AbstractThe rapid expansion of satellite constellations and the continued accumulation of orbitaldebris have significantly increased the risk of satellite–satellite and satellite–debriscollisions. While modern Space Situational Awareness (SSA) systems provide increasinglyaccurate tracking and conjunction warnings, collision prevention remains largely advisoryand dependent on human-in-the-loop decision processes. This approach introduceslatency, inconsistency, and poor scalability in dense orbital environments.This paper proposes a drift-aware, autonomous collision avoidance architecture designedto operate directly onboard future satellite systems. The framework integrates dynamicproximity safety envelopes, layered detection and sensing, corrected drift modeling forimproved orbital prediction, autonomous maneuver decision authority, and cooperativebehavior between maneuverable satellites. By explicitly accounting for environmental driftand accumulated prediction error, the system enables earlier, lower-energy avoidancemaneuvers while reducing uncertainty-driven false alarms.The proposed architecture shifts collision prevention from passive warning to active,autonomous execution, providing a scalable foundation for safe orbital operations assatellite populations continue to grow. This approach offers a path toward sustainablespace traffic management before collision cascades impose irreversible consequences onthe near-Earth orbital