Shape Estimation and Control of a Soft Continuum Robot Under External Payloads

Finding the shape of a continuum robot is a challenging task, especially in a constrained environment where external sensors like cameras are unable to provide sufficient information. In particular, accurate robot control becomes more difficult when the robot needs to interact with uncertain external payloads. In this article, we present the development of a soft robotic system with embedded sensors to help reconstruct the shape of the robot under different external disturbances. First, strain gauges were employed to perceive the current deformation state of the robot due to actuator inputs and external payloads. Then, a shape reconstruction (SR) method based on a spatial curve fitting approach was proposed, where multiple control points along the robot were predicted using Neural Networks. With the estimated shape, a local inverse kinematics model was developed so that the robot end-effector can be precisely controlled in quasi-static state to reach various positions, and position of end effector calculated via SR module was employed to improve the path following performance. Simulation and experiments were conducted and the results confirm that the proposed control system is effective to predict the shape and properly control the robot with or without external payloads.