, and due to its nonlinear nature, accurate estimation remains difficult. In this study, we employ an EMG-driven model to estimate arm stiffness during a ball-bouncing task, followed by computing the damping profile using an Unscented Kalman Filter. The estimated impedance profiles are then validated against existing literature and used to replicate a ball-bouncing movement recorded during experiments. A task chosen as it presents cyclic upper-limb motion with intermittent environmental interaction.

Keywords: Biomedical Instruments and Systems
Abstract: This paper presents the development of a graphical user interface (GUI)-based software tool to assist in the design of DNA sequences for use in molecular logic circuits. DNA computing, which exploits strand displacement reactions, offers a promising framework for constructing logic functions at the molecular level and is typically utilized in the design of molecular robots. However, designing appropriate base sequences remains a labor-intensive and error-prone process largely because of the lack of systematic methodologies and the combinatorial complexity of unintended interactions. The proposed application addresses these issues by integrating domain and strand definitions, secondary structure specifications, and thermodynamic simulations into a unified interface by leveraging the NUPACK analysis library. Designed for accessibility, the software enables users to perform automated DNA sequence design by specifying target complexes and reactions, and by applying structural and thermodynamic constraints—all without coding. An example design of a DNA-based AND gate demonstrates the software's ability to generate sequences that satisfy the given design constraints.

Keywords: Biomedical Instruments and Systems, Artificial Intelligence Systems, Information and Networking
Abstract: This study demonstrates the feasibility of employing a Q-learning model to personalize hint delay times in digital learning tasks for children with autism spectrum disorder (ASD). Leveraging data from a previous experiment, the model was developed to address the wide variability in ASD learning profiles. K-means clustering classified performance metrics (e.g., accuracy, completion time) into distinct learning profiles, which defined the Q-learning states. The model dynamically adjusted hint delays based on each child's real-time performance, ensuring individualized support aligned with their cognitive and motor abilities. Simulation results confirmed the model's effective adaptation across learning groups: fast learners (average accuracy 85.98%) received shorter delays to maintain engagement, while moderate learners (average accuracy 74.70%) received optimal consistent delays, and slow learners (average accuracy 59.36%) were given longer delays to support task completion. The converged Q-table values consistently reflected these tailored adjustments, highlighting the model's potential for real-time personalization. This adaptive approach shows promise for enhancing sustained attention, working memory, processing speed, and motor control in children with ASD.

Keywords: Biomedical Instruments and Systems, Artificial Intelligence Systems, Information and Networking
Abstract: Motor difficulties are common in children with Autism Spectrum Disorder (ASD), particularly in tasks that require both cognitive and motor control, such as manual dexterity, eye–hand coordination, and visual–motor integration, and these challenges are often associated with the severity of symptoms. This pilot study aimed to investigate the feasibility of classifying ASD severity by evaluating learning abilities through motor execution data collected from tablet-based dragging tasks. Sixteen children with ASD participated in a 4-week game-based intervention, during which they performed daily routine tasks in a digital game. Gameplay involved selecting and dragging objects to complete task steps. Three distinct dragging styles — straight, curved, and zigzag — were extracted to assess learning performance. Features such as trajectory accuracy, curviness, duration, and number of attempts were analyzed, and machine-learning classifiers were applied to differentiate between mild, moderate, and severe ASD. The zigzag task demonstrated the strongest potential (81.6% F1-score in mild vs. moderate severity), indicating that task complexity is crucial for revealing motor differences. Although preliminary and with limited effectiveness for simpler tasks, these findings establish a proof of concept for tablet-based assessments supporting individualized ASD monitoring. Future work will expand the dataset and refine models to enhance their clinical utility and personalize therapeutic interventions.

Keywords: Industrial Applications of Control
Abstract: This study proposes a model predictive control (MPC)-based energy management strategy for fuel cell electric vehicles (FCEVs), formulated using a unified state-space representation that captures the dynamic behavior of both the fuel cell system and the battery. A key feature of the proposed method is the inclusion of a cost function term that encourages operation around the fuel cell’s maximum efficiency point. The control framework is designed to concurrently reduce hydrogen consumption, regulate the battery state-of-charge (SOC), suppress abrupt variations in fuel cell power output, and guide the fuel cell stack power toward a reference associated with the maximum efficiency point. Simulation results confirm that the proposed approach achieves a 2.9% improvement in fuel economy and a 56% reduction in fuel cell stack output fluctuation compared to a rule-based strategy.

Keywords: Industrial Applications of Control, Navigation, Guidance and Control, Human-Robot Interaction
Abstract: The dismantling of end-of-life vehicles relies on manual labor and crushing, and is difficult to automate owing to differences among vehicle types. Among these tasks, bolt removal is a fundamental part of the dismantling process, and it is particularly challenging due to the need to align the tool with the bolt axis in all six degrees of freedom. The teleoperation of a robotic arm by a human is a feasible solution. This study aimed to develop a work support system to assist in aligning the position and orientation of a robotic arm with a bolt for removal. The proposed system guides the operation to facilitate the alignment of the bolt and tool, and estimates the bolt’s position even when it is obscured by obstacles. Experiments were conducted to evaluate the accuracy of the position and posture alignment systems. The results showed that the system achieved an average position error of 0.84 mm and an average posture error of 3.22°, demonstrating sufficient accuracy to support remote bolt removal tasks.

Keywords: Industrial Applications of Control
Abstract: The internal model control as a notable tool of robust control rejects external disturbances and compensates model uncertainties of controlled systems. In this paper, a control parameter design guideline of internal model control is proposed targeting uncertain linear time invariant systems with high-speed actuator, which makes the control algorithm simple but satisfies the performance requirement of control designers. The effectiveness of the design method is validated and evaluated using the example of a gas turbine system with uncertain parameters.

Keywords: Industrial Applications of Control, Artificial Intelligence Systems, Control Devices and Instruments
Abstract: This paper presents an ontology that incorporates the Failure Mode and Effect Analysis (FMEA) including its extension for Monitoring and System Response (FMEA-MSR) based on the AIAG/VDA 2019 standard. Building on this foundation, the work introduces PFMEA-MSR, a novel semantic extension of Process Failure Mode and Effects Analysis (PFMEA) that enables the runtime integration of failure knowledge in automated production systems. Traditional FMEA approaches are limited to static risk assessments during the design and planning phases. PFMEA-MSR allows access to formalized fault handling strategies during operation and supports automated adaptive system responses in Industry 4.0 environments. A semantic knowledge graph is developed to represent the relationships between failure modes, system functions, risk metrics, and corrective actions. This graph underpins a control architecture in which programmable logic controllers (PLCs) interact with a skill orchestrator via semantic queries to perform context-sensitive exception handling. Moreover, this research introduces an inference mechanism for dynamic risk assessment that is evaluated using an industry-related demonstration system. By establishing a seamless and automated information flow between failure documentation and operational control, the PFMEA-MSR framework enables scalable, adaptive manufacturing processes while enhancing traceability, risk transparency, and runtime resilience.

Keywords: Exoskeleton Robot, Rehabilitation Robot, Robot Mechanism and Control
Abstract: Walking is not only essential for mobility but also contributes to quality of life and emotional well-being. However, for many older adults, age-related muscle fatigue makes walking physically demanding. To address this, we developed a wearable ankle-assist device designed to reduce walking burden and enhance comfort. The system supports multiple control strategies, including speed-based assistance and a combined speed-and-torque approach, each providing distinct levels of support across the gait cycle. To evaluate their effectiveness, six participants performed treadmill walking under four conditions: one without assist and three assistive modes. Among these, the mixed-control method, which integrates torque support with speed guidance, demonstrated the most significant improvement in walking distance and reduction in heart rate change. These findings suggest that targeted assistance at specific gait phases can effectively reduce physical strain and promote more comfortable and sustained walking.

Keywords: Exoskeleton Robot, Artificial Intelligence Systems, Human-Robot Interaction
Abstract: Accurate estimation of human-exoskeleton interaction forces is essential for enhancing the safety, comfort, and effectiveness of assistance in functional rehabilitation. This paper presents a novel approach based on artificial neural networks to estimate interaction forces using the input and output data of an upper-limb exoskeleton. Unlike conventional methods that rely on explicit biomechanical models, the proposed method utilizes machine learning techniques to automatically extract relevant features from sensor signals, eliminating the need for prior knowledge of the musculoskeletal system. Experimental results demonstrate a significant improvement in estimation accuracy compared to a classical physics-based model, confirming the potential of data-driven neural approaches for real-time human-exoskeleton interaction modeling.

Keywords: Human-Robot Interaction, Artificial Intelligence Systems, Robot Vision
Abstract: This project aimed to develop HRI system for a Humanoid Robot that works on AI. The robot understands voice commands and body gestures and navigates by itself without a human need. The robot uses a chatbot built on Mistral API to process the speech. The Chatbot also Controls the arms and the navigation. The robot understands human gestures through MediaPipe. This tool has been programmed to identify specific gestures like waving and handshaking. Moreover, the robot works on ROS2 Humbel for arms control and navigation. The hardware includes an NVIDIA Jetson Orin Nano, a YDLIDAR X2 sensor, and an ESP32-based servo controller for the robot arm. All the parts are designed for scalability

Keywords: Human-Robot Interaction, Exoskeleton Robot, Robotic Applications
Abstract: Conventional motor-driven exoskeletons for ankle assistance add significant mass and bulk. This extra weight often causes discomfort and limits wearer mobility. To address these limitations, the Artificial Assistive Achilles Tendon (A3T) was developed as a lightweight, Switching Stiffness Pneumatic Spring (SSPS) system that delivers plantar-flexion assistance exclusively during the stance phase. The A3T pairs a double-acting pneumatic cylinder with a solenoid valve. When the valve closes, the mechanism operates like spring. When it opens, the ankle moves freely. Gait phase is detected using a single, wrist-worn IMU. By exploiting synchronized arm and leg swings from angular momentum conservation, complex algorithms and high-performance processors are not needed. At 0.45 kg per leg and minimal electrical draw, A3T compares favorably with motor based walking assist robots reported in previous research. In healthy young adults, surface EMG showed a 16.9 % drop in plantar-flexor activation. Indirect calorimetry measured a 10.6 % reduction in metabolic cost versus unassisted walking. These findings confirm that intent-driven switching stiffness pneumatic assistance can deliver effective ankle support with much less hardware complexity. Future work will refine sensor placement and power modules and extend trials to elderly and pediatric

Keywords: Human-Robot Interaction, Multimedia Systems
Abstract: This study develops a VR-based full-body motion learning support system that accounts for individual physical differences. By combining motion capture technology with virtual reality, the system aims to provide real-time consistency evaluation of movements and intuitive visual feedback. Specifically, we introduce full-body motion reproduction using a small number of trackers, real-time processing of 3D data, and an algorithm to compensate for individual body dimension differences. This enables personalized motion evaluation tailored to each learner's characteristics. Conventional systems have faced challenges where variations in body size and flexibility influenced motion assessment. Our system, however, is designed to apply corrections according to the characteristics of the learner, thereby achieving fair and highly accurate evaluations. This system is expected to be applied in a wide range of fields, including sports training, dance education, and rehabilitation.

