Assessing Autonomous Inspection Regimes: Active Versus Passive Satellite Inspection

TL;DR

The paper tackles autonomous satellite inspection of a rotating RSO under uncertainty by contrasting passive (NMC) and active (RL-based waypoint transfers) strategies. It combines Monte Carlo evaluation for passive methods with a reinforcement-learning framework to train active policies, examining fuel use and surface coverage across RSO dynamic modes and illumination scenarios. The results show significant interaction effects, with active RL generally delivering lower fuel consumption and higher inspection progress, though passive strategies remain safer and more predictable in certain settings. These findings inform strategy selection for on-orbit inspection missions by balancing robustness, efficiency, and operational safety in the presence of environmental uncertainty.

Abstract

This paper addresses the problem of satellite inspection, where one or more satellites (inspectors) are tasked with imaging or inspecting a resident space object (RSO) due to potential malfunctions or anomalies. Inspection strategies are often reduced to a discretized action space with predefined waypoints, facilitating tractability in both classical optimization and machine learning based approaches. However, this discretization can lead to suboptimal guidance in certain scenarios. This study presents a comparative simulation to explore the tradeoffs of passive versus active strategies in multi-agent missions. Key factors considered include RSO dynamic mode, state uncertainty, unmodeled entrance criteria, and inspector motion types. The evaluation is conducted with a focus on fuel utilization and surface coverage. Building on a Monte-Carlo based evaluator of passive strategies and a reinforcement learning framework for training active inspection policies, this study investigates conditions under which passive strategies, such as Natural Motion Circumnavigation (NMC), may perform comparably to active strategies like Reinforcement Learning based waypoint transfers.

Figures (12)

• Figure 1: The top-left figure demonstrates an action space specified for waypoint sequence and waypoint hold inspection strategy. The top-middle figure demonstrates an example trajectory corresponding to a waypoint sequence strategy; the top-right figure shows an example trajectory corresponding to a point-hold. The bottom panel demonstrates an example action specification for an NMC hold.
• Figure 2: The figure in the upper panel contains the MonteCarlo results used for policy determination of PH and NMC strategies. The Pareto front for PH is highlighted in red, whereas results for NMC are highlighted black. The bottom left figure contains the strategy selected for PH whereas the bottom right figure contains the strategy selected for NMC.
• Figure 3: Sample policy rollout with overlayed HL decisions generated by the trained RL policy used for comparisons.
• Figure 4: The top panel shows sampling of RSO attitude over the 30 trials conducted for each RSO dynamic mode. The bottom panel demonstrates initial angular velocity sampled for the same trials.
• Figure 5: Cumulative fuel usage over time for the three agent inspection mission across 30 trials. Specified by RSO dynamic mode and policy combination. Green traces indicate early mission success. The top panel is for the Stable Tumble mode while the bottom is for Static CWH.