Dynamic Modeling and Motion Control of a Soft Robotic Arm Segment

Soft robotics has shown great potential in manipulation and human-robot interaction due to its compliant nature. However, soft systems usually have a large degree of freedom and strong nonlinearities, which pose significant challenges for precise modeling and control. In this paper, a linear parameter-varying (LPV) model is developed to describe the dynamics of a soft robotic arm segment. Given the different actuation mechanisms, the LPV models for elongation and bending motions are identified through experimental data. A state-feedback H∞ controller is designed for the LPV model using a linear matrix inequality (LMI). Simulation of the state-feedback controller indicates that the closed-loop system is stable but with steady-state errors. As a result, an iterative learning control (ILC) with P-type learning function is implemented to improve the tracking performance. Simulation results of the ILC+state-feedback controller show steady-state errors are significantly reduced with iterations. The ILCs+state-feedback controller successfully moves the soft robotic arm segment to its desired position within several iterations in experiments.