基于注意力机制的无人机集群协同分群控制算法

Cooperative Fission Control Algorithm of UAV Swarm Based on Attention Mechanism

针对基于避碰、组队和聚集规则的无人机集群无法响应部分感知引起的多重刺激问题,本文提出一种基于注意力机制的无人机集群协同分群控制算法.为确保无人机对邻居信息的高效选择,文中考虑视线遮挡因素设计感知规则,然后引入注意力机制计算交互邻居中无人机对局部群体有序性的贡献;为解决多重刺激下无人机决策冲突的问题,文中设计一种基于注意力机制的状态转换模型,特别地设计亚激活状态计算刺激源的动态权重,以提高集群响应多重刺激的灵敏度;为实现无人机对目标持续准确的跟踪,基于无人机的运动状态确定运动策略,然后调整运动模型;通过仿真分析集群响应多重刺激时运动轨迹和运动方向的变化过程,并利用应激精度、子群序参量和分群耗时对分群运动过程进行分析.仿真结果表明:所提算法使无人机集群在5 s左右即可完成分群运动,实现对刺激快速准确的跟踪.

Aiming at the problem that unmanned aerial vehicle(UAV)swarm based on separation,alignment and co⁃hesion rules cannot respond to multiple stimuli caused by partial perception,a cooperative fission control algorithm of UAV swarm based on attention mechanism is proposed.In order to ensure the efficient selection of neighbor information by UAV,the perception rule is designed considering the sight occlusion factor,and then the attention mechanism is introduced to calculate the contribution of UAV to the orderliness of local group in interactive neighbors.To solve the problem of UAV decision conflict under multiple stimuli,a state transition model based on attention mechanism is designed.In particu⁃lar,the sub-activation state is designed to calculate the dynamic weight of stimulus source,so as to improve the sensitivity of swarm response to multiple stimuli.To achieve continuous and accurate tracking of the target by UAV,the motion strate⁃gy is determined based on the motion state of UAV,and then the motion model is adjusted.The changes process of motion trajectory and movement direction of the swarm in response to multiple stimuli is analyzed by simulation,and the process of fission motion is analyzed by stress precision,subgroup order parameter and fission time.The simulation results show that the proposed algorithm enables the UAV swarm to complete the fission motion in about 5 s,and realize fast and accurate tracking of stimuli.