Keywords: Human-Robot Interaction, Navigation, Guidance and Control, Robot Vision
Abstract: The force interaction in the teleoperation system is a significant realization. This paper presents a design and implementation of the contact force control. An omnidirectional mobile robot is designated as a haptic device to facilitate manipulation and contact force interaction between two human operators stationed in different locations. The modeling analysis of this system is based on a linear time invariant. With the unique design mechanism for contact force interaction, the haptic device can sense force in all directions. The network communication management in the teleoperation system is conducted under the VPN cloud and UDP protocol. The classical controller known as PID is employed for force feedback control. The feasibility of the force controller is evaluated using a common desired input across four function types: step change, linear, nonlinear, and tracking. The experimental results confirm that the PID controller effectively regulates contact force, enabling accurate tracking of the desired force.

Keywords: Human-Robot Interaction, Industrial Applications of Control, Navigation, Guidance and Control
Abstract: The paper explores the use of finger gestures as a means of human-machine interaction (HMI), with an emphasis on the ergonomics and usability of such systems. A parametric methodology was developed to create a gesture lexicon based on six critical criteria: comfort, intuitiveness, recognition accuracy, fatigue, ease of learning, and memorability. The process includes modeling the relationship between mental intention and the visual representation of the gesture, as well as employing HMM and DTW algorithms for real-time recognition. Initial experimental results showed a 12% improvement in recognition accuracy compared to conventional methods, a reduction in subjective user fatigue of up to 30% based on the Borg scale, high memorability rates (>85%) after a 7-day non-use period, and intuitiveness >80% on first attempts—indicating the naturalness of gesture use. The findings demonstrate that finger gestures, when designed according to ergonomic principles, can serve as an efficient and natural interaction medium in applications such as robotics, augmented reality, and accessible systems. This work lays the foundation for the development of intelligent interfaces with optimized user experience and technological maturity.

Keywords: Exoskeleton Robot, Human-Robot Interaction, Robot Mechanism and Control
Abstract: As demand for wearable assistive devices increases, the need for soft actuators with adaptive, directional stiffness control has grown correspondingly. This paper presents a novel soft actuator that combines a twisted string actuator (TSA) with a spiral layer jamming mechanism to enable multi-degree-of-freedom (DOF) motion with tunable stiffness. Unlike conventional jamming mechanisms that rely on pneumatic components, the proposed mechanism achieves stepwise stiffness modulation through motor-driven rotation alone, eliminating pneumatic component requirements.

Analytical models were developed to predict stiffness changes in both axial and bending directions and were validated experimentally using flexible materials. The results confirm that jamming significantly enhances stiffness. Bending stiffness increased by up to 69.3% across jamming states and by up to 48.5% compared to the silicone-only design. Axial stiffness showed up to 82.9% improvement between states and up to 29.1% over the baseline. The proposed design demonstrates proof-of-concept validation as a lightweight, compact solution for wearable robotic systems, with experimental results indicating suitability for back-support applications requiring adaptable and compliant actuation.

Keywords: Autonomous Vehicle Systems, Robotic Applications
Abstract: The safety of airspace around critical infrastructure (e.g., airports) can be enhanced by networked surveillance sensors. However, limited sensor numbers and narrow fields of view (FOV) allow adversaries to deploy mobile robots (e.g., drones) to evade detection. Improving surveillance effectiveness, therefore, requires identifying vulnerabilities in sensor placement and scanning patterns. This paper presents an attacker-centric framework, motivated by game-theoretic principles, to assess such vulnerabilities and enable iterative deployment analysis. We model the attacker’s strategy as a space–time path-planning problem under perfect knowledge of surveillance patterns and propose a novel variant of rapidly-exploring random trees (RRT*), termed space–time parallel RRT* (STP-RRT*). Our algorithm outperforms existing RRT* and ST-RRT* methods for surveillance-evasion planning. Effectiveness is demonstrated through extensive simulations and validated experimentally using a Crazyflie 2.1 quadrotor and a Reolink pan-tilt-zoom (PTZ) camera system.

Keywords: Autonomous Vehicle Systems, Control Theory and Applications, Sensors and Signal Processing
Abstract: This study proposes a novel Fail-safe mechanism for autonomous vehicles that is designed to minimize excessive emergency stops and unnecessary control interventions during sensor fault situations, while ensuring that a Minimal Risk Maneuver (MRM) is executed swiftly when actual risk accumulates. To this end, we define an exponentially accumulating risk assessment function that accounts for both the duration and the criticality of sensor failures. The decision-making logic is implemented using a Behavior Tree structure, enabling intuitive debugging of Fail-safe transitions through its modular, node-based flow. For validation, we constructed a CARLA-based digital twin environment. Using OpenStreetMap data, we generated a 3D simulation map of a complex commercial district in Korea and implemented a sensor fault interface to replicate diverse failure scenarios. Experimental results show that the proposed Fail-safe mechanism successfully satisfied the fault handling time intervals, completing MRM execution within 100 ms for critical sensors and 120 ms for noncritical sensors.

Keywords: Autonomous Vehicle Systems, Control Theory and Applications, Industrial Applications of Control
Abstract: This paper proposes a new formulation of distributed proportional–integral–derivative (PID) controller based on the dominant pole placement method for vehicle platoon systems with directed and undirected topologies. We have derived the analytical expressions of PID controller gains to place dominant and nondominant closed-loop poles at specified locations in the complex s-plane, enabling dominant pole placement via coefficient matching method to ensure internal and string stability of vehicle platoons. A genetic algorithm (GA) based optimization is used as sampler to obtain approx- imate stabilizable region within the design parameter space that satisfies analytically derived pole placement expressions,string stability criteria and minimizes the integral squared error (ISE). These obtained design parameters are then used to determine the optimal PID controller gains. The effectiveness of proposed approach is validated and compared via numerical study on a platoon with five followers under predecessor-following (PF) i.e. directed and bidirectional (BD) i.e. undirected topologies.

Keywords: Autonomous Vehicle Systems, Control Theory and Applications, Robotic Applications
Abstract: Accurate estimation of lateral wheel-soil interaction is essential for rover mobility control on deformable terrain. While terramechanics-based formulations offer physically grounded modeling, their reliance on numerical integration of anisotropic shear stress distributions renders them unsuitable for real-time control or estimation. Existing studies have attempted to overcome this via quasi-static simplifications, but these typically assume steady-state conditions and lack validation in time-varying scenarios. This paper proposes a real-time capable lateral force model by structurally simplifying the anisotropic shear stress distribution into an analytically integrable form. The resulting closed-form quasi-static force is then extended into a second-order dynamic model using a transfer function, allowing for transient response modeling under time-varying steering inputs. Additionally, we incorporate a physically motivated bulldozing resistance term, whose asymmetric effects across left/right wheels are captured via a side-slip-dependent scaling factor The proposed model is validated through full-rover simulations under varing steering amplitudes and frequencies, demonstrating accurate time-domain tracking with less than 10.5% normalized RMS error across conditios. The model requires only a single tuned parameter, ensuring pratical deployability with physical interpretability, and forms a foundation for onboard lateral force estimation and control on soft soil.

Keywords: Autonomous Vehicle Systems, Control Theory and Applications, Artificial Intelligence Systems
Abstract: This paper presents a robust lane-keeping control framework that integrates a traditional Linear Quadratic Regulator (LQR) with a Reinforcement Learning–based Winding Road Disturbance Compensator (RL-WRDC). While conventional LQR controllers exhibit excellent performance on road segments with nearly zero curvature, lane-tracking errors can arise on curved roads due to unmatched disturbances. To address this issue, the proposed compensator employs the Deep Deterministic Policy Gradient (DDPG) algorithm to learn additive control inputs that actively mitigate curvature-induced disturbances and sensor uncertainty in real time. An Error Characteristic–based Ornstein-Uhlenbeck (EC-OU) exploration strategy is introduced to balance exploration and exploitation. The OU noise covariance parameters are adaptively updated by analyzing the agent’s output during the initial training phase, enabling informed adjustments in subsequent sessions. This approach accelerates policy convergence while preserving exploration efficiency. Training is conducted in a simulation environment that replicates camera-derived road curvature profiles collected from the Gwangan Bridge. Evaluation results confirm that the proposed LQR with RL-WRDC architecture significantly enhances tracking accuracy and maintains closed-loop stability across a range of road geometries and sensor uncertainty scenarios, demonstrating its promise for next-generation autonomous driving systems.

Keywords: Autonomous Vehicle Systems, Control Theory and Applications, Robot Mechanism and Control
Abstract: GrThis paper presents a hybrid Fault Tree–Bayesian Network (FT-BN) framework for the systematic reliability assessment and real-time health monitoring of autonomous ground rovers. While conventional Fault Tree Analysis (FTA) is effective for identifying fault propagation paths, it is inherently static. To overcome this limitation, we integrate FTA with Bayesian Networks (BNs), which offer dynamic inference capabilities under uncertainty. Our methodology utilizes Mission Planner log data to extract anomalous events, which are then used to manually construct fault trees. These trees are subsequently converted into BNs to enable probabilistic reasoning. Experimental results demonstrate that magnetometer failure is the most significant contributor to the overall mission failure probability. The proposed FT-BN approach facilitates both structural fault analysis and dynamic risk evaluation, enhancing the safety and reliability of autonomous ground systems.

Keywords: Control Theory and Applications, Process Control Systems
Abstract: The spread of COVID-19 poses a major challenge due to uncertain parameters, stochastic effects, and complex regional dynamics. Traditional control strategies often fail to handle mismatched disturbances and time-varying uncertainties that arise in real-world epidemic systems. To address this issue, we propose an adaptive sliding mode control (ASMC) framework designed for a stochastic epidemic model with mobility. The control design incorporates integral-type sliding surfaces and adaptive switching gains to ensure robustness against both deterministic mismatched uncertainty and stochastic fluctuations. An Extended Kalman Filter (EKF) is employed for real-time estimation of unknown model parameters based on partial observational data. The ASMC method is applied to regulate vaccination and isolation efforts in Jakarta and West Java. Simulation results indicate that the proposed control scheme significantly reduces infection levels In Jakarta, the total infections in the regular and mutant compartments are reduced by 72.90% and 62.22%, respectively. In West Java, the reductions are 71.42% and 62.89%. These results demonstrate the effectiveness and resilience of the proposed method in epidemic control under uncertainty.

Keywords: Control Theory and Applications, Industrial Applications of Control
Abstract: The control of an aircraft’s attack angle is critical for flight safety. This study presents a control framework that accounts for nonlinear dynamics and discrete input constraints while remaining feasible for real-time implementation. Focusing on the F-8 Crusader aircraft, the original Mixed-Integer Nonlinear Programming (MINLP) problem is relaxed into a continuous Quadratic Program (QP) and addressed using a Linear Parameter-Varying MPC (LPV-MPC) structure, with discrete constraints approximated via Rule-Based Control (RBC). The framework is validated through Processorin-the-Loop (PiL) simulation on a Teensy 4.1 microcontroller. Both strategies stabilized the attack angle to 0 rad for prediction horizons of 4 and 8 with a 0.05 s sampling time, and the LPV-MPC+RBC method met the 50 ms real-time control requirement on the embedded platform.

