Dynamic Control of Soft Robotic Arm

TL;DR

This paper compares various control strategies for a pneumatically actuated soft robotic arm, highlighting adaptive passivity control with sigma modification as the most effective in simulation, and discusses challenges like sensor noise and hysteresis.

Contribution

It evaluates multiple control approaches for soft robotic arms, emphasizing adaptive passivity control with sigma modification for improved performance in realistic scenarios.

Findings

Adaptive passivity control outperforms other methods in simulations.

High-gain observer enhances control robustness under sensor noise.

Hysteresis effects significantly impact control performance.

Abstract

In this article, the control problem of one section pneumatically actuated soft robotic arm is investigated in detail. To date, extensive prior work has been done in soft robotics kinematics and dynamics modeling. Proper controller designs can complement the modeling part since they are able to compensate other effects that have not been considered in the modeling, such as the model uncertainties, system parameter identification error, hysteresis, etc. In this paper, we explored different control approaches (kinematic control, PD+feedback linearization, passivity control, adaptive passivity control) and summarized the advantages and disadvantages of each controller. We further investigated the robot control problem in the practical scenarios when the sensor noise exists, actuator velocity measurement is not available, and the hysteresis effect is non-neglectable. Our simulation results indicated that the adaptive passivity control with sigma modification terms, along with a high-gain observer presents a better performance in comparison with other approaches. Although this paper mainly presented the simulation results of various controllers, the work will pave the way for practical implementation of soft robot control.