Motion Planning for Autonomous Spacecraft Inspection: A Review of Single and Multi-Agent Approaches

This paper reviews recent advances in motion planning algorithms for spacecraft inspection via free-flying vehicles, a critical capability for future space operations as agencies transition toward intermittently crewed and commercially operated platforms. We examine both single and multi-agent planning frameworks, analyzing their suitability for autonomous inspection missions in the challenging orbital environment. For single-agent systems, we evaluate optimization- based methods, sampling-based planners, and machine learning approaches, particularly recent deep reinforcement learning implementations. In multi-agent contexts, we contrast centralized and decentralized planning architectures, evaluate emerging hybrid methods combining Voronoi partitioning with learning-based navigation, and analyze task allocation and decomposition strategies for coordinating multi-robot teams. Our findings indicate that while sampling-based methods currently offer the most reliable performance for operational deployment, integration of deep reinforcement learning with physics-based constraints and run-time assurance frameworks represents a promising direction for next-generation inspection systems. We identify key challenges in simulation-to-reality transfer, communication-constrained coordination, computational efficiency, and verification methodologies that must be addressed to fully realize autonomous inspection capabilities for future spacecrafts, space stations in particular. These technologies will significantly enhance orbital operations across commercial, scientific, and defense domains by enabling comprehensive monitoring, rapid damage assessment, and improved space situational awareness with minimal human intervention.