基于双重预判机制的无人机集群蜂拥控制避障方法  

A Method for Obstacle Avoidance in Swarm Control of UAVs Based on Dual Prediction Mechanism

作　　者：杨振[1] 李昕聪 张婉滢 王华山 庞小龙 周德云[1] YANG Zhen;LI Xincong;ZHANG Wanying;WANG Huashan;PANG Xiaolong;ZHOU Deyun(College of Electronics and Information,Northwestern Polytechnical University,Xi’an 710129,China;College of Microelectronics,Northwestern Polytechnical University,Xi’an 710129,China)

机构地区：[1]西北工业大学电子信息学院,西安710129 [2]西北工业大学微电子学院,西安710129

出　　处：《无人系统技术》2025年第1期68-87,共20页Unmanned Systems Technology

基　　金：国家自然科学基金(62103338);陕西省重点研发计划(2024GX-YBXM-115);航空科学基金(2022Z023053001)。

摘　　要：针对传统无人机集群避障方法在应对动静态障碍物时存在避障效率低、系统资源浪费等问题,提出了一种基于双重预判机制的无人机集群避障方法。首先,基于无人机与障碍物间的相对位置和速度信息进行初次碰撞预判;其次,采用图时序长短时记忆网络模型预测动态障碍物的未来轨迹,并结合无人机当前状态进行二次碰撞预判,其中图时序长短时记忆网络模型训练损失值最终在0.01左右;最后依据双重预判结果做出避障决策,并引入动态分层牵制策略作为集群控制方法。仿真实验结果表明,与无预判机制的避障算法相比,所提方法在单动态障碍物作战场景下,由50架无人机组成的集群在避障运动过程中,系统资源消耗有了明显的降低,且能更快地实现集群速度匹配,同时避障成功率达到100%,验证了所提策略在避障任务中的高效性与可靠性,也为大规模无人机集群在复杂环境中的协同飞行提供了新的技术途径。In order to solve the problems of low obstacle avoidance efficiency and waste of system resources in the traditional Unmanned Aerial Vehicle(UAV)swarm obstacle avoidance method when dealing with dynamic and static obstacles,this paper proposes a UAV swarm obstacle avoidance method based on dual prediction mechanism.Firstly,the initial collision prediction is carried out based on the relative position and velocity information between the UAV and the obstacle.Secondly,the Graph Temporal Long Short-Term Memory(GTLSTM)model is used to predict the future trajectory of dynamic obstacles,and the secondary collision prediction is carried out in combination with the current state of the UAV,in which the training loss value of the GTLSTM model is finally about 0.01.Finally,obstacle avoidance decisions are made based on the dual prediction results,and a dynamic hierarchical pinning constraint strategy is introduced as the cluster control method.The simulation results show that compared with the obstacle avoidance algorithm without prediction mechanism,the proposed method in the single dynamic obstacle combat scenario,the system resource consumption of the swarm composed of 50 UAVs in the process of obstacle avoidance movement is significantly reduced,and the swarm speed matching can be achieved faster,and the obstacle avoidance success rate reaches 100%,which verifies the efficiency and reliability of the proposed strategy in the obstacle avoidance task,and also provides a new technical approach for the cooperative flight of large-scale UAV swarms in complex environments.

关 键 词：无人机集群 蜂拥控制 分层牵制策略 双重预判机制 轨迹预测 图时序长短时记忆网络模型 避障 

分 类 号：V249.12[航空宇航科学与技术—飞行器设计]