[Paper Review] Iterative Learning Control for Fast and Accurate Position Tracking with a Soft Robotic Arm.

Matthias Hofer, Lukas Spannagl+1 more|arXiv (Cornell University)|Jan 29, 2019

Iterative Learning Control Systems14 references4 citations

TL;DR

This paper proposes a norm-optimal iterative learning control (ILC) scheme combined with feedback control to enhance position tracking accuracy in a soft robotic arm actuated by antagonistic inflatable bellows. By applying the ILC scheme over fewer than 30 iterations, the root-mean-square tracking error was reduced from 13° to under 2° during aggressive 60° set-point shifts in 0.2 seconds.

ABSTRACT

This paper presents the application of an iterative learning control scheme to improve the position tracking performance for a soft robotic arm during aggressive maneuvers. Two antagonistically arranged, inflatable bellows actuate the robotic arm and provide high compliance while enabling fast actuation. Low-cost switching valves are used for pressure control of the soft actuators. A norm-optimal iterative learning control scheme based on a linear model of the system is presented and applied in parallel with a feedback controller. The learning scheme is experimentally evaluated on an aggressive trajectory involving set point shifts of 60 degrees within 0.2 seconds. The effectiveness of the learning approach is demonstrated by a reduction of the root-mean-square tracking error from 13 degrees to less than 2 degrees after applying the learning scheme for less than 30 iterations.

Motivation & Objective

• To improve position tracking performance of a soft robotic arm during high-speed, aggressive maneuvers.
• To address the challenge of high tracking error in soft robotic systems with compliant, fast-acting pneumatic actuators.
• To develop and validate an iterative learning control scheme that reduces repetitive tracking errors without requiring high-precision system modeling.
• To integrate ILC with a feedback controller for robust performance using low-cost switching valves for pressure control.

Proposed method

• A linearized model of the soft robotic arm's dynamics is used to design a norm-optimal iterative learning control law.
• The ILC scheme is applied in parallel with a feedback controller to stabilize the system and improve transient response.
• Low-cost switching valves are employed to control pressure in antagonistic inflatable bellows, enabling fast actuation with high compliance.
• The learning law minimizes the tracking error over repeated trials by updating control inputs based on past error data.
• The system is experimentally tested on a trajectory involving 60° set-point shifts completed in 0.2 seconds.
• The ILC update rule is derived to minimize the error norm across iterations, ensuring convergence and robustness.

Experimental results

Research questions

• RQ1 Can iterative learning control effectively reduce tracking error in a soft robotic arm during fast, aggressive maneuvers?
• RQ2 How does the integration of ILC with a feedback controller improve performance compared to feedback control alone?
• RQ3 What level of tracking accuracy can be achieved using low-cost pneumatic valves and a linear model-based ILC approach?
• RQ4 How many iterations are required for the ILC scheme to achieve significant error reduction in a repetitive task?

Key findings

• The norm-optimal ILC scheme reduced the root-mean-square (RMS) tracking error from 13 degrees to less than 2 degrees after fewer than 30 iterations.
• The system achieved high-accuracy position tracking despite the use of low-cost switching valves for pressure control.
• The combination of ILC and feedback control enabled stable and precise tracking during rapid set-point changes of 60 degrees in 0.2 seconds.
• The learning scheme demonstrated rapid convergence, with significant error reduction observed within the first 10–15 iterations.
• The linear model-based ILC approach was effective even with the inherent nonlinearities of soft pneumatic actuators.
• The experimental results confirm that iterative learning control is a viable method for improving accuracy in soft robotic systems performing repetitive tasks.