Control-Oriented Learning on the Fly

TL;DR

This paper introduces a control strategy called myopic control for systems with unknown dynamics, enabling real-time decision-making to achieve control objectives during critical failures.

Contribution

It develops a theory and algorithm for myopic control that learns local dynamics through small perturbations while optimizing system trajectory direction in real-time.

Findings

Algorithm effectively learns local dynamics during operation.

Ensures near-optimal control trajectory in unknown systems.

Validated on aircraft crash avoidance and Van der Pol oscillator simulations.

Abstract

This paper focuses on developing a strategy for control of systems whose dynamics are almost entirely unknown. This situation arises naturally in a scenario where a system undergoes a critical failure. In that case, it is imperative to retain the ability to satisfy basic control objectives in order to avert an imminent catastrophe. A prime example of such an objective is the reach-avoid problem, where a system needs to move to a certain state in a constrained state space. To deal with limitations on our knowledge of system dynamics, we develop a theory of myopic control. The primary goal of myopic control is to, at any given time, optimize the current direction of the system trajectory, given solely the information obtained about the system until that time. We propose an algorithm that uses small perturbations in the control effort to learn local dynamics while simultaneously ensuring that the system moves in a direction that appears to be nearly optimal, and provide hard bounds for its suboptimality. We additionally verify the usefulness of the algorithm on a simulation of a damaged aircraft seeking to avoid a crash, as well as on an example of a Van der Pol oscillator.