DiffLoop: Tuning PID controllers by differentiating through the feedback loop

Abstract

Since most industrial control applications use PID controllers, PID tuning and anti-windup measures are significant problems. This paper investigates tuning the feedback gains of a PID controller via back-calculation and automatic differentiation tools. In particular, we episodically use a cost function to generate gradients and perform gradient descent to improve controller performance. We provide a theoretical framework for analyzing this non-convex optimization and establish a relationship between back-calculation and disturbance feedback policies. We include numerical experiments on linear systems with actuator saturation to show the efficacy of this approach.

Extension of paper in 2021 55th Annual Conference on Information Sciences and Systems (CISS). IEEE, 2021

References in corpus (1)

• Improper Learning for Non-Stochastic Control

Cited by in corpus (3)

• Machine learning for industrial sensing and control: A survey and practical perspective
• Adaptive Force-Based Control of Dynamic Legged Locomotion over Uneven Terrain
• DiffTune-MPC: Closed-Loop Learning for Model Predictive Control