Satellite Scheduling Algorithms for Earth Observation Systems

Satellite scheduling for Earth observation involves the allocation of limited orbital, energy and sensor resources to a set of imaging tasks subject to a multitude of constraints. These include visibility windows determined by orbital geometry, sensor pointing and agility limits, on‐board memory and power budgets, and ground‐station contact periods. The objective is often to maximise coverage, temporal revisit frequency, data quality or mission value, sometimes under multi‐objective criteria such as sensitivity to high‐priority targets and responsiveness to dynamic events. Classical approaches have employed integer programming and exact methods such as branch and bound, which ensure optimality but can struggle with the combinatorial explosion of modern multi‐satellite constellations. Heuristic and metaheuristic techniques—genetic algorithms, tabu search, simulated annealing and adaptive large neighbourhood search—have gained prominence by providing high‐quality solutions within operational time frames. More recently, machine‐learning paradigms such as deep reinforcement learning have been explored to achieve real‐time adaptability and improved scalability. Dynamic or rolling‐horizon frameworks address the arrival of emergent requests, allowing schedules to be revised on the fly. Across these methods, the trend is towards greater autonomy, integration of uncertainty modelling, multi‐satellite coordination and robust handling of emergency tasks. Advances in agility—satellite slewing and rapid retargeting—further complicate scheduling but also open up new mission profiles. This field is of global significance for climate monitoring, disaster response, agricultural management and security applications, and continues to evolve in response to the growth of small‐satellite constellations and demands for real‐time situational awareness.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent developments outside the Nature family include the application of deep reinforcement learning to continuous‐time satellite scheduling, wherein a deep deterministic policy gradient approach is combined with a graph‐partition preprocessing step to cluster tasks, yielding superior performance over traditional metaheuristics in large-scale simulations. Another study introduces a hybrid adaptive large neighbourhood search integrated with tabu search, demonstrating robust performance across diverse scheduling domains and outperforming general-purpose mixed integer programming and constraint-programming methods, including a dedicated test case on multi-orbit agile Earth observation scheduling. Foundational work in exact methods presents a specialised branch and bound algorithm for agile satellite tasking, leveraging look-ahead initial bounds and multiple pruning strategies to solve moderately sized instances to optimality in seconds, thereby providing benchmarks and insights into the limits of exact optimisation for modern constellations.

Satellite Scheduling Algorithms for Earth Observation Systems publication trend

The graph below shows the total number of articles in satellite scheduling algorithms for earth observation systems across all publications each year (not limited to Nature Index journals).

Technical terms

Branch and Bound Algorithm: An exact optimisation method that systematically explores and prunes a search tree to find optimal solutions to combinatorial problems.

Metaheuristic: A high-level solution framework (e.g. genetic algorithms, tabu search) designed to find near-optimal solutions for hard optimisation problems within practical time limits.

Deep Reinforcement Learning: A machine-learning technique where an agent learns a policy to make sequential decisions by maximising cumulative reward, often through neural-network approximations.

Adaptive Large Neighbourhood Search (ALNS): A metaheuristic that iteratively destroys and repairs parts of a candidate solution using a set of dynamically selected neighbourhood operators.

Tabu Search: A metaheuristic that guides local search by maintaining a short-term memory of recent moves (tabu list) to avoid cycling and promote exploration.

References

1. A Branch and Bound Algorithm for Agile Earth Observation Satellite Scheduling. Discrete Dynamics in Nature and Society (2017).
2. Time/sequence-dependent scheduling: the design and evaluation of a general purpose tabu-based adaptive large neighbourhood search algorithm. Journal of Intelligent Manufacturing (2019).
3. Revising the Observation Satellite Scheduling Problem Based on Deep Reinforcement Learning. Remote Sensing (2021).
4. A Dynamic Scheduling Method of Earth‐Observing Satellites by Employing Rolling Horizon Strategy. The Scientific World JOURNAL (2013).