Keywords: Control Theory and Applications, Artificial Intelligence Systems
Abstract: In this paper, the existence and uniqueness of solution to initial value problems for a class of fuzzy fractional differential equations involving the fractional generalized Hukuhara Caputo derivative is studied. By means of some fixed point theorems for weakly contractive mappings on partially ordered sets, some sufficient conditions are obtained.

Keywords: Control Theory and Applications, Control Devices and Instruments, Sensors and Signal Processing
Abstract: This study presents a speed control of permanent magnet synchronous motor (PMSM) under parameter uncertainties and external disturbance via an adaptive super twisting sliding mode control (ASTSMC) integrated with a sigmoid neural network (SNN). The ASTSMC is proposed to reduce chattering and enhance robustness compared with conventional super twisting sliding mode control (STSMC). Additional, the SNN model is used to estimate and compensate for the unknown part of mathematical model including parameter uncertainty and external disturbance. The stability of the method is proven using Lyapunov theory. To demonstrate the effectiveness of the proposed method, experiments on real PMSM were conducted.

Keywords: Control Theory and Applications, Robot Mechanism and Control, Robotic Applications
Abstract: An output feedback trajectory tracking controller is presented for a belt-driven wafer transfer robot (WTR) with unknown dynamics, achieving precise tracking. The controller employs an extended high-gain observer (EHGO) to effectively estimate the lumped effect of unknown dynamics and disturbances. The robot dynamic model is formulated using the Euler-Lagrange equation incorporating the elongation and contraction of the belts, as these factors are the dominant contributors to the tracking error. In contrast, coupling effects, frictional forces, and parametric uncertainties are assumed to be unknown, as they are not the primary determinants of tracking precision and are difficult to measure. This strategic distinction between known and unknown model components contributes to the performance of the proposed controller. The simulation results illustrate the minimization of the tracking error for the base link.

Keywords: Control Theory and Applications, Artificial Intelligence Systems
Abstract: Real-time control of hybrid systems is fundamentally constrained by the computational burden of solving mixed-integer quadratic programming (MIQP) problems. Classical approaches such as branch and bound and explicit model predictive control (MPC) experience exponential growth in computation time and memory usage as the prediction horizon or the number of discrete variables increases. To address this scalability bottleneck, we propose a novel learning-based control framework that employs a Transformer architecture to approximate MIQP solutions. The model is trained offline to map problem instances to high-quality control inputs, allowing fast inference during deployment without requiring online optimization. Through extensive experiments, ranging from a simple benchmark scenario to a complex multi-agent trajectory planning, we demonstrate that our method achieves near-optimal performance while satisfying real-time constraints. This work presents a promising direction for hybrid control, where learned optimization surrogates overcome the fundamental limitations of traditional solvers.

Keywords: Robot Mechanism and Control, Robot Vision, Robotic Applications
Abstract: One method for a surgeon to control a camera-holding robot during laparoscopic surgery without removing their hands from the surgical tools involves recognizing and tracking the surgical instruments. Although image processing is crucial, a control algorithm that smoothly operates based on the pixel coordinates obtained from the image processing is equally important. This paper proposes a control algorithm combining a Hammerstein model and model predictive control (HMPC) to maintain consistent control performance despite the zoom in/out of the endoscopic camera and the changes in the distance between the camera and the instruments, significantly affecting the control performance. Through experiments, a Hammerstein model was derived by extracting a nonlinear static model and a linear dynamic model. The performance of the HMPC in tracking surgical instruments was verified under static operations such as step inputs and dynamic operations following a trajectory. The experimental results indicated that the proposed algorithm enhanced the adaptability and consistency in stabilization time under various conditions, ensuring robust and stable tracking performance while effectively mitigating the impact of distance changes.

Keywords: Artificial Intelligence Systems
Abstract: This paper introduces a streamlined one-file Isaac Sim implementation for Reinforcement Learning (RL)-based control of a 12-DoF ant robot. The code is specifically designed to reduce the initial understanding and setup time for new Isaac Sim users, helping quicker adaptation and experimentation with RL methodologies. By simplifying the underlying code structure, we facilitates the development and deployment of RL-driven physical robotic systems. Experimental results demonstrate robust velocity following capabilities and successful obstacle traversal by the 12-DoF ant robot across simple environments, demonstrating its effectiveness, practical utility, and educational value.

Keywords: Robot Mechanism and Control, Robotic Applications, Information and Networking
Abstract: In recent years, there has been a growing demand for multiple autonomous mobile robots to collaborate in real-world environments such as logistics warehouses and construction sites. When multiple robots operate simultaneously in the same space, a certain system is required for each robot to avoid collisions autonomously. Conventional methods could solve this problem using several sensors, however it is still challenging to avoid the collisions with invisible situation in which sensors cannot detect each robot due to obstacles such as walls. To tackle this problem, in this paper, we propose Information Sharing Dynamic Window Approach (IsDMA in short) in which each robot exchanges their information using network and predicts their trajectories. Then each robot can avoid the collision even though sensors on the robots cannot detect each other. We use Distributed Streaming data Sharing Manager~(DSSM) to exchange information such as position, orientation, velocity and battery level as time-synchronized manner and extend the existing Dynamic Window Approach~(DWA). Moreover, we conduct experiments using real robots and show its effectiveness with detailed results.

Keywords: Sensors and Signal Processing, Robot Vision
Abstract: In this study, we considered using the depth data to recognize tableware on a table, extract the tableware parts from the acquired data, obtain the shape data and recognize the type of tableware. However, small items such as cutlery could not be successfully obtained because the difference in height from the table surface was small. Therefore, we proposed a method to successfully obtain the shape data of cutlery by modifying the table environment and using a glass table, which creates a difference in the measurement data of cutlery and other parts, and confirmed its effectiveness through experiments. We also considered the generalization of the proposed method, and considered a method that absorbs infrared laser light to the table surface, which is expected to have the same characteristics and effects as a glass table.

Keywords: Artificial Intelligence Systems
Abstract: In recent years, the environmental burden caused by the disposal of plastic bottles has become a growing concern, making recycling increasingly necessary. However, many of the collected bottles still have caps attached or contain other contaminants, necessitating proper sorting. Currently, the sorting of plastic bottles from other waste in recycling plants is largely done manually. In Japan, however, the declining labor force caused by an aging society and low birth rate has exacerbated the labor shortage, creating an urgent need for automation of such tasks. To address this problem, automated sorting systems using robotic arms in conjunction with bottle recognition have been proposed. This study focuses on the recognition of plastic bottles. Specifically, we improve recognition performance by adding depth information to RGB images using a monocular depth estimation model, Depth-Anything-v2, and perform semantic segmentation with DeepLab v3+. Compared to conventional methods, the proposed approach improves both recognition accuracy and boundary clarity between the bottle body and cap.

Keywords: Artificial Intelligence Systems, Information and Networking, Industrial Applications of Control
Abstract: Anomaly detection in images is an important task, especially in real-time computer vision applications. In this research, we address this task by improving EfficientAD, a lightweight and efficient anomaly detection model that combines high detection accuracy with fast inference. EfficientAD is based on the Student-Teacher architecture, where the student imitates the teacher over the distribution of normal images. By introducing a training loss that prevents the students from imitating the teacher beyond the distribution of normal images, EfficientAD improves the detection performance of anomalous features while significantly reducing the computational cost. Furthermore, to enable the detection of complex logical anomalies, an autoencoder capable of analyzing the entire image is efficiently integrated. In this paper, we attempt to further improve the detection performance of structural and logical anomalies by introducing new attention mechanisms, such as DANet and FPA, into the autoencoder. Through experiments, we will investigate and evaluate the effectiveness of this method in a real-world environment that requires highly accurate and fast anomaly detection.

Keywords: Artificial Intelligence Systems
Abstract: In recent years, the global adoption of renewable energy has accelerated, with solar power drawing particular attention due to its sustainability and ease of deployment. However, accurately identifying the distribution and condition of solar panels remains a significant challenge, especially when dealing with unregistered, aging, or damaged installations. These limitations hinder effective energy management and safety monitoring. To address this issue, image recognition technologies using satellite and aerial imagery have gained attention. While deep learning approaches have significantly improved detection accuracy, variations in environmental conditions and image quality caused by weather continue to present difficulties. This study proposes a robust solar panel detection method based on the YOLOv8-OBB model, enhanced with Deformable Convolutional Networks v2 (DCNv2) to better handle the geometric variability of solar panel installations. By incorporating DCNv2, we enable more adaptive feature sampling that can effectively capture the irregular shapes and orientations of solar panels. As a result, the proposed method achieves a mean Average Precision (mAP@0.5) of 0.926, demonstrating strong detection capabilities even under diverse conditions. This approach contributes to the automation and accuracy of solar infrastructure monitoring, supporting safer and more efficient management of renewable energy systems.

Keywords: Biomedical Instruments and Systems, Sensors and Signal Processing, Artificial Intelligence Systems
Abstract: Stroke is currently the third leading cause of death worldwide, and even when it is not fatal, it often leaves people with permanent physical disabilities. Therefore, early detection and treatment are important issues to reduce the impact of stroke. X-ray computed tomography (CT) is used to diagnose stroke. Contrast is used to confirm detailed vascular structures. However, this examination increases the number of images required to view detailed blood vessels and places a heavy burden on the physician. To solve these problems, a computer-aided diagnosis (CAD) system is being developed. Because the CAD system can use the results of quantitative computer analysis, it is expected to improve the reproducibility and accuracy of diagnosis and reduce the burden on physicians. Therefore, this paper proposes a temporal subtraction technique of non-contrast and contrast-enhanced CT image to develop a CAD system. After the registration of the two steps, we plotted the difference values in the extracted parenchymal regions of the brain. The proposed method was applied to 5 cases of CT images and NCC=0.996 was obtained. It was also confirmed that it is possible to generate pseudo-color difference images.

Keywords: Artificial Intelligence Systems, Multimedia Systems
Abstract: The need for advanced diagnostic tools is underscored by the 8 million deaths annually caused by respiratory diseases. Auscultation is a non-invasive and repeatable method, but it is difficult to quantify because it relies on skilled practitioners. In addition, auscultation is limited in resource-constrained settings such as developing regions or disaster areas. In this study, we propose a deep learning-based respiratory sound classification system using the ICBHI 2017 dataset, which consists of 920 recordings categorized into normal, crackle, wheeze, and both (crackle and wheeze). We employ a ResNet34 model that uses residual connections to mitigate gradient vanishing, enhanced with a Convolutional Block Attention Module (CBAM) to highlight critical channel and spatial features for disease-specific patterns. Patient metadata (age, gender) processed via one-hot encoding and z-score standardization are integrated to enhance feature extraction. A multimodal learning framework incorporates data augmentation, including ±10 degree random spectrogram rotation, simulating microphone angle or patient posture variations to improve robustness. The model achieves accuracy of 0.763, sensitivity of 0.656, specificity of 0.859, and ICBHI score of 0.757, outperforming baseline models in accuracy and generalization. Metadata integration and augmentation significantly enhance diagnostic performance, validating the system's effectiveness for non-invasive diagnosis. This approach holds promises to support respiratory disease diagnosis in resource-limited settings.

