Integrated multiple satellite scheduling framework for earth observation and data download planning

Abstract

Satellite constellations are useful for obtaining images of requested locations on the Earth. The operator of Earth observation satellites may handle hundreds or even thousands of imaging requests every day with various priority levels, which typically cannot be fully addressed. Instead, the operator can selectively conduct high-priority tasks to maximize the scheduling mission's value. This optimization process is referred to as the satellite scheduling problem (SSP), which is a highly challenging problem that encompasses the selection of imaging task points, determination of task sequence, and download of acquired data – under a set of operational constraints on satellite performance (e.g., capacity, uplink/downlink speed, and dynamic characteristics). This thesis proposes an Integrated Multiple Satellite Scheduling Framework (IMSSF) to handle the SSP considering various practical constraints effectively. IMSSF consists of three modules – the Time Window Generation Module (TWGM), the Attitude Control Module (ACM), and the Task Scheduling Module (TSM). Four task scheduling algorithms – Mixed Integer Linear Programming (MILP) model in continuous-time and discrete-time domains, a First-In-First-Out (FIFO) heuristic, and a Receding Horizon Greedy Heuristic (RHGH) – were formulated. The proposed IMSSF has three distinctive characteristics. First, control laws are utilized to generate a Guidance Profile Function (GPF), which models the maneuver (transition) time accurately between two successive imaging requests. The GPF is incorporated into the task scheduling algorithms to grant a feasible scheduling solution that the satellites can execute. Second, the imaging and downloading schedules are optimized simultaneously to ensure all acquired data is downloaded to accessible ground stations. Lastly, the IMSSF is modularized so that the control laws for generating GPFs or the task scheduling algorithms can be changed freely. An intensive case study for virtual scenarios involving up to a thousand task locations and six satellites were carried out for four different task geometric distributions. The proposed MILP algorithms were compared with FIFO and RHGH in terms of computation time and the solution's quality (i.e., the sum of priority values assigned to the selected tasks). The study results demonstrated the efficacy of the proposed framework.