Keywords: Soft Robotics, Design, Grasping and Manipulation
Abstract: Rapid and versatile object manipulation in air is an open challenge. An energy-efficient and adaptive soft gripper combined with an agile aerial vehicle could revolutionize aerial robotic manipulation in areas such as warehousing. This paper presents a bio-inspired gripper powered by hydraulically amplified electrostatic actuators mounted to a quadcopter that can interact safely and naturally with its environment. Our gripping concept is motivated by an eagle's foot. Our custom multi-actuator concept is inspired by a scorpion tail design (consisting of a base electrode with pouches stacked adjacently) and spider-inspired joints (classic pouch motors with a flexible hinge layer). A hybrid of these two designs realizes a higher force output under moderate deflections of up to 25° compared to single-hinge concepts. In addition, sandwiching the hinge layer improves the robustness of the gripper. For the first time, we show that soft manipulation in air is possible using electrostatic actuation. This study demonstrates the potential of untethered hydraulically amplified actuators in aerial robotic manipulation. Our proof of concept opens up the use of hydraulic electrostatic actuators in mobile aerial systems.

Keywords: Humanoid Robot Systems, AI-enabled Robotics, Bioinspired Robotics
Abstract: Reinforcement learning recently shows great progress on legged robots, while bipedal robots in high dimensions but narrow solution space are still challenging to learn. The typical methods introduce the reference joints motion to guide the learning process; however, obtaining a high-quality reference trajectory is nontrivial, and imitation suffers from the local minimum. For general reference-free scenarios, the bipedal robot is discouraged by the early termination and biased sample collection. Inspired by the assistive learning commonly shown in biped animals, we introduce the assistive force to aid the learning process without the requirement of reference trajectories. The learned assistant could be a curricula to lead motor skills learning and is eliminated in the end to shape the learned motion to be plausible. We analyze the assistive system and verify its effectiveness in multiple challenging bipedal skills.

Keywords: Robot Learning, Soft Robotics, Design
Abstract: Roboy 3.0 is an open-source tendon-driven humanoid robot that mimics the musculoskeletal system of the human body. Roboy 3.0 is being developed as a remote robotic body - or a robotic avatar - for humans to achieve remote physical presence. Artificial muscles and tendons allow it to closely resemble human morphology with 3-DoF neck, shoulders and wrists. Roboy 3.0’s 3-DoF joints are implemented as ball-and-socket joints. While industry provides a clear solution for 1-DoF joint pose sensing, it is not the case for the ball-and-socket joint type. In this paper we present a custom solution to estimate the pose of a ball-and-socket joint. We embed an array of magnets into the ball and an array of 3D magnetic sensors into the socket. We then, based on the changes in the magnetic field as the joint rotates, are able to estimate the orientation of the joint. We evaluate the performance of two neural network approaches using the LSTM and Bayesian-filter like DVBF. Results show that in order to achieve the same MSE DVBFs require significantly more time training and hyperparameter tuning compared to LSTMs, while DVBF cope with sensor noise better. Both methods are capable of real-time joint pose estimation at 37 Hz with MSE of around 0.03 rad for all three degrees of freedom combined. The LSTM model is deployed and used for joint pose estimation of Roboy 3.0's shoulder and neck joints. The software implementation and PCB designs are open-sourced under https://github.com/Roboy/ball_and_socket_estimator

Keywords: Robot Learning, Grasping and Manipulation
Abstract: We present BulletArm, a novel benchmark and learning-environment for robotic manipulation. BulletArm is designed around two key principles: reproducibility and extensibility. We aim to encourage more direct comparisons between robotic learning methods by providing a set of standardized benchmark tasks in simulation alongside a collection of baseline algorithms. The framework consists of 31 different manipulation tasks of varying difficulty, ranging from simple reaching and picking tasks to more realistic tasks such as bin packing and pallet stacking. In addition to the provided tasks, BulletArm has been built to facilitate easy expansion and provides a suite of tools to assist users when adding new tasks to the framework. Moreover, we introduce a set of five benchmarks and evaluate them using a series of state-of-the-art baseline algorithms. By including these algorithms as part of our framework, we hope to encourage users to benchmark their work on any new tasks against these baselines.

Keywords: Robot Learning, Control
Abstract: The motion of an autonomous underwater vehicle (AUV) is affected by its surrounding water flows, so an accurate estimation of the flow field could be used to assist the vehicle's navigation. We propose an optimization-based approach to the problem of online flow field learning with limited amounts of data. To compensate for the shortage of online measurements, we identify two types of physically meaningful constraints from eddy geometry of the flow field and the property of fluid incompressibility respectively. By parameterizing the flow field as a polynomial vector field, the optimization problem could be solved efficiently via semi-definite programming (SDP). The effectiveness of the proposed algorithm in terms of flow field estimation is experimentally validated on real-world ocean data by providing performance comparisons with a related method. Further, the proposed estimation algorithm is proved to be able to be combined with a motion planning method to allow an AUV to navigate efficiently in an underwater environment where the flow field is unknown beforehand.

Keywords: Control, Robot Vision
Abstract: At CERN, radiation surveys of equipment and beam lines are important for safety and analysis throughout the accelerator complex. Radiation measurements are highly dependent on the distance between the sensor and the radiation source. If this distance can be accurately established, the measurements can be used to better understand the radiation levels of the components and can be used for calibration purposes. When surveys are undertaken by the Train Inspection Monorail (TIM) robot, the sensor is at a constant distance from the rail, which means that it is at a known distance and height from the centre of the beam line. However, the distance of the sensor to the closest surface of the beam line varies according to what kind of equipment is installed on the beam line at this point. Ideally, a robotic survey would be completed with online adaption of the sensor position according to the equipment present in the LHC. This new approach establishes a scan of the surface with a 2D LiDAR while moving along the tunnel axis in order to obtain a 3D scan of the environment. This 3D scan will be used to generate online trajectories that will allow the robot to accurately follow the beam line and thus measure the radiation levels.