The Intelligent Decision-Making and Planning of Multi-Satellite Game Under a Single Fault

Encirclement is a typical form of spacecraft cluster game. The spacecrafts face the complex and time-varying environment in the game. This poses the challenge to the decision-making and planning of spacecraft cluster. For better addressing this problem, especially in cases where a single spacecraft experiences malfunctions, an end-to-end cooperative hunting method combined with the fault detection and tolerance mechanism based on the distance deviation between the current and deductive positions is proposed. A CTDE architecture named MADDPG is adapted to interact with the environment to train the decision-making and planning model of each spacecraft. During the interaction with the environment in the execution stage, the fault is injected and detected based on the minimum deviation threshold. Then, the healthy spacecrafts deduce the states with unpowered relative orbital dynamic until the fault disappears and the mission can proceed normally. The purpose of the proposed algorithm in this way is to generate an algorithm that can collaborate to complete the encirclement task even a brief fault occurs. Finally, a 3vs1 hunting experiment of spacecraft cluster is conducted and the result indicates that the proposed hunting method can effectively handle spacecraft game decision-making and planning tasks under one fault condition.