Control of Small Spacecraft by Optimal Output Regulation: A Reinforcement Learning Approach

TL;DR

This paper presents a reinforcement learning-based optimal output regulation method for controlling small satellite swarms in space debris removal, emphasizing low-cost, modular, and disturbance-rejecting control strategies validated through high-fidelity simulations.

Contribution

It introduces a novel control approach combining tracking, adaptive optimal control, and output regulation for satellite swarms, with a focus on practical implementation and disturbance rejection.

Findings

The proposed controller guarantees asymptotic tracking of reference inputs.

Simulation results show improved performance over classical LQR controllers.

The method is practical for high-fidelity space applications with accurate spacecraft dynamics.

Abstract

The growing number of noncooperative flying objects has prompted interest in sample-return and space debris removal missions. Current solutions are both costly and largely dependent on specific object identification and capture methods. In this paper, a low-cost modular approach for control of a swarm flight of small satellites in rendezvous and capture missions is proposed by solving the optimal output regulation problem. By integrating the theories of tracking control, adaptive optimal control, and output regulation, the optimal control policy is designed as a feedback-feedforward controller to guarantee the asymptotic tracking of a class of reference input generated by the leader. The estimated state vector of the space object of interest and communication within satellites is assumed to be available. The controller rejects the nonvanishing disturbances injected into the follower satellite while maintaining the closed-loop stability of the overall leader-follower system. The simulation results under the Basilisk-ROS2 framework environment for high-fidelity space applications with accurate spacecraft dynamics, are compared with those from a classical linear quadratic regulator controller, and the results reveal the efficiency and practicality of the proposed method.