Hybrid Visual Servoing Control of a Soft Robot With Compliant Obstacle Avoidance

Soft robots demonstrate endowment in reducing unexpected impaction effects due to the compliant mechanism. This characteristic makes it possible to meet the safety demand even though contacts are generated between the robot and the external, making soft robots more competent with tasks in unstructured and interactive environments compared to their rigid counterparts. To drive a robot to execute in a constrained environment, conventional methods usually require preknowledge of the environment to plan the path that avoids obstacles, and deliberately control the motion of the robot. This article investigates the vision-based control problem of a soft robot in unknown constrained environments. Considering the unique characteristics of a soft robot, complete obstacle-avoiding motion is sometimes too conservative and may degrade the performance. Instead, this article proposes a compliant obstacle-avoiding (COA) algorithm, taking contact forces as a metric to evaluate whether the robot is under safe interaction. And if not, the optimization mechanism is designed based on the idea borrowed from the control barrier function to actively adjust the controls to ensure safety without impeding positioning performance. The proposed algorithm has experimentally validated the performance of visual servoing and COA in an eight-tendon driving soft robot platform. The results indicate that the controller can ensure good positioning accuracy, with the final image error converging to the subpixel scale. Meanwhile, the controller guarantees safety during the interaction as evidenced by the contact forces consistently remaining within the predefined allowable set.