Robust Replanning for Multi-Agent SmallSat Inspection in Failure Scenarios

Abstract

Future space structures like the proposed Lunar Gateway may benefit from autonomous inspection and maintenance, especially during extended uncrewed periods. Small satellites (SmallSats) offer a low-cost alternative to human extravehicular activities (EVAs) for performing these tasks. However, SmallSats have a high failure rate, particularly with propulsion systems, often causing mission failures. This work is motivated by this failure rate and considers a multi-agent approach to SmallSat inspection as a way of increasing mission success through redundancy. The added risk of the multi-agent approach, namely collision risk during thruster or other system failures, is addressed. We propose a collision-aware path-planning strategy based on the Anytime Dynamic A* (AD*) algorithm, enabling efficient initial path generation and dynamic replanning in response to failures. Implementation considerations like collision-checking in a high-fidelity simulation environment are discussed. The proposed strategy is demonstrated in a case study multi-agent inspection scenario around simulated Lunar Gateway geometry, where one SmallSat experiences a thruster failure, prompting other agents to replan their paths.