The Autonomous Scheduling Problem in Satellite Constellations for EO Missions. A Robust Distributed Optimization Approach

Satellite operations, including tasking and commanding, are traditionally managed by ground-based mission control and planning centers. These centers leverage a comprehensive, system-wide perspective to determine operations and dispatch telecommands to orbiting satellites via TT&C ground stations. Accurate knowledge of the system’s status is crucial for scheduling and planning; however, as the scale of satellite constellations expands, maintaining this level of coordination becomes increasingly challenging, particularly in highly dynamic environments. Managing real-time status updates and future task scheduling for a large number of satellites becomes more demanding for a central coordinator. The rise of large-scale and mega-constellations in remote sensing and telecommunications calls for a significant shift in operational strategies. Advanced mission control and planning approaches must evolve beyond traditional centralized models. Critical enabling technologies, such as onboard satellite computing and inter-satellite communications, now make distributed and autonomous system operations feasible. By utilizing space-edge and cloud computing, intelligence can be decentralized across the constellation, reducing the dependence on a central control structure. This paper discusses the application of distributed optimization methods in multi-agent systems, where communication networks evolve dynamically based on orbital mechanics. Integer linear programming is employed to minimize system latency in Earth Observation missions by optimizing image acquisition and data downlink, with a robust version that accounts for uncertainties in local satellite constraints. We perform numerical simulations that model constellation dynamics, acquisition, downlink, and communications to provide a comprehensive assessment of system performance.