Keywords: Civil and Urban Control Systems, Sensors and Signal Processing, Information and Networking
Abstract: Infrastructure safety inspections typically rely on visual inspections and hammering tests conducted by inspectors. However, an important challenge is variation of inspection results caused by differences in inspectors' technical expertise. To address this variation, we propose an inspection method and preventive work using a coating-type resin sensor combined with an infrared camera. Thermography is increasingly popular as a nondestructive evaluation technique for maintaining concrete structures. Most inspections evaluate only the locations of surface shapes and defects. No report has described assessment of defect depth.

Keywords: Industrial Applications of Control, Artificial Intelligence Systems, Control Theory and Applications
Abstract: Designing disturbance observers (DOBs) involves a trade-off between estimation accuracy and noise sensitivity, and tuning the DOB gain does not guarantee optimal performance. This paper proposes a dynamic gain DOB that uses deep reinforcement learning (DRL) to adaptively adjust the gain based on system conditions. A variable gain DOB is first derived by modifying a conventional structure, and a DRL-based policy is trained to control the gain. Case studies show that the proposed method adjusts the gain intelligently and outperforms conventional DOBs with fixed gains.

Keywords: Industrial Applications of Control, Human-Robot Interaction, Control Theory and Applications
Abstract: This paper proposes a haptic-shared control method using a force-feedback joystick to provide operational assistance and facilitate skill acquisition for slewing crane operators. In the proposed method, we develop a novel Model Predictive Control (MPC) that unifies swing suppression and positioning to accurately guide the payload to its target while mitigating load swing. This MPC generates haptic commands for the joystick, enabling the payload's position and swing angle to track their respective reference values. Furthermore, it incorporates a compensator to suppress control output drift caused by modeling errors. The proposed method realizes a control that simultaneously achieves payload swing suppression and target position tracking in a swing-free manner by applying the control output as torque commands to the force-feedback joystick. This establishes an effective haptic-shared system for operator assistance and training. We validate its effectiveness through numerical simulations and Hardware-in-the-loop (HIL) experiments.

Keywords: Industrial Applications of Control, Robotic Applications
Abstract: Feature extraction is a key step in the robotic welding process that has a critical impact on the welding quality and productivity. However, existing systems mainly use expensive sensors, which is a factor that hinders the spread of this technology to industrial sites. To address these limitations, this paper proposes a novel welding system that integrates a low-cost RGB-D sensor with a CAD model to recognize weld seams and gaps in real-time precisely. The proposed method employs a geometric model-based point cloud denoising approach, combining 2D and 3D data to enable robust feature extraction even from noisy data. Experiments demonstrated that our system achieves an optimal balance between speed and performance, attaining significantly faster processing times (0.190 s) and high accuracy (0.31 mm RMSE) compared to previous methods. The results of physics-based simulations verified that our system can reliably generate welding trajectories for various geometries, including planar and curved surfaces. This research demonstrates the feasibility of a cost-effective and flexible automation solution for complex welding processes.

Keywords: Industrial Applications of Control, Control Devices and Instruments, Control Theory and Applications
Abstract: In this paper, we propose a robust control strategy for electric brake booster~ (Ebooster) systems based on a backstepping controller with disturbance observer (DOB). The backstepping controller provides stable tracking performance under nonlinear load conditions, while the DOB estimates unmodeled disturbances such as hydraulic and spring forces. Unlike existing methods that apply disturbance compensation unconditionally, the proposed approach incorporates a conditional compensation strategy based on the direction of the tracking error and the estimated disturbances. This strategy avoids unnecessary compensation of useful forces, such as resilience during brake release, and improves control efficiency. To validate the proposed method, simulations were performed under three pedal input scenarios (fast, medium, and slow). The results show that the proposed controller achieves faster pressure release, lower cumulative current consumption, and improved responsiveness without degrading tracking accuracy.

Keywords: Industrial Applications of Control, Human-Robot Interaction, Autonomous Vehicle Systems
Abstract: This paper proposes a method for constructing a driver model that can reproduce the steering behavior of individual drivers in real time by identifying model parameters online. Furthermore, a driving assistance approach is presented that utilizes this personalized driver model. To adapt to the fact that a driver’s characteristics may vary during driving, the proposed method employs the recursive least squares (RLS) algorithm to update the driver model parameters sequentially. These updated parameters are then used to provide individualized driving assistance that reflects each driver's preferences and behaviors. To evaluate the effectiveness of the proposed method, three types of substitute driver models based on state feedback were developed, and simulation experiments were carried out. The results showed that, in comparison with assistance using only model predictive control (MPC) without considering personal characteristics, the proposed method with the incorporated corrector was able to reproduce the driver’s behavior more accurately and reduce the driver’s workload during driving.

Keywords: Rehabilitation Robot, Human-Robot Interaction, Robot Mechanism and Control
Abstract: In this research, we are developing owl-shaped pet robots capable of communicating with humans. We developed two types of robots, one that blinks physically and one that blinks via a display. These robots have multiple degrees of freedom in its neck and wings, recognizes humans and the outside world using built-in sensors, and communicates with humans using movements and cries. Furthermore, even when there is no contact with humans, it will autonomously generate emotions, allowing it to perform movements that give the impression of being more lifelike. We verify the effectiveness of displaying emotions and changes in emotions over the long term with the developed robots, as well as the effectiveness of displaying emotions by changing the display of its eyes, to verify the effectiveness of this robot.

Keywords: Human-Robot Interaction, Control Theory and Applications, Robot Mechanism and Control
Abstract: This paper proposes a novel system configuration to enhance human operator proficiency in bilateral control systems. Our approach treats the entire virtual internal model (VIM)-based bilateral control system as the controlled plant, augmenting it with an external assistance control system. We conducted real-device experiments to evaluate the effectiveness of this proficiency support control from two key perspectives: positioning and force regulation. For positioning tasks, experimental results confirmed that a PID compensator effectively adjusts the compensation force based on the operator's skill level. This adjustment significantly aids in precisely tracking the displacement to the target value. Furthermore, for force regulation tasks, an integral compensator was shown to provide targeted compensation according to the operator's proficiency. This enables the operator to accurately apply the desired target force. These findings demonstrate the proposed system's capability to provide adaptable assistance, thereby improving operator performance in diverse bilateral control scenarios.

Keywords: Human-Robot Interaction, Control Devices and Instruments, Robotic Applications
Abstract: This paper presents a human-drone interaction system that integrates a hand-gesture controller with a wearable vibrotactile feedback device. The gesture controller uses an inertial measurement unit (IMU) and a gated recurrent unit (GRU) neural network classifier with an error-correcting output code (ECOC) scheme to recognize user commands with 95.65% accuracy. Mapped gestures correspond to the drone’s movements. For vibrotactile feedback, we developed a belt with a 3×12 array of coin-type vibration motors that encodes the drone’s relative azimuth, altitude, and distance via activation position and vibration intensity. In a user study with 12 participants, participants achieved accuracy of 82.75% in identifying the indicated drone position using the tactile device, with a mean response time of 3.1 seconds. The results demonstrate that combining intuitive gesture inputs with vibrotactile feedback can significantly enhance teleoperation and situational awareness in unmanned aerial vehicle control.

Keywords: Human-Robot Interaction, Artificial Intelligence Systems, Sensors and Signal Processing
Abstract: This study proposes a real-time, audio-based method for detecting human–robot collisions in industrial en vironments. The system employs a 1D Convolutional Neural Network (1D-CNN) with channel attention to classify vibration signals from raw audio using Short-Time Fourier Transform (STFT). A novel hop-length adjustment strategy addresses class imbalance, achieving 95% overall accuracy and 100% recall for human collisions in real-time testing.

Keywords: Human-Robot Interaction, Autonomous Vehicle Systems, Robotic Applications
Abstract: In this paper, we propose an optimal longitudinal control system for personal mobility service robots, considering human factors derived from human-robot interaction (HRI) experiments. First, we design velocity profiles for different modes using Bezier curves to improve the driving safety of personal mobility service robots. We analyze individual mode preferences through usability evaluation in HRI experiments and derive a human factor that reflects user needs. Based on the derived human factor, we construct an optimal velocity profile that ensures user preference and convenience. To accurately track the optimal velocity profile while ensuring both driving performance and user safety, we design a model-based optimal control system using a Linear Quadratic Regulator (LQR) integrated with an error-based compensator. The proposed optimal longitudinal control system, considering human factors, is implemented and validated on an autonomous electric wheelchair equipped with autonomous driving sensors, demonstrating its effectiveness in providing a control framework for personal mobility service robots.

Keywords: Human-Robot Interaction, Robotic Applications, Artificial Intelligence Systems
Abstract: Effective emotional body expressions enhance Human-Robot Interaction (HRI), yet existing studies lack continuous emotion-body motion mapping. This work addresses this gap by proposing a music-driven framework that leverages instruments playing data to generate emotionally expressive motions. We construct a novel motion dataset annotated with Valence-Arousal labels via music emotion recognition, enabling the exploration of relationships between motion trajectories and emotional characteristics. Key trajectory features are identified to capture emotional expressions, and an innovative framework combining Dynamic Movement Primitives (DMP) and Deep Deterministic Policy Gradient (DDPG) is proposed for generating emotional body motions. We propose an emotion-driven humanoid robot named Emoid, and validate our approach through simulation experiments. Experimental results demonstrate its effectiveness in enabling robots to naturally express different emotional states and enhance HRI quality.

Keywords: Autonomous Vehicle Systems, Navigation, Guidance and Control, Industrial Applications of Control
Abstract: This paper examines the control system design problem of autonomous vehicles. In particular, we focus on the trajectory tracking control problem of articulated vehicles. Model predictive control is a well-established method in which the control performance can be optimized with taking constrained condition into account. This paper proposes a model predictive control method for a nonlinear mathematical model of articulated vehicles to achieve the trajectory tracking with obstacle avoidance. The effectiveness of the proposed method is verified by numerical simulations.

Keywords: Autonomous Vehicle Systems, Sensors and Signal Processing, Information and Networking
Abstract: Reconfigurable intelligent surfaces (RIS) are emerging as a key technology for sixth-generation (6G) wireless networks. By manipulating and reflecting electromagnetic waves using adjustable reflecting elements, RIS can actively reshape the wireless environment in real time. This capability proves most effective when the RIS is positioned in line-of-sight (LoS) and proximity to the transmitter and receiver. To meet these criteria in real-world applications, the RIS is mounted on an unmanned aerial vehicle (UAV). Combining UAV with RIS is particularly advantageous due to the low weight and energy-efficient design of the RIS. However, a UAV is exposed to many disturbances during flight, which can compromise the performance benefits of the RIS-enabled link. We show the RIS position can be estimated based on the UAV’s extended Kalman filter (EKF). By conducting experimental measurements and recording the UAV position and orientation during flight, we quantify the uncertainties regarding the position estimation of the RIS and analyse their impact on the quality of the RIS link through numerical simulations.

