基于高斯过程混合专家的非合作目标自主 规避方法

To address the challenge of autonomous threat avoidance for on-orbit spacecraft facing non-cooperative targets，an intelligent threat avoidance strategy under the“prediction-decision”framework is developed based on the pursuit-evasion game theory with incomplete information. To tackle the diversity of target intents and trajectory uncertainty，a Dirichlet-Gaussian process mixture model（DPGPMM）is constructed for multimodal dynamic modeling of typical intents and maneuvering strategies of non-cooperative targets， which enables simultaneous clustering and regression，thus realizing accurate recognition of target intents and precise prediction of their maneuvering strategies. Subsequently，the prediction results are incorporated into a stochastic model predictive control（MPC）framework，and the optimal avoidance strategy for on-orbit spacecraft is designed to effectively counter the potential threats from non-cooperative targets. Numerical simulations show that the strategy can quickly and accurately recognize and predict the intents of non-cooperative targets，maintain excellent avoidance performance in various complex scenarios，and its effectiveness and practical value are fully verified.