Featured with high power density, excellent flexibility, shock absorption capacity, etc., pneumatic muscles (PMs) promote the development of exoskeleton robots and rehabilitation equipment. However, the complex nonlinearities of PMs limit efficiency optimization in closed-loop control, while the force-displacement coupling, soft materials, deficient workspace, etc., make it more difficult to simultaneously increase motion speeds and ensure the safety of multiple PM-actuated (PMA) robots. Although force sensors can currently be replaced by applying state estimation techniques, the amplification effects of measurement noises still compromise control accuracy and stability in practice. To this end, this article proposes a reinforcement learning-based robust motion control method with the prescribed performance, which achieves efficient and satisfactory tracking control for PMA robotic arms. In particular, by elaborately incorporating an integral term, a robust generalized proportional integral observer is used to eliminate measurement noises. Meanwhile, by using an actor–critic network to optimize control performance, an error-transformation-based continuous controller is designed to guarantee the uniformly ultimately boundedness of tracking errors. Compared with most existing methods, this article provides the first solution to restrict the entire transient and steady-state performance of PMA robotic arms, improve the noise suppression capability, and optimize the control efficiency simultaneously. Finally, complete stability analysis based on Lyapunov techniques is provided, and several groups of hardware experiments demonstrate the practicability and robustness of the proposed method.

Reinforcement Learning-Based Prescribed Performance Motion Control of Pneumatic Muscle Actuated Robotic Arms With Measurement Noises

In a virtual reality (VR) space, wearing a head-mounted display can help with the visualization of objects; however, users cannot experience realistic tactile sensations. Recently, several force feedback devices have been developed, including wearable devices that use straight-fiber-type pneumatic muscles and magnetorheological fluids. This allows the devices to render elastic, frictional, and viscous forces during spatially unrestricted movement. Nevertheless, a problem remains in that previous devices could handle many bilateral upper limb movement tasks. Therefore, this study aims to develop a device that can handle movements that interact with both arms. Based on experiments concerning the pushing motion in a VR space, the influence of the pseudo force sense was determined to not be small. In addition, we confirmed that the force sensation presented by this system was more realistic when the robot was operated with both arms than when operated with the right arm.

Development of a Bimanual Wearable Force Feedback Device with Pneumatic Artificial Muscles, MR Fluid Brakes, and Sensibility Evaluation Based on Pushing Motion

Featured with high power density, excellent flexibility, shock absorption capacity, etc., pneumatic muscles (PMs) promote the development of exoskeleton robots and rehabilitation equipment. However, the complex nonlinearities of PMs limit efficiency optimization in closed-loop control, while the force-displacement coupling, soft materials, deficient workspace, etc., make it more difficult to simultaneously increase motion speeds and ensure the safety of multiple PM-actuated (PMA) robots. Although force sensors can currently be replaced by applying state estimation techniques, the amplification effects of measurement noises still compromise control accuracy and stability in practice. To this end, this article proposes a reinforcement learning-based robust motion control method with the prescribed performance, which achieves efficient and satisfactory tracking control for PMA robotic arms. In particular, by elaborately incorporating an integral term, a robust generalized proportional integral observer is used to eliminate measurement noises. Meanwhile, by using an actor–critic network to optimize control performance, an error-transformation-based continuous controller is designed to guarantee the uniformly ultimately boundedness of tracking errors. Compared with most existing methods, this article provides the first solution to restrict the entire transient and steady-state performance of PMA robotic arms, improve the noise suppression capability, and optimize the control efficiency simultaneously. Finally, complete stability analysis based on Lyapunov techniques is provided, and several groups of hardware experiments demonstrate the practicability and robustness of the proposed method. 

We have been developing a wearable force-feedback device using pneumatic artificial muscles and magnetorheological fluid brakes. In a previous study, we confirmed that this device can present the stiffness of a virtual object. However, based on the results of a questionnaire administered during the experiment, we considered that the subjects' awareness of the device's weight suppressed the reality of the VR experience. Therefore, we redesigned the device to reduce the physical burden on the user. From simulation results, we confirmed that this redesign reduced the shoulder joint torque required for flexion and increased the workspace of the force-feedback device. In addition, the torque transmission characteristics of the link mechanism involved in horizontal shoulder motion was tested. We also measured the vertical load applied to the user's hand during horizontal shoulder motion. The results show that the device transmits torque close to the target value in the horizontal flexing direction but does not reach the target value in the horizontal extending direction. We also confirmed that the vertical load during horizontal motion could be reduced by redesigning the device.

Development of Shoulder Mechanism for a Wearable Upper Limb Force-Feedback Device with Pneumatic Artificial Muscles and MR Brakes

We have been developing a wearable force-feedback device using pneumatic artificial muscles and magnetorheological fluid brakes. In a previous study, we confirmed that this device can present the stiffness of a virtual object. However, based on the results of a questionnaire administered during the experiment, we considered that the subjects' awareness of the device's weight suppressed the reality of the VR experience. Therefore, we redesigned the device to reduce the physical burden on the user. From simulation results, we confirmed that this redesign reduced the shoulder joint torque required for flexion and increased the workspace of the force-feedback device. In addition, the torque transmission characteristics of the link mechanism involved in horizontal shoulder motion was tested. We also measured the vertical load applied to the user's hand during horizontal shoulder motion. The results show that the device transmits torque close to the target value in the horizontal flexing direction but does not reach the target value in the horizontal extending direction. We also confirmed that the vertical load during horizontal motion could be reduced by redesigning the device. 

In a virtual reality (VR) space, wearing a head-mounted display can help with the visualization of objects although users cannot experience realistic tactile sensations. Recently, several force feedback devices have been developed, including wearable devices that use straight-fiber-type pneumatic muscles and magnetorheological fluids. This allows the devices to render elastic, frictional, and viscous forces during spatially unrestricted movement. However, during friction and viscosity rendering, the elasticity of the artificial muscle is influenced by the elastic element of the muscle, preventing proper presentation of the force. Therefore, this study proposed a clutch mechanism to switch the force presentation of elastic elements by the contraction of artificial muscles. The mechanism was incorporated into a wearable four-degrees-of-freedom force feedback device, and its elastic, frictional, and viscous force feedback performances were quantitatively assessed via fundamental property experiments. Furthermore, a VR space was constructed to present the operator with force perceptions of virtual elastic, frictional, and viscous objects within that space, and the system’s performance was qualitatively assessed. The results confirmed a reduction in the effect of elastic elements and an improvement in the presentation performance. Additionally, the use of this device in the VR space was confirmed to improve the realism of virtual objects in terms of friction and viscosity.

Ryunosuke Sawahashi

Papers

Development of a Wearable Four-Degrees-of-Freedom Force Feedback Device with a Clutch Mechanism Using Artificial Muscle Contraction