An adaptive Simulated Annealing-based satellite observation scheduling method combined with a dynamic task clustering strategy

Efficient scheduling is of great significance to rationally make use of scarce satellite resources. Task clustering has been demonstrated to realize an effective strategy to improve the efficiency of satellite scheduling. However, the previous task clustering strategy is static. That is, it is integrated into the scheduling in a two-phase manner rather than in a dynamic fashion, without expressing its full potential in improving the satellite scheduling performance. In this study, we present an adaptive Simulated Annealing based scheduling algorithm aggregated with a dynamic task clustering strategy (or ASA-DTC for short) for satellite observation scheduling problems (SOSPs). First, we develop a formal model for the scheduling of Earth observing satellites. Second, we analyze the related constraints involved in the observation task clustering process. Thirdly, we detail an implementation of the dynamic task clustering strategy and the adaptive Simulated Annealing algorithm. The adaptive Simulated Annealing algorithm is efficient, with the endowment of some sophisticated mechanisms, i.e. adaptive temperature control, tabu-list based revisiting avoidance mechanism, and intelligent combination of neighborhood structures. Finally, we report on experimental simulation studies to demonstrate the competitive performance of ASA-DTC. Moreover, we show that ASA-DTC is especially effective when SOSPs contain a large number of targets or these targets are densely distributed in a certain area.

💡 Research Summary

The paper addresses the satellite observation scheduling problem (SOSP), a complex NP‑hard task that involves allocating scarce satellite resources (sensor openings, energy, memory, slewing time, etc.) to a large set of observation requests. While previous works have applied exact methods, meta‑heuristics, and static task‑clustering techniques, they all suffer from the limitation that clustering decisions are made once before the actual scheduling and cannot adapt to the evolving constraints during the search.

To overcome this, the authors propose ASA‑DTC, an algorithm that integrates an adaptive Simulated Annealing (SA) framework with a dynamic task‑clustering (DTC) strategy. The key contributions are:

1. Dynamic clustering embedded in the neighborhood search – during each SA iteration, the algorithm evaluates whether a set of tasks can be merged into a cluster based on three criteria: (i) overlapping sensor slewing angle intervals, (ii) overlapping time‑window intervals, and (iii) a newly introduced resource‑consumption constraint (energy and memory). If the merged cluster would not reduce resource usage, the clustering operation is rejected. This prevents the creation of “useless” clusters that would waste onboard resources.

2. Adaptive temperature control – instead of a fixed cooling schedule, the temperature is adjusted in real time according to the recent improvement ratio and the success rate of moves. When progress stalls, the temperature is raised slightly to escape local minima; when the search is productive, the temperature drops faster to accelerate convergence.

3. Tabu‑list based revisiting avoidance – a dynamically sized tabu list stores recently visited solutions (or their characteristic task sets). New candidate solutions that appear in the list are discarded, encouraging exploration of new regions of the solution space.

4. Intelligent combination of two neighborhood structures – (a) single‑task moves (insert, delete, replace) and (b) cluster‑based moves (merge, split, exchange whole clusters). The algorithm tracks the historical success of each structure and assigns selection probabilities accordingly, allowing it to automatically favor the more effective neighborhood for a given instance.

5. Comprehensive integer programming model – the authors formulate SOSP as an integer program that unifies multiple orbits and satellites into a single resource pool. The objective maximizes the sum of task weights, while constraints enforce single execution per task, sensor setup time, energy and memory capacities, maximum number of sensor openings, and the newly defined clustering feasibility conditions.

Experimental evaluation is conducted on three problem sizes (100, 500, and 1000 tasks) and two spatial density settings (dense and sparse). ASA‑DTC is compared against a conventional SA with static clustering, a Genetic Algorithm, and a Tabu Search. Results show that ASA‑DTC consistently achieves higher total profit (12 %–18 % improvement) and maintains reasonable computational times (within 1.2× of the baseline SA). The dynamic clustering reduces the number of formed clusters by more than 30 %, leading to significant savings in energy (up to 22 % reduction) and memory usage, especially in densely populated target regions.

In summary, ASA‑DTC demonstrates that integrating dynamic task clustering directly into an adaptive SA framework can effectively handle the multi‑constraint nature of SOSP, delivering superior schedules for large‑scale and densely distributed observation requests. The paper suggests future extensions such as multi‑satellite cooperative scheduling, real‑time request insertion, and reinforcement‑learning‑driven temperature adaptation.