Keywords: Autonomous Vehicle Systems, Navigation, Guidance and Control, Industrial Applications of Control
Abstract: This study addresses the control problem of a four-wheeled vehicle in the presence of lateral slip and tire blowout. A four-wheeled vehicle is a dynamical system with various nonlinearities and constraints. Model predictive control is a type of optimal feedback control in which the control performance over a finite future is optimized and its performance index has a moving initial time and a moving terminal time. Nonlinear model predictive control method is known as one of the most successful control methodologies because it enables control performance to be optimized while taking nonlinearities and constraints into account. The objective of this study is to propose a control system design method for driving a four-wheeled vehicle on a slope during tire blowout. This paper provides a numerical solution method based on an algorithm called the C/GMRES method to solve the model predictive control problem for four-wheeled vehicle dynamics with considering lateral slip, tire blowout, and gradient resistance. The effectiveness of the proposed method is verified by numerical simulations.

Keywords: Autonomous Vehicle Systems, Control Theory and Applications
Abstract: In this paper, we present a data-driven control approach for autonomous track driving of an RC car using Data-enabled Predictive Control (DeePC). This method directly utilizes system input/output data without requiring modeling or machine learning. The cost function design for DeePC of this work includes contouring and lag error terms arising from RC car reference tracking in a racing track. To evaluate our method, we consider a control task where the vehicle follows a parameterized reference trajectory along the track while staying within the track. The experiment setup consists of a 1/27 scale remote control car, a remote controller, a motion capture system, and a desktop computer for solving the optimization problem associated with DeePC. The experimental results show that the proposed DeePC framework, operating at an 80ms sampling time, successfully completed four laps while tracking the centerline and remaining within the track boundaries.

Keywords: Autonomous Vehicle Systems, Artificial Intelligence Systems, Navigation, Guidance and Control
Abstract: Reinforcement Learning (RL) has shown promise in addressing mobile robot navigation tasks in the presence of dynamic and stationary obstacles, particularly in multi-robot scenarios. In this paper, we demonstrate the zero-shot generalization capability of deep RL, through successful deployment of trained policies in perturbed environments with- out any retraining or adaptation. We implement the Soft Actor-Critic (SAC) framework to train agents for navigation and decentralized collision avoidance, and evaluate its performance under dynamic goal conditions and kinematic perturba- tions. Each robot is modeled as a non-holonomic robot moving at a constant linear speed, with the SAC agent producing continuous, bounded angular velocities that takes it to its goal while avoiding collisions with obstacles. The observations include the poses of nearby agents, the robot’s own pose, and the location of the goal. The reward structure is designed to guide the agent based on angular alignment with the goal and avoidance of other agents. Through simulation studies based on the non-holonomic kinematic model, we show that the learned SAC policies generalize effectively across dy- namic target motion and model drift, achieving robust and collision-free behavior in previously unseen conditions. Our results highlight the potential of SAC for zero-shot deployment in real-world robotic systems.

Keywords: Control Theory and Applications, Autonomous Vehicle Systems, Sensors and Signal Processing
Abstract: This paper presents a novel reinterpretation of the Model Predictive Path Integral (MPPI) algorithm, traditionally employed as a sampling-based stochastic control method, as a real-time state estimator for fusing Ultra-Wideband (UWB) and Inertial Measurement Unit (IMU) sensor data. Departing from conventional MPPI applications that focus on optimizing control actions via trajectory sampling and cost minimization, our approach leverages the importance-sampling framework of MPPI to directly smooth and integrate measurement data within the cost function. By embedding both UWB range measurements and IMU dynamics into the cost evaluation of each sampled trajectory, our method achieves implicit sensor fusion with built-in smoothing and drift correction. To ensure real-time performance, we utilize GPU parallelization with CUDA, enabling the simultaneous evaluation of hundreds of thousands of trajectory samples. Experimental results on an unmanned aerial vehicle (UAV) platform demonstrate robust, drift-free localization in GPS-denied environments. These findings underscore the versatility of MPPI as a unified framework for both control and state estimation in aerial robotics, offering a powerful solution for real-time sensor fusion and localization.

Keywords: Control Theory and Applications, Autonomous Vehicle Systems
Abstract: This paper proposes a nonlinear control strategy based on Sontag’s formula for one-dimensional systems. To address the issue of chattering caused by signum function in control inputs, a modified version of Sontags formula is developed. The proposed control law guarantees asymptotic stability of the closed-loop system, which is verified using Lyapunov analysis. To demonstrate the effectiveness of the method, the controller is applied to a velocity control of an unmanned ground vehicle. Simulation results confirm that the proposed control input significantly reduces chattering while maintaining accurate tracking performance. This work lays the foundation for future research involving real-time implementation and experimental validation in practical environments.

Keywords: Control Theory and Applications, Control Devices and Instruments
Abstract: A model reference adaptive control (MRAC) based speed control framework is proposed for permanent magnet synchronous motors (PMSMs). The proposed scheme utilizes an acceleration-model-based MRAC structure to compensate for model uncertainties in both the mechanical and current dynamics of the PMSM. To further enhance robustness against external disturbances and parameter variations, a lumped disturbance is explicitly modeled and estimated using an extended state observer (ESO), which is integrated into the adaptive control loop. The combined MRAC-ESO approach improves disturbance rejection and maintains reliable tracking performance without sacrificing transient response. Lyapunov-based analysis is employed to theoretically guarantee the closed-loop stability of the system, and the boundedness of both the tracking and disturbance estimation errors are ensured. The effectiveness of the proposed control strategy is validated through MATLAB/Simulink simulations.

Keywords: Control Theory and Applications, Artificial Intelligence Systems, Process Control Systems
Abstract: Smart energy systems are one of those fields where, due to the complex dynamics involved, learning-based (optimal) control is regarded as a viable option to handle such complexity. Yet, an aspect common to most smart energy systems is the presence of switched dynamics like thermostatic behavior, on/off or discrete regimes. Using a case study based on thermostatically controlled loads (TCLs), this work shows that the lack of regularity/smoothness arising from switched dynamics can put learning-based control at stake. Classical optimal linear quadratic control (LQC) can be competitive or even superior in terms of robustness and generalization. To provide theoretical support to this observed phenomenon, we analyze the closed loop resulting from control-oriented dynamics of the TCLs and its LQC law, proving the existence of robustness against a wide set of unmodeled dynamics (including possibly switched dynamics).

Keywords: Control Theory and Applications, Robot Mechanism and Control, Robotic Applications
Abstract: Instead of relying on static layers and centralized structures, SWAPM dynamically adapts the level of hardware abstraction based on the requirements of each process. This allows robotic systems to move beyond rigid architectures and toward adaptive, context-aware behavior. A key component of our approach is the use of digital twins, not as monolithic control elements but as semantically structured representations of physical devices. These models enable flexible integration and coordination through a shared information layer. The framework introduces virtual devices as software entities that manage physical components and coordinate among themselves. This supports distributed orchestration, modular design, and process-driven control. SWAPM further allows for concurrent task execution and swarm-like behavior while maintaining goal-oriented structure.We demonstrate the system using a heterogeneous robot swarm that autonomously assembles visualized brick structures.

Keywords: Control Theory and Applications
Abstract: This paper investigates the effect of output measurement noise on the robustness of one-step predictors in Subspace Predictive Control (SPC), a data-driven framework that constructs predictors directly from input-output trajectories. When measurement noise is present, the data matrix is corrupted in such a way that its rank is increased from that of the clean data matrix, leading to degraded prediction accuracy. To mitigate this, we examine two rank-reduction techniques---Low-Rank Approximation (LRA) and Generalized Total Least Squares (GTLS). Through simulations on six representative systems and 5000 randomly generated systems, we show that GTLS consistently reduces predictor error, especially when the minimum nonzero singular value of the clean data matrix is moderately small. Furthermore, the same trend holds for higher-order systems, indicating that the observed relationship generalizes beyond the second-order case. These findings reveal a strong empirical correlation between the smallest nonzero singular value and predictor sensitivity to noise, and demonstrate that GTLS enhances predictor robustness to measurement noise.

Keywords: Control Theory and Applications, Autonomous Vehicle Systems, Industrial Applications of Control
Abstract: This study investigates robust control for electric propulsion in cycloidal rotor systems, a largely unexplored area despite their promise for sustainable marine and aerial transport. We propose a novel design integrates the distributed electric propulsion with cycloidal rotor, where electric motors serve as individual blade actuators. A nonlinear adaptive integral backstepping controller, with stability proven via Lyapunov theory, is developed for the cycloidal rotor's electromechanical dynamics under parameter uncertainties. Real-time adaptation laws estimate unknown parameters, and the resulting control laws are expressed in d– and q–axis voltages. Numerical simulations confirm the controller's effectiveness and validate the model, showing a 23 % reduction in pitch-angle tracking error under parameter uncertainties compared to the existing nonlinear approach.

Keywords: Control Theory and Applications, Information and Networking
Abstract: This paper presents a novel mode-based approach to address the leader-following consensus problem of multi-agent systems (MASs) under Denial-of-Service (DoS) attacks. Given that consensus in MASs heavily depends on communication networks, these systems are inherently vulnerable to cyber attacks, which may disrupt information exchange and degrade overall system performance. Consequently, ensuring secure consensus control under DoS attacks has attracted increasing research interest. To deal with this challenge, this paper introduces a mode-based framework that accounts for multiple types of DoS attack scenarios, along with a corresponding state-feedback controller designed for each mode while considering the agent connectivity. Furthermore, based on the proposed controllers and derived sufficient conditions, an efficient algorithm is developed to improve the performance of our mode-based approach. Eventually, the effectiveness of the proposed approach is demonstrated through two numerical examples.

Keywords: Control Theory and Applications, Robotic Applications, Robot Mechanism and Control
Abstract: This paper presents an integrated attitude control strategy for a one-degree-of-freedom(1-DoF) CubeSat, combining energy-based control and Linear Quadratic Regulator(LQR). To ensure the successful completion of satellite missions in space, the ground-based test is essential. For this purpose, the motion of both the mission-performing satellite and its target objects, such as in docking or debris removal scenarios, has to be represented under microgravity conditions. Therefore, 1-DoF CubeSat with a reaction wheel is utilized to execute and sustain a designated attitude. To achieve the swing-up of CubeSat to a target position, the energy-based controller is employed, and the system's stability is analyzed using passivity theory. In addition, the LQR controller is applied for stabilization, and a switching condition is proposed to integrate both control strategies. In order to validate the effectiveness of the proposed strategies, simulations for 1-DoF CubeSat system are conducted.

Keywords: Control Theory and Applications
Abstract: This paper presents estimating launching points of unmanned aerial vehicles (UAVs) equipped with boosters. When UAVs are detected and tracked by sensors such as radars, fast identification of vehicle launching points is needed for air traffic control and defense systems. When UAVs in a boosting phase are controlled by first-order pitch programming, this paper leverages inverse first-order pitch programming to analytically solve launching points. Furthermore, a two-point robust measurement selection (T-RMS) scheme is developed to reduce errors such as random noise and bias by utilizing multiple moving average filters. The proposed work is verified by various simulation results.

