基于分区间强化学习的集群导弹快速任务分配EI北大核心CSCD

针对集群导弹在线任务分配面临的环境不确定、耗时过长等问题,本文研究了一种基于分区间强化学习的集群导弹快速任务分配算法.首先,建立集群导弹的综合攻防性能模型,并将存在环境不确定性的集群导弹任务分配问题表述为马尔可夫决策过程.其次,针对该过程采用分区间强化学习,通过将搜索空间划分成若干个子区间,降低搜索维度,加快算法的收敛过程,并通过理论证明给出了最优区间划分依据.最后,通过3组仿真实验,分别从收敛速度、不确定条件下的寻优能力以及导弹和目标数量可变情况下的决策能力3个方面,验证了所提算法的快速性和优化性能.

Fast task allocation for missile swarm based on sectioned reinforcement learning

Aiming at the problems of uncertain environment and long time-consuming in online task allocation of missile swarm, this paper studies a fast task allocation algorithm of missile swarm based on the sectioned reinforcement learning. Firstly, the comprehensive attack and defense performance model of missile swarm is established, and the task allocation problem of missile swarm in the presence of environment uncertainty is expressed as the Markov decision process. Secondly, the sectioned reinforcement learning is adopted for this process. By dividing the search space into several subintervals, the search dimension is reduced and the convergence process of the algorithm is accelerated. In addition, the basis for optimal interval division is given through theoretical proof. Finally, through three groups of simulation experiments, the rapidity and the optimization performance of the proposed algorithm are verified from three aspects: convergence speed, optimization ability under uncertain conditions, and decision-making ability under variable number of missiles and targets.