面向高超声速滑翔目标的多模型多意图融合轨迹预测

To address the problem of low trajectory prediction accuracy and limited prediction window caused by the high speed and strong maneuverability of hypersonic gliding vehicles，a trajectory prediction method based on multi-model and multi-intent fusion is proposed. By analyzing the control mechanism of targets，a time-varying prediction model set with longitudinal and lateral decoupling is constructed. Taking the lift variable，the drag variable and the amplitude of bank angle as the characteristic parameters，longitudinal maneuver models are designed through the least square method and Gaussian assumption. Taking the sign of bank angle as the characteristic parameter，lateral maneuver models are designed for typical actions such as attacking intents，avoiding no-fly zones，C-type and S-type maneuvers. Furthermore，the intent cost function is designed by combining the intent's distance，heading angles and avoiding no-fly zones. The posterior probability recursive formulas of maneuver patterns and intents are derived based on the Bayesian theory. Through double-level fusion of multiple maneuver patterns and multiple intents，the accuracy and robustness of trajectory prediction are improved. Simulation results show that the proposed method can cope with complex maneuvers of hypersonic gliding targets and effectively improve the performance of long-term trajectory prediction.