Keywords: Control Theory and Applications
Abstract: This paper addresses the design problem of disturbance observers for first-order uncertain linear plants. The standing assumption is that the disturbance acting on the plant is modeled as an output of a second-order anti-stable linear system; for example, the disturbance can be sinusoidal or polynomial-in-time of degree less than two. While conventional disturbance observers are capable of attenuating the effect of disturbances on the closed-loop system, this paper primarily focuses on the asymptotic rejection of disturbances and hence, the internal model principle is explicitly incorporated into the proposed design. A systematic design procedure is provided to ensure the robust stability of the resulting closed-loop system.

Keywords: Autonomous Vehicle Systems
Abstract: 。 道路上の物体と運転中の物体までの距離 自動車の完全自動運転を実現するクルマ。 私たちの手法は、単眼カメラのみを使用し、 目の高さに基づいた簡単な計算で。したがって、 処理負荷とコストを削減できます。この論文では、 必要な消失点検出方法が考慮される 私たちの方法の目の高さを推定します。によってキャプチャされた画像 走行中に車両に搭載されたカメラには物体が含まれている さまざまな形や方向の。したがって、すべての 画像内の線 by Hough 変換と発見 その状態で交差点を判別するのは難しい たくさんあるので消失点です 交差点。次に、 走行中のカメラの向きは基本的に平行です 道路へ。に基づいて消失点を&

Keywords: Robotic Applications, Robot Mechanism and Control
Abstract: Previous calibration techniques have often depended on high-precision end-effector tracking devices, such as laser trackers, which are costly and impractical in constrained or field environments. In the absence of such devices, measurement errors are more likely to occur. Even with precision instruments like laser trackers, external disturbances, such as illuminations, can cause measurement noise. These noise factors degrade the reliability of conventional kinematic calibration methods. To overcome these limitations, this study proposes a calibration approach that integrates the Quasi-Differential (QD) method with Competitive Swarm Optimization (CSO). CSO serves as a global optimization framework, efficiently exploring the solution space while preserving search diversity and avoiding premature convergence to local minima. The proposed method reduces the required number of calibration poses and improves accuracy in identifying axis deviations. Its effectiveness and robustness under measurement noise are validated through simulation on a six-degree-of-freedom serial robot.

Keywords: Artificial Intelligence Systems, Industrial Applications of Control, Information and Networking
Abstract: This study presents a multivariate LSTM-Attention-VAE model for anomaly detection in ionic piston hydrogen compressors. Ionic piston compressors hydraulically actuate ionic-liquid pistons to compress flammable hydrogen gas, requiring accurate anomaly detection for safe operation. Multivariate sensor data—including pressure, temperature, flow rate, current, and voltage—are collected via Serial, MQTT, and Modbus-TCP interfaces at 7 Hz. The proposed unsupervised model, trained exclusively on normal data, encodes sensor data windows using LSTM and refines latent representations through a multi-head attention mechanism highlighting critical temporal and inter-sensor dynamics. Reconstruction errors quantify deviations from normal behavior, serving as robust anomaly indicators. Experiments with artificially injected anomalies based on the Westgard rule confirmed the model’s effectiveness, demonstrating high sensitivity with minimal false alarms. Future work aims to integrate this anomaly detection framework into real-time monitoring systems for safer and more reliable hydrogen refueling operations

Keywords: Robotic Applications
Abstract: This paper presents the development and field evaluation of a digital twin-based snow removal support system designed for deep-snow, low-visibility environments. The system reconstructs a snow-free 3D terrain using pre-surveyed data and visualizes the real-time position and orientation of snow removal vehicles using RTK-GNSS and IMU-based localization. To enhance operational safety and efficiency, two key features were implemented: multi-vehicle visualization and a step-cutting construction model. The multi-vehicle visualization enables operators to monitor the positions of all vehicles in a virtual environment, facilitating coordination in collaborative tasks. The step-cutting construction model provides visual guidance based on the target slope geometry, assisting operators in performing precise cutting operations that were previously dependent on experience. A field experiment was conducted with three bulldozers operating over two days in mountainous regions. The results demonstrate the system’s effectiveness in supporting safe and efficient snow removal in complex.

Keywords: Robot Mechanism and Control, Robotic Applications
Abstract: This study investigates the manipulation of pouch packages using a method we call caging-like grasping. This form of grasping can manipulate pouch packages simpler and cheaper than conventional methods of manipulation. In this method, the robotic hand holds the object by touching it, but also incorporates caging by using the hands to clamp the object as much as geometrically possible. This allows the robots to immobilize the packages geometrically and handle them as if they are rigid bodies. In turn, the pouch package would no longer fall since its deformation during operation is impossible. The only thing needed to establish caging-like grasping is the hand’s minimum opening width that is geometrically possible. To determine the required opening width, we created 3D shape models of real pouch packages using B-spline surfaces. This was accomplished by optimizing the shape to gain the minimum width of the robotic hands when it is clamped. After conducting physical experiments with a robotic jaw to validate the minimum width of the robotic hands when clamped, we demonstrated that using these models to find the opening width for robotic hands is effective, as the errors between the models and reality were small enough to achieve caging-like grasping in practice.

Keywords: Artificial Intelligence Systems, Robot Vision, Biomedical Instruments and Systems
Abstract: Lung cancer is the leading cause of cancer-related deaths worldwide. Chest CT imaging is a critical tool for early detection of lung cancer. However, the large number of CT slices can overwhelm radiologists, increasing the risk of overlooking something due to fatigue or inexperience. Computer-aided diagnosis (CAD) systems are increasingly being used as a "second opinion" to support clinical decision-making and address this issue. In this study, we propose a CAD system that uses temporal subtraction images generated from current and prior CT scans. This system emphasizes temporal changes, such as emerging pulmonary nodules, while suppressing stable structures, like vessels and bones. We extract radiomics features from these images. Radiomics features are high-dimensional, quantitative descriptors of texture, shape, and intensity. To avoid overfitting and reduce redundancy, we use Lasso regression to automatically select the most informative features. Finally, a neural network (ANN) performs classification. When applied to 21 cases, our method, which combines features from the original images and the wavelet transformed images, achieved a true positive rate (TPR) of 74.8%, a false positive rate (FPR) of 30.6%, and an accuracy rate of 72.1%. These results demonstrate the effectiveness of frequency-based and shape-aware features in nodule detection.

Keywords: Artificial Intelligence Systems, Biomedical Instruments and Systems
Abstract: Many people are concerned about the way improving their teeth aesthetics. Many approaches have been proposed to change the appearance of teeth, but all of them have the problem that it is difficult to provide a detailed picture of the post-treatment result before the aesthetic dental treatment. To overcome this problem, we propose a new tool that objectively suggests a method for evaluating suitable tooth color to the patient. Specifically, we extract the tooth regions from the whole face of the patient in video images, make color-tone changes, and create new video images of patients after treatment. The video images are satisfied by possible practical medical environment. However, the dental region is small and difficult to extract directly. Therefore, we limit the search region of the dental region by using a lip rectangle region. In this paper, we use YOLO11 as the base model and improve it by including Attention Gate. After extraction, we apply post-processing such as inter-fame processing to improve the extraction accuracy, and evaluate their effectiveness.

Keywords: Artificial Intelligence Systems, Information and Networking, Biomedical Instruments and Systems
Abstract: In recent years, there has been an increasing interest in the aesthetics of teeth, particularly regarding tooth color. Among the various approaches to improve the appearance of teeth, crown restoration treatments such as laminate veneers are widely practiced. However, if the result does not meet the expectations of patient, the treatment may not be considered successful. Since the ideal tooth color varies with age and skin tone, patients often have difficulty imagining the result, leading to anxiety or dissatisfaction. To address this problem, we propose an image processing system that generates simulated post-treatment images from facial or intraoral photographs. This system allows both the patient and the dentist to share a clear vision of the expected outcome. Accurate extraction of the tooth region is essential, as global color transformation would undesirably affect surrounding regions such as skin and untreated teeth. Manual extraction is impractical in clinical settings, highlighting the need for automation. To detect individual teeth, we employ Mask R-CNN, a deep learning-based instance segmentation model. We use ResNet-50 as the backbone and incorporate Deformable Convolutional Networks (DCN) to improve adaptability to different tooth shapes and scales. We evaluate the performance of the model and confirm its effectiveness for aesthetic dentistry.

Keywords: Artificial Intelligence Systems, Information and Networking, Biomedical Instruments and Systems
Abstract: Computed tomography (CT) is a medical imaging procedure that uses X-rays from different angles to create clear cross-sectional images of the body. Although CT is especially useful, it exposes patients to radiation, which can increase the risk of health problems like cancer. To reduce this risk, low-dose CT (LDCT) is used, but this makes the images noisier and less clear, which can make diagnosis more difficult. Traditional methods to clean up noisy images usually require both normal-dose and low-dose image pairs, but obtaining such data is difficult and raises ethical concerns. In this study, we propose a self-supervised Neighbor2Neighbor denoising method that uses only single low-dose images for training. We use ResUNet as the base model and build three improved versions by adding Efficient Channel Attention (ECA) and Pyramid Pooling Module (PPM). We evaluated the models on whole-body CT images of piglets acquired with only 10% of the usual radiation dose [1]. We measured image quality using PSNR and SSIM compared to normal dose images. The results show that all our models perform better than the originalUNet and ResUNet, with less noise and clearer images.

Keywords: Navigation, Guidance and Control, Robotic Applications, Autonomous Vehicle Systems
Abstract: This paper presents a collaborative path planning method for multiple underwater gliders (UGs) aimed at energy-optimal task execution. Paths are encoded using path points, pitch angles, and dive depths as decision variables. An energy cost fitness function is then formulated by integrating the UG energy consumption model with ocean environmental data. To identify high-quality solutions, a multi-task differential evolution algorithm with adaptive knowledge transfer (MTDE-AKT) is proposed as the optimizer. MTDE-AKT employs a multi-population framework in which each subpopulation is assigned to a specific UG path planning task. To enhance co-evolution, an adaptive knowledge transfer mechanism is designed, enabling subpopulations to probabilistically acquire beneficial information from others during the mutation process. Our method is evaluated using two sets of task scenarios and benchmarked against four latest multi-task evolutionary algorithms. Experimental results prove that MTDE-AKT achieves the lowest energy consumption and significantly outperforms the comparison algorithms in statistical tests. Moreover, ablation studies confirm the effectiveness of the adaptive knowledge transfer mechanism.

Keywords: Navigation, Guidance and Control, Autonomous Vehicle Systems, Robotic Applications
Abstract: We revisit the Particle Filter (PF) solution to robot localization and present an alternative approach to the calculation of weights of particles; this crucial step determines which particles are retained and resampled and which are depleted. The weights of particles (with random pose values) are determined as the inverse of the power of a vector norm, defined in a space spanned by the angles of the sensor at which range measurements are made. Those with high weights (low norm value) are retained as they have a pose close to the actual robot. This step does not require a probability density function to define the sensor model. In addition, to prevent sudden depletion of particles, those with high weights are roughened using distortions emerging from a uniform distribution. Experimental results show that this approach leads to convergence of the particles to the true pose of the robot.

