Integrating High-Level Priority Decisions with Distributed Safety Filter for Multi-Satellite Collision Avoidance

Miniaturization and dense constellation deployments exacerbate collision risks of future satellites. While numerous collision avoidance strategies have been proposed, few reconcile agent-level safety with mission-level efficiency. In this paper, we propose a distributed inter-satellite collision avoidance framework where low-level safety control is guided by high-level priority decisions. First, we formulate “safe protocol” constraints among satellites and enforce these constraints on nominal controllers through distributed safety filters, establishing collision-free coordination of the swarm. By introducing tunable priority parameters within the safety filter, collision evasion responsibilities become dynamically adjustable, enabling swarm behavior adaptation. We further demonstrate two methods to integrate with high-level decisions: using optimization to approximate global reference behaviors and using Large Language Models to accommodate to tasks, respectively. Theoretical analysis proves the safety guarantees, while numerical simulations demonstrate the framework’s efficacy.