基于近似动态规划与零和博弈的空战机动决策

Maneuver Decision of Air Combat Based on Approximate Dynamic Programming and Zero-sum Game

为应对机动策略未知的空中目标,针对以往机动决策建模通常存在"维数灾难"的不足,提出了一种结合近似动态规划与零和博弈的在线积分策略迭代算法。首先描述了近似动态规划存在问题及求解方法,提出了双人零和微分对策的近似动态规划算法,然后分析了博弈双方的均衡策略和在线策略迭代的求解方法,并结合空战应用场景设计了评价函数,最后进行了空战格斗的算例仿真。仿真结果表明,该算法通过运用函数逼近可以解决动态规划中难以处理的大规模模型"维数灾难"问题。

In order to against the aerial target with unknown maneuver strategy, an online integral policy iteration algorithm is proposed to consider the objections of the ancient air combat maneuver model with the ＂dimension curse＂ problem. First, the approximate dynamic programming （ADP） problems and its solving method are described, and the algorithm combining the ADP with the two player zero-sum differential game is presented. Then, the solution methods of the equilibrium strategy and on-line strategy iterative in both sides are analyzed. Design evaluation function based on air combat application scene, and carry out example simulation of air combat. The simulation results show that the algorithm uses the function approximation to deal with the large-scale ＂dimension curse＂ problem in dynamic programming.