Heuristic tree search method augmented with motion characteristics for solving the impulsive orbital pursuit evasion game

Orbital pursuit-evasion games (OPEG) have attracted growing academic interest owing to their significance in the domain of on-orbit servicing. However, existing solutions for impulsive OPEG often suffer from limited interpretability and algorithm stability. To overcome these limitations, this paper proposes a heuristic tree search method augmented with motion characteristics. Different from the existing methods, this study focuses on the description of the orbital dynamics information in the OPEG process and how it can be introduced into the algorithmic mechanism of tree search for the first time. By constructing quantitative reward functions that represent orbital motion characteristics, key motion information is introduced as heuristic factors into the Monte Carlo tree search (MCTS) framework, aiming to guide the tree search and prioritize the exploration of more promising maneuver spaces. Compared with the random strategy adopted by the standard MCTS during node expansion, the proposed heuristic tree search framework can generate optimal maneuver decisions more stably within a limited search time. In this study, the performance of different algorithms is comprehensively evaluated in terms of capture success rate, number of game rounds, and terminal relative distance. Simulation results show that, compared to the standard MCTS, the proposed heuristic tree search method enables the pursuer to achieve fewer game rounds and shorter terminal relative distances at the end of the game while also exhibiting greater decision stability across multiple repeated experiments. In addition, extended experiments are conducted under a three-dimensional orbital dynamics model, further showing that the proposed method has the potential to be adapted to multi-degree-of-freedom space mission scenarios.