Keywords: Navigation, Guidance and Control, Autonomous Vehicle Systems, Artificial Intelligence Systems
Abstract: This paper proposes a reinforcement learning-based path planning framework that enables mobile robots to autonomously navigate narrow space environments by avoiding obstacles and proactively seeking safe open spaces. The proposed method encodes spatial information, such as the relative distance and motion between the robot and surrounding obstacles, into a compact state representation that enhances environmental awareness. The reward function is carefully designed to promote goal reaching, avoidance of both static and dynamic obstacles, and preference for safe and navigable directions. The learning and evaluation were conducted in a simulated environment built with NVIDIA Isaac Sim, which replicates complex scenarios including human-robot coexistence. Unlike traditional planners which are limited in handling dynamic obstacles, the trained policy demonstrates reliable performance by avoiding dynamic obstacles and reaching the goal without collisions. This work provides a practical foundation for safe robot navigation in real-world narrow environments.

Keywords: Navigation, Guidance and Control
Abstract: Three-axis slew maneuvers of spacecraft are used in a variety of missions, including observation tasks. In such maneuvers, accounting for the spacecraft’s constraints on angular velocity and angular acceleration and minimizing the maneuver time are crucial for mission efficiency. This paper addresses the approximate minimum time for three-axis slew maneuvers of spacecraft. In this context, the spacecraft is subject to spin-to-spin boundary conditions and is constrained by limits on body angular velocity and angular acceleration. In general, three-axis spin-to-spin maneuvers of spacecraft are non-eigen-axis, and therefore no analytical solution exists. Therefore, for onboard implementation, it is necessary to derive an approximate minimum-time estimate. To do this, the optimal minimum time is first obtained by employing the commercial optimization solver, General Purpose Optimal Control Software (GPOPS). Then, we transform the maneuver under the same boundary conditions into an equivalent eigen-axis rotation and calculate its maneuver time using the known analytical solution. Finally, we compare and analyze the trends between the optimal time computed by the solver and the eigen-axis approximation across the specified simulation cases. The difference between the two results was found to be, on average, approximately 0.02~0.03 s depending on the case.

Keywords: Navigation, Guidance and Control, Sensors and Signal Processing, Control Theory and Applications
Abstract: This article presents VALINOR (Velocity-Aided Leg Inertial Nonlinear Odometry and Registration), a method for Leg-Inertial odometry for humanoid robots addressing the challenge of lightweight yet accurate and certifiable state estimation. VALINOR associates Leg odometry with the Tilt Observer, a computationally efficient complementary filter, which provides accurate estimates of the IMU's tilt and linear velocity with strong mathematical convergence guarantees. We introduce an axis-agnostic method for the fusion of the Leg odometry's yaw with the Tilt Observer's tilt estimate. We argue that this method is less arbitrary and more mathematically sound than those based on other orientation representations, especially on Euler angles. We present an evaluation of the proposed estimator through real-world data on two humanoid robots. We show that, while being 7.5 times faster than the state-of-the-art method used for comparison, VALINOR improves tilt estimation by over 25%, making it a well-suited feedback for balance and walking controllers.

Keywords: Robot Mechanism and Control, Sensors and Signal Processing, Robotic Applications
Abstract: This paper presents a novel spring design that enhances the Torque Capacity-to-Stiffness without increasing actuator volume or introducing morphological complexity. The proposed design employs multiple cantilever beams arranged in parallel, enabling the stiffness to be preserved while increasing torque capacity through beam thinning. Analytical expressions for stiffness and torque capacity are derived based on cantilever beam theory, and the theoretical relationship between the number of beams and the Torque Capacity-to-Stiffness is established. To validate the proposed methodology, Finite Element Analysis (FEA) was performed on ten spring samples with increasing beam numbers. The results confirm that higher beam counts lead to improved torque capacity under fixed stiffness conditions.

Keywords: Robot Mechanism and Control, Control Theory and Applications, Artificial Intelligence Systems
Abstract: Advancements in optimization solvers and computing power have led to growing interest in applying whole-body model predictive control (WB-MPC) to bipedal robots. However, the high degrees of freedom and inherent model complexity of bipedal robots pose significant challenges in achieving fast and stable control cycles for real-time performance. This paper introduces a novel kino-dynamic model and warm-start strategy for real-time WB-MPC in bipedal robots. Our proposed kino-dynamic model combines the linear inverted pendulum plus flywheel and full-body kinematics model. Unlike the conventional whole-body model that rely on the concept of contact wrenches, our model utilizes the zero-moment point (ZMP), reducing baseline computational costs and ensuring consistently low latency during contact state transitions. Additionally, a modularized multi-layer perceptron (MLP) based warm-start strategy is proposed, leveraging a lightweight neural network to provide a good initial guess for each control cycle. Furthermore, we present a ZMP-based whole-body controller (WBC) that extends the existing WBC for explicitly controlling impulses and ZMP, integrating it into the real-time WB-MPC framework. Simulations and real robot experiments further validate that the proposed framework demonstrates robustness to perturbation and satisfies real-time control requirements during walking.

Keywords: Robot Mechanism and Control, Control Theory and Applications, Sensors and Signal Processing
Abstract: スネークのようなロボットは、車輪付きロボットよりも機動性が高くなります。しかし、関節を駆動するサーボモーターが多いため、生体ヘビのような起伏のある移動を再現すると消費電力が増加するという問題があります。私たちは、前述の問題に対処するために、ロボットのナビゲーションに重心(COG)シフトを統合する効率的な方法を提案します。提案手法では、ヘビのようなロボットをタイヤの形に変形させて並列二輪車を実現し、リンクを変形させることでCOGを変化させて転がり運動を発生させる。この実装により、平地か荒れた地面かに応じて、転がり移動と起伏のある移動のどちらかを選択できます。これまでの研究&#

Keywords: Robot Mechanism and Control, Robotic Applications, Industrial Applications of Control
Abstract: Collaborative manipulations in industrial setups such as assembly lines and warehouse management are essential for advanced automation. This paper presents a mathematically robust, kinematics-based method for real-time distance calculation between manipulator links without requiring external sensors. We first derive a comprehensive geometric framework for multi-arm systems that classifies link configurations as parallel, intersecting, or skew, enabling precise minimum distance calculations through line segment representation. The general n-arm mathematical framework is then specialized for dual-arm systems, focusing on collision-critical segments in table-mounted configurations. Our approach constructs distance matrices for systematic collision monitoring and reduces computational overhead by targeting inter-arm link pairs. Real-time hardware validation on dual Kinova Gen3 Lite manipulators demonstrates the method’s effectiveness for computing inter-arm distances in shared workspace operations, confirming reliable collision avoidance performance in collaborative robotic tasks. The scalable mathematical foundation combined with efficient dual-arm implementation enables safe and effective collaborative automation in industrial applications.

Keywords: Robot Mechanism and Control
Abstract: This paper presents the design and evaluation of the LUTA Hand, a lightweight, underactuated, three-fingered robotic hand capable of adaptive grasping. The hand employs a shared tendon-routing scheme to coordinate flexion of three fingers using a single actuator, while rolling contact joints and elastic ligaments provide passive restoring forces, eliminating the need for antagonistic actuation. All components are compactly integrated, resulting in a portable and self-contained system. Experimental results demonstrate consistent finger movement, fingertip force, and demonstration of grasping a range of objects and shapes, enabled by the adaptive grasping capability. These findings highlight the potential of combining minimal actuation with compliant mechanisms for efficient and versatile robotic hands.

Keywords: Robot Mechanism and Control, Robotic Applications
Abstract: Kinematic validation, structural optimization, and gait planning are needed to build compact bipedal robots. This work designs, models, and simulates a 6-DOF bipedal robot for kinematic analysis and trajectory development. Joint configurations for stable locomotion are computed using the Denavit-Hartenberg (D-H) convention and geometric inverse kinematics. Smooth joint trajectories are created using cubic spline interpolation. MATLAB Simscape simulations verify motion feasibility and provide joint velocity patterns and torque needs. Topology optimization under gravitational self-weight loading assumptions reduces weight without compromising stability, improving mechanical efficiency. Dynamic stability is qualitatively examined by studying the Center of Mass (CoM) trajectory and determining the Zero Moment Point (ZMP) using the Linear Inverted Pendulum Model (LIPM). Simulations show steady gait patterns, synchronized joint motions, appropriate torque distribution across the ankle, knee, and hip joints, and a CoM-ZMP trajectory deviation of less than 6 mm, confirming gait stability. This research bridges theoretical modeling with real-world applications to produce agile, energy-efficient bipedal robots for humanoid locomotion.

Keywords: Control Theory and Applications, Navigation, Guidance and Control, Industrial Applications of Control
Abstract: Quadcopter UAV formations have been widely employed across diverse fields. However, UAV formations face challenges such as random and varied actuator failures, stringent tracking error constraints, and restricted power redundancy, which have significantly hindered their practical application. To address these issues, this paper presents a distributed fault-tolerant control strategy for UAV formations based on active fault-tolerant control allocation (AFTCA). By capitalizing on known fault information, the control efficiency matrix of UAVs is reconstructed. The control matrix weights are dynamically adjusted in response to real-time fault information, leading to the acquisition of a fault-tolerant control allocation matrix that enables effective fault-tolerant control allocation for quadcopter UAVs. The proposed algorithm has been contrasted with several existing algorithms, and the results starkly demonstrate its superiority in fault-tolerant flight under various single faults. Moreover, hardware-in-the-loop (HIL) formation flight experiments have been conducted. The experimental results substantiate the efficacy of the proposed algorithm in achieving fault-tolerant flight missions for quadcopter UAV formations under compound faults.

Keywords: Control Theory and Applications
Abstract: A non-iterative data-driven tuning method is proposed for linear fractional-order controllers. The formulation of the tuning process adopts a model reference control problem structure and is subsequently reformulated as a numerical optimization problem on the basis of a fictitious reference signal. This signal is computed using the controller under evaluation and a single set of input and output data from the controlled plant. We analyze the effect of noise corrupting the data to the data-based optimization problem. To mitigate noise corrupting the data used for controller tuning, the data is preprocessed using the L2 total variation regularization technique. The proposed approach is simple in that it requires neither mathematical modeling nor repeating closed-loop control tests. The incorporation of the L2 total variation denoising improves the noise tolerance of the proposed tuning technique. The validity of the proposed approach is demonstrated through numerical simulations.

Keywords: Control Theory and Applications
Abstract: Hidden attractors represent a new interesting topic in chaos. These hidden attractors have a basin of attraction that does not intersect with small neighborhoods of any equilibrium points. Oscillations in dynamical systems can be easily localized numerically if initial conditions from its open neighborhood lead to a long-time oscillation. For example, hidden attractors are attractors in systems with no equilibria or with only one stable equilibrium, or with a special case of multi-stability and coexistence of attractors. In this paper, we study the reset control of the chaotic systems with hidden attractors. We think chaotic systems with an equilibrium point and without the equilibrium points, and we design the reset controller to the stable response for this chaotic system by numerical simulations. It is more difficult to control chaos with Hidden Attractor than widely known chaos, and various phenomena, such as jump phenomena, are shown to occur.

