关键词: 多无人机编队, 动态目标跟踪, 分布式粒子滤波, 轨迹优化, 目标失锁重捕

Abstract: For the persistent tracking task of dynamic targets by UAV swarms in unknown obstacle environments, existing methods are prone to target loss when confronted with occlusions from dense obstacles or active evasion maneuvers of the target. To address this issue, this paper proposes an integrated formation tracking framework featuring “Cooperative Estimation–Distributed Planning–Target Reacquisition”. First, at the state estimation layer, a distributed particle filtering framework based on Gaussian mixture model is applied, where each UAV achieves consistent estimation of the target state posterior distribution through exchanging parameter statistics and consensus iteration. Secondly, at the planning and control layer, a virtual rigid-body coordinate system is introduced to describe the formation configuration. Tracking and obstacle avoidance requirements are concerned in both front-end dynamic search and soft-constrained trajectory optimization based on MINCO curves, forming a 'search–optimize–execute' three-level safety mechanism combined with execution-side verification. Finally, for target loss scenarios, a frontier search and re-acquisition strategy is designed. When observation is lost, conventional filtering is stopped and historical particles are retained; local negative observations and obstacle information are fused; the weights of particles in invalid areas are globally weakened through consensus; frontier hotspots are then extracted and search direction commitments are assigned, driving the swarm to quickly and cooperatively reacquire the target. Multi-scenario simulation results demonstrate that the proposed method exhibits superior performance in persistent tracking, formation maintenance, and target reacquisition under lock-loss conditions when dealing with complex obstacle avoidance and target evasive maneuvers.

Key words: UAV Formation, Target tracking, Distributed particle filtering, Trajectory optimization, Target reacquisition