Earth observation satellite imaging task scheduling with metaheuristics: Multi-level clustering and priority-driven pre-scheduling

Abstract

Daily planning and scheduling of Agile Earth Observation Satellite (AEOS) observation tasks are enormously challenging due to their inherent complexity. This complexity stems from the vast number of observations that need to be coordinated and the multiple constraints that must be considered. In this study, we propose a comprehensive framework consisting of three steps to effectively tackle these challenges. To begin with, we employ a Multi-Level Clustering technique (MLC) to mitigate the problem complexity. Additionally, we integrate a Fixed High Priority First (FHPF) algorithm for tasks pre-scheduling. This algorithm effectively manages conflicts in time resource allocation for observation tasks at the lowest cluster level. Finally, we conduct a comparative analysis of four metaheuristic algorithms, integrating the pre-scheduled task clusters from the MLC-FHPF module as input for the main scheduling process. Our strategy aims to overcome the limitations of existing methods by incorporating continuous time modeling and accounting for “Fixed Time Maneuver” for time-dependent transition time constraints. Real-world evaluations highlight the resilience and effectiveness of our MLC-FHPF framework when integrated with various metaheuristic algorithms. Our approach significantly outperforms standard metaheuristics in terms of efficacy and efficiency. Notably, the GA MLC-FHPF algorithm consistently surpasses BDP-ILS and ALNS, especially in large-scale scenarios. It adapts effectively to different task densities and scales, maintaining optimized scheduling performance and efficient processing times.