Safety-Constrained Optimal Control for Unknown System Dynamics

TL;DR

This paper develops a control framework that accounts for model inaccuracies in unknown system dynamics, ensuring safety and optimality in real robotic applications.

Contribution

It introduces a Pontryagin's Minimum Principle-based control strategy with a deviation penalty, bridging model-based and true system optimal control.

Findings

Control strategy aligns with true system optimal control under mild convexity.

Framework successfully applied to robotic cruise control with safety constraints.

Demonstrates robustness of the approach in real-world experiments.

Abstract

In this paper, we present a framework for solving continuous optimal control problems when the true system dynamics are approximated through an imperfect model. We derive a control strategy by applying Pontryagin's Minimum Principle to the model-based Hamiltonian functional, which includes an additional penalty term that captures the deviation between the model and the true system. We then derive conditions under which this model-based strategy coincides with the optimal control strategy for the true system under mild convexity assumptions. We demonstrate the framework on a real robotic testbed for the cruise control application with safety distance constraints.