Reinforcement learning-based decision-making for spacecraft pursuit-evasion game in elliptical orbits

The orbital game theory is a fundamental technology for the cleanup of space debris to improve the safety of useful spacecraft in future, thus, this work develops a decision-making method by reinforcement learning technology to implement the pursuit-evasion game in elliptical orbits. The linearized Tschauner-Hempel equation describes the spacecraft's motion and the problem is formulated by game theory. Subsequently, an impulsive maneuvering model in a complete three-dimensional elliptical orbit is established. Then an algorithm based on deep deterministic policy gradient is designed to solve the optimal strategy for the pursuit-evasion game. For the successful decision of the pursuer, an extensive reward function is designed and improved considering the shortest time, optimal fuel, and collision avoidance. Finally, numerical simulations of a pursuit-evasion mission are performed to demonstrate the effectiveness and superiority of the proposed decision-making algorithm. The game success rate of the algorithm against targets with different maneuvering abilities is verified, which implies that the algorithm can be applied in extended scenarios.