Multi-stage optimization framework of satellite scheduling for large areas of interest

The satellites are assigned to perform scanning tasks by maneuvering sensors in roll and pitch directions within resource constraints. However, these task regions may be larger than the scanned region by the satellites in a single attempt, so multiple attempts are needed to complete the task. Previous literature has proposed strip-based segregation for larger task regions with parallel strips. However, satellite trajectories may not be parallel to each other, and multiple visits from multiple satellites may be required to complete a scanning task. This results in higher usage of resources due to overlapping strips in strip-based segregation. Therefore, a novel three-stage scheduling methodology is proposed for better performance in terms of the size of the scanned region and resource utilization. The initial stage assigns satellites to the tasks irrespective of task segmentation for each scan. The second stage optimizes the roll and pitch angles of assigned satellites to maximize the scanned region with parallel computation since the tasks are independent. The third stage handles uncertainty associated with task failures. An elitist mixed-coded evolutionary solution technique with a constrained non-linear optimization and Markov decision process is proposed. The efficiency of the proposed methodology in small and large-scale scenarios is illustrated with simulations that show up to 20% improvement in the reward collection and 16% improvement in resource utilization compared to the traditional strip-based method.