Search

The development of large-scale cluster intelligence will inevitably lead to new problems of adversarial game control. Aiming at the problem of high dimension and high dynamics in the process of unmanned aerial vehicle (UAV) cluster confrontation game, and the traditional optimal control algorithm cannot meet the requirements of timeliness, the evolution strategies (ESs) optimization method is proposed and applied to large-scale UAV cluster. It effectively avoids the problem that it is difficult to obtain accurate gradients when using reinforcement learning to deal with high-dimensional models, and promotes autonomous UAVs to find strategies with higher performance. First, the confrontation game models including UAV motion, cluster behavior patterns and interaction are established. Second, two UAV cluster game algorithms using the OpenAI evolution strategy (OpenAI ES) and integrated evolution strategy (IES) are presented. Finally, the large-scale UAV attack and defense confrontation scenarios have been established, and different sampling proportions and different numbers of UAVs are fully simulated. The results show that the two proposed algorithms can effectively solve large-scale UAV cluster confrontation game problems, especially the adaptive IES algorithm, which has better performance and shows more strategic behavior for the UAVs, which improves the effectiveness and robustness of confrontation strategies. [ABSTRACT FROM AUTHOR]