巡飞弹集群协同路径规划方法研究

Research on Collaborative Path Planning Method for Cluster of Cruise Missiles

针对传统启发式算法难以从大样本地形数据中及时提取经验模型的问题,提出了一种基于注意力强化学习的多巡飞弹协同路径规划方法。该协同优化方法综合考虑了生存概率、路径长度、负载平衡和耐力约束等影响因素。注意力神经网络被用来生成巡飞弹的协同侦察策略,并对大量的模拟数据进行测试,利用REINFORCE算法对注意力网络进行优化。实验结果表明,所提出的方法能有效解决实时性要求高的多巡飞弹路径规划问题,且求解时间小于传统算法。

For the problem that traditional heuristic algorithms are difficult to extract empirical models from large sample terrain data in a timely manner,a cooperative path planning method based on attention reinforcement learning for multi-patrol missiles is proposed.This collaborative optimization method integrates the influencing factors such as survival probability,path length,load balance and endurance constraints.Attention neural networks are used to generate cooperative reconnaissance strategies for patrol missiles,and a large amount of simulated data is tested to optimize the attention network using the REINFORCE algorithm.The experimental results show that the proposed method can effectively solve the multi-patrol missile path planning problem with high real-time requirements,and the solution time is smaller than that of traditional algorithms.