Soft robots are considered intrinsically safe with regard to human–robot interaction. This has motivated the development and investigation of soft medical robots, such as soft robotic gloves for stroke rehabilitation. However, the output force of conventional purely soft actuators is usually limited. This restricts their application in stroke rehabilitation, which requires a large force and bidirectional movement. In addition, accurate control of soft actuators is difficult owing to the nonlinearity of purely soft actuators. In this study, a soft robotic glove is designed based on a soft-elastic composite actuator (SECA) that integrates an elastic torque compensating layer to increase the output force as well as achieving bidirectional movement. Such a hybrid design also significantly reduces the degree of nonlinearity compared with a purely soft actuator. A model-based online learning and adaptive control algorithm is proposed for the wearable soft robotic glove, taking its interaction environment into account, namely, the human hand/finger. The designed hybrid controller enables the soft robotic glove to adapt to different hand conditions for reference tracking. Experimental results show that satisfactory tracking performance can be achieved on both healthy subjects and stroke subjects (with the tracking root mean square error (RMSE) < 0.05 rad). Meanwhile, the controller can output an actuator–finger model for each individual subject (with the learning error RMSE < 0.06 rad), which provides information on the condition of the finger and, thus, has further potential clinical application.

A high-level perceptual model found in the human brain is essential to guide robotic control when facing perception-intensive interactive tasks. Soft robots with inherent softness may benefit from such mechanisms when interacting with their surroundings. Here, we propose an expected-actual perception-action loop and demonstrate the model on a sensorized soft continuum robot. By sensing and matching expected and actual shape (1.4% estimation error on average), at each perception loop, our robot system rapidly (detection within 0.4 s) and robustly detects contact and distinguishes deformation sources, whether external and internal actions are applied separately or simultaneously. We also show that our soft arm can accurately perceive contact direction in both static and dynamic configurations (error below 10°), even in interactive environments without vision. The potential of our method are demonstrated in two experimental scenarios: learning to autonomously navigate by touching the walls, and teaching and repeating desired configurations of position and force through interaction with human operators.

Objective: Stroke survivors are usually unable to perform activities of daily living (ADL) independently due to loss of hand functions. Soft pneumatic gloves provide a promising assistance approach for stroke survivors to conduct ADL tasks. However, few studies have explored effective control strategies for the ‘human-soft robot’ integrated system due to challenges in the nonlinearities of soft robots and uncertainties of human intentions. Therefore, this work aims to develop control approaches for the system to improve stroke survivors’ hand functions. Methods: Firstly, a soft pneumatic glove was utilized to aid with stroke-impaired hands. Secondly, a probabilistic model-based learning control approach was proposed to overcome the challenges. Then a task-oriented intention-driven training modality was designed. Finally, the control performance was evaluated on three able-bodied subjects and three stroke survivors who attended 20-session rehabilitation training. Results: The proposed approach could enable the soft pneumatic glove to provide adaptive assistance for all participants to accomplish different tasks. The tracking error and muscle co-contraction index showed decreasing trends while the hand gesture index showed an increasing tendency over training sessions. All stroke survivors showed improved hand functions and better muscle coordinations after training. Conclusion: This work developed a learning-based soft robotic glove training system and demonstrated its potential in post-stroke hand rehabilitation. Significance: This work promotes the application of soft robotic training systems in stroke rehabilitation.

Soft robots with a well-balanced performance in terms of dexterity, accuracy, and payload have a great potential for application. Balancing safe human-robot interaction with operation performance enables the use of soft robot in biomedical fields, among others, such as surgery, rehabilitation and elder care. In this letter, we present a rod-driven soft robot (RDSR) where the flexible rods embedded in the silicone-based body provide a double push-pull actuation, to satisfy the needs of balanced performance. The RDSR can realize multi DOFs movement with elongation, contraction and bending. For trajectory tracking, Gaussian process model enables RDSR to achieve more accurate motion control with an RMSE within 2.8 mm compared to the constant curvature model with that within 10.8 mm. Comparative experiments demonstrate that the workspace, vertical and lateral stiffness of the RDSR are up to 5, 2.6 and 5.2 times that of an analogous tendon-driven soft robot (TDSR), respectively. Additionally, the RDSR possesses the ability, that the TDSR does not have, to actively apply pushing perpendicular to an inclined plane. Based on numerical Jacobian matrix, the maximum force along target direction is greater than 4 N for a 30 $^\circ$ plane. Furthermore, pick-and-place tests validate that our soft robot, with large workspace, precise and steady motion, is capable of conducting object manipulation tasks.

Soft robots are increasingly being employed in different fields and various designs are created to satisfy relevant requirements. The wide ranges of design bring challenges to soft robotic control in that a unified control framework is difficult to derive. Traditional model-driven approaches for soft robots are usually design-specific which highly depend on specific design structures. Our approach to such challenges involves a probabilistic model that learns a mapping between soft actuator states and control actions. Then an optimal control policy is derived by minimizing a cost function based on the probabilistic model. We demonstrate the efficiency of our approach through simulations with parameter analysis and real-robot experiments involving three different designs of soft pneumatic actuators. Comparisons with previous model-based controllers are also provided to show advantages of the proposed method. Overall, this work provides a promising design-independent control approach for the soft robotics community.