Keywords: Control Devices and Instruments, Industrial Applications of Control
Abstract: This paper presents the implementation of a real-time neural network controller to compensate for the uncertainties of a 2 axis-gimbal system. The accurate attitude control performance can be improved by neural network controller in online fashion, which does not require offline learning. The neural network outputs are added at the trajectory level to reject the disturbance for the better attitude control performance. Experimental studies are performed to verify the outperformance of the neural network controller.

Keywords: Control Devices and Instruments, Human-Robot Interaction, Sensors and Signal Processing
Abstract: This paper proposes a novel human-intent-driven assistive control framework for mobile transport robots that aims to support collaborative physical interaction without relying on dedicated force sensors. Conventional autonomous mo-bile robots (AMRs) suffer from high initial costs, poor adaptability in complex environments, and limited capability for human cooperation. To address these limitations, the proposed method estimates user input using only wheel en-coder measurements and employs a force observer combined with a classification algorithm to separate user-intended signals from external disturbances in real time. The classified user force is converted into compliant motion com-mands using a virtual impedance model and then integrated into a Power Assistive Controller (PAC) that generates final motion commands while actively suppressing disturbances. Comparative simulation results demonstrate that the proposed system enables stable and accurate motion assistance, selectively amplifying user intent while rejecting en-vironmental interference. The system's simple hardware structure, low implementation cost, and robustness in un-structured environments make it a practical solution for assistive mobility applications, particularly for elderly and physically impaired users.

Keywords: Sensors and Signal Processing, Industrial Applications of Control
Abstract: This paper presents a state estimation algorithm for permanent magnet synchronous motors (PMSMs) using the interacting multiple model-kalman Filter (IMM-KF). Accurate state estimation based on output feedback is a significant challenge in PMSM control due to the system’s inherent nonlinearities. To address this, the nonlinear state space is decomposed into multiple linear submodels, each defined around specific operating points. This method, referred to as polytopic decomposition, allows each submodel to represent local dynamics more accurately. Each linear submodel is paired with an individual Kalman filter, and the IMM-KF algorithm combines their outputs by evaluating the likelihood of each model with respect to the current measurements. The interaction among multiple filters enables dynamic adjustment of estimation weights, improving accuracy and robustness compared to conventional single-model approaches. Simulation results demonstrate that the proposed method provides reliable estimation of key internal states, including stator currents and rotor speed, in response to gradual variations in input frequency. The estimator maintains high performance using only limited output signals, making it suitable for PMSM applications that require reliable state information under constrained sensing environments.

Keywords: Navigation, Guidance and Control, Autonomous Vehicle Systems, Artificial Intelligence Systems
Abstract: Long-term path planning in large-scale outdoor environments is a critical task for autonomous driving. Path planning algorithms require outstanding computational efficiency and real-time responsiveness. However, due to the complexity of considering terrain and obstacles in 3D map, computational requirements increase significantly as the scale expands. To address these challenges, this paper proposes RUSH (Recursive and Scalable 3D Coarse To Fine Path Planning). First, RUSH introduces a hierarchical Coarse To Fine approach that breaks down large-scale path planning problems into smaller segments, maximizing computational efficiency through parallel processing. It then leverages a diffusion model to rapidly generate paths even in complex, multi-resolution environments. Extensive evaluations on diverse, large-scale 3D datasets demonstrate that RUSH achieves path planning approximately 565 times faster than the 3D A* algorithm without Coarse To Fine method, while maintaining high-quality and feasible paths. These results highlight RUSH’s potential as a practical solution for real-time, long-range path planning in autonomous systems.

Keywords: Robot Vision
Abstract: Simultaneous Localization and Mapping (SLAM) with dense representation is a key technology in robotics, Virtual Reality (VR), and Augmented Reality (AR). Recent developments in this field emphasize the advantages of neural scene representations and 3D Gaussian models for achieving high-fidelity spatial understanding. In this paper, we introduce a novel dense SLAM framework that integrates Generalized Iterative Closest Point (G-ICP) with 3D Gaussian Splatting (3DGS). Unlike prior methods, our approach employs a single unified Gaussian map for both tracking and mapping, enabling mutual enhancement. By exchanging covariance information and applying scale alignment between tracking and mapping, we reduce computational redundancy and improve system efficiency. Furthermore, our tailored keyframe selection strategy boosts both tracking robustness and mapping accuracy. Experimental results validate the effectiveness of our approach, demonstrating outstanding runtime performance with speeds reaching up to 107 FPS and delivering high-quality 3D reconstructions.

Keywords: Autonomous Vehicle Systems, Robotic Applications, Navigation, Guidance and Control
Abstract: .자율 주행 기술이 빠르게 채택되었습니다. 군사, 건축에서 사용되는 무거운 장비, 및 오프로드에서 작동하는 농업 응용 분야 환경. 그러나 지형이 불규칙하고 빈번한 지면 상태의 변화는 종종 반복적인 미끄러짐을 유발합니다 이벤트를 발생시켜 누적 오류를 악화시키고 다음을 기반으로 한 single-sensor localization 방법의 신뢰성 주행 거리 측정 또는 GPS. 이러한 문제를 해결하기 위해 센서 융합 확장 칼만 필터(EKF)와 같은 접근 방식, 입자 필터(PF) 및 적응형 칼만 필터(adaptive Kalman Filters)가 있습니다. 제안; 그럼에도 불구하고, 계층적 비선형을 동시에 고려하는 아키텍처 스키드 스티어링 차량의 조향 특성과 오프로드 지형의 복잡한 불확실성은 여전히 부족합니다. 안으로 특히 작업 현장 환경은 종종 미끄러짐을 유

Keywords: Robot Vision
Abstract: A depth image was generated by fusing high-resolution camera images with LiDAR distance information. A CNN-based model was used to compensate for the sparse LiDAR point cloud and reconstruct accurate depth information

Keywords: Robot Vision
Abstract: This study applied GLIM, a lightweight LiDAR-based SLAM algorithm, for field robot localization and implemented real-time visualization and remote monitoring using the Foxglove tool

Keywords: Autonomous Vehicle Systems, Robot Vision, Sensors and Signal Processing
Abstract: Autonomous ground vehicles (AGVs) are increasingly being deployed in unstructured environments such as agricultural fields, construction sites, and mining areas, where terrain complexity poses significant challenges for navigation. A key requirement for reliable off-road autonomy is the accurate estimation of traversable areas. Although LiDAR has been widely adopted for this purpose due to its 3D perception capability, its high cost, limited resolution, and hardware complexity hinder widespread adoption in practical, cost-sensitive industrial applications. To address these limitations, this study proposes a real-time traversability estimation method based on depth images acquired from a monocular camera. The proposed approach eliminates the need for expensive sensing hardware and enables efficient terrain assessment with a lightweight perception pipeline. It is designed to operate effectively in diverse off-road conditions while maintaining computational efficiency, making it suitable for integration into low-cost robotic platforms. This work contributes a scalable and cost-effective alternative to LiDAR-based systems, with potential applications across a broad range of off-road automation scenarios. The proposed method lays the foundation for more accessible and adaptable off-road autonomous systems.

Keywords: Navigation, Guidance and Control
Abstract: This paper proposes an Adaptive Maximum Correntropy Unscented Kalman Filter (A-MCUKF) for GNSS/INS navigation in urban environments, where signal degradation due to multipath and non-line-of-sight conditions poses significant challenges. Traditional Kalman filters assume Gaussian noise, rendering them susceptible to non-Gaussian disturbances and outliers. To address this limitation, the proposed method introduces a novel adaptive kernel model that combines Gaussian and Cauchy kernels using a weighting function dependent on the error magnitude. The kernel bandwidth is dynamically adjusted via a maximum likelihood approach derived from the Student-t distribution, enabling real-time adaptation to varying noise characteristics. This adaptive kernel is integrated into the Unscented Kalman Filter framework to enhance robustness against outliers while preserving the information from normal data. The algorithm is validated using real-world urban navigation data collected from an ADIS16448 IMU and a Ublox F9P GNSS receiver on a Cortex-M4-based MCU platform. Experimental results across diverse urban scenarios demonstrate that the proposed A-MCUKF outperforms conventional EKF, UKF, and fixed-kernel MCUKF variants, particularly in multipath-rich areas. The method exhibits superior performance in both position and yaw estimation, confirming its effectiveness for robust and adaptive state estimation in challenging urban navigation environments.

Keywords: Navigation, Guidance and Control, Autonomous Vehicle Systems, Robot Vision
Abstract: In the development of an autonomous driving system, it is essential to find accurate transformation between an embedded GNSS/INS (EGI), which measures global navigation information, and cameras that visually capture the surrounding environment. Since EGI cannot directly perceive the surrounding environment but only measures the motion of the ego vehicle, extrinsic calibration between the two sensors is typically accomplished by comparing the motion of each sensor. However, the conventional motion-based methods that rely solely on comparing camera motion estimated from just two views overlook multi-view constraints, resulting in suboptimal accuracy. In this study, we propose a refined EGI-camera calibration pipeline that comprises a coarse calibration step using an improved motion-based method and a fine calibration step that determines extrinsic parameters to minimize reprojection error directly without calculating camera motion. Through Monte Carlo simulations and real vehicle tests, we demonstrate that the proposed pipeline yields more accurate results than the conventional motion-based method.

Keywords: Navigation, Guidance and Control, Sensors and Signal Processing, Robotic Applications
Abstract: This paper presents a localization approach based on an Error-state Kalman Filter (EsKF) that integrates Ultra Wide Band (UWB), Radar, and Inertial Measurement Unit (IMU) sensors for use in harsh indoor environments. Radar was employed to overcome the environmental limitations of conventional cameras and LiDAR, while UWB was utilized to correct the accumulated navigation errors. However, UWB is difficult to employ in environments without prior information. Therefore, we address the infrastructure-less environments where no prior information is available about the UWB anchors’ positions by proposing a particle filter-based method to efficiently estimate the unknown anchor positions. By fusing the two proposed techniques, robust localization is achieved even in challenging indoor scenarios without prior knowledge of the UWB anchor positions. To validate the system, Radar and IMU measurements were collected through experiments, and simulated UWB anchor measurements were generated based on the experimental ground truth. The proposed method improves both the performance and practical applicability of localization in indoor and infrastructure-less environments.

Keywords: Navigation, Guidance and Control, Artificial Intelligence Systems, Autonomous Vehicle Systems
Abstract: This paper proposes a transformer-based particle resampling technique to improve visual odometry accuracy for autonomous vehicles. To address the particle diversity loss problem of conventional systematic resampling, we developed a hybrid resampling strategy combining weight-aware multi-head attention with variational autoencoders. Experimental results on the KITTI dataset demonstrate that the proposed method achieves 15.3 % reduction in position error, 10.8 % reduction in rotation error, and 22.1 % improvement in particle variance compared to conventional methods.