Research On Dynamic Collision Avoidance Algorithms For UAV Cluster

With the development of new sensors,autonomous control systems and communications technologies,UAV cluster research has become a research focus.Compared with single UAV combat mode,UAV cluster combat mode has the characteristics of diversification,high efficiency,high reliability and high intelligence.This paper focus on the issue of dynamic collision avoidance technology of UAV cluster.For the diversity and unpredictability of battlefield environment,targeting at the known static terrain obstacles,real-time detection of terrain obstacles and collision avoidance within the cluster,a collision avoidance scheme is proposed,which effectively solves the contradiction between accuracy and real-time in the flight avoidance of UAV cluster,and improves the security and stability of UAV cluster.For the collision avoidance problem of UAV cluster in known terrain,a multi-objective optimal path assignment strategy is designed,aiming at improving the accuracy of the algorithm.A* algorithm is used to obtain the optimal path between a single UAV and a single target point.The advantages and disadvantages of different evaluation functions under complex terrain are analyzed.Finally,Manhattan distance function is selected as the optimal evaluation function.On this basis,KM algorithm is introduced to distribute multiple optimal paths within UAV cluster,so as to obtain the optimal path combination to the whole.For the real-time detection of terrain obstacles and collision avoidance,a real-time path fine-tuning algorithm based on virtual force principle is designed.Based on the airborne radar detection data,the terrain obstacle model is constructed,and the virtual repulsion force between the three-dimensional terrain obstacle and the UAV is generated on this basis,so as to quickly fine-tune the UAV route.This algorithm has the advantages of small computation and high reliability,and is suitable for real-time avoidance of unknown obstacles.For collision avoidance between UAVs in cluster,a fast nearest point matching(CPA)detection algorithm based on projection matrix is proposed to predict the collision probability between drones.After that,it can combine the virtual force collision avoidance idea to minimize the route change for collision avoidance UAVs in cluster.In order to reduce the impact of collision avoidance on the flight efficiency of the cluster,this paper proposes a heuristic RRT path fast planning algorithm from the perspective of completeness and rapidity,combined with the heuristic function of A* algorithm and the fast random tree generation of RRT algorithm.It can quickly get a better solution of the UAV and target point path under complex terrain.Combined with the KM algorithm,the optimal adjustment of the flight path of the UAV cluster during flight can be realized,and the task execution efficiency is improved.The dynamic collision avoidance simulation system of UAV cluster is established,and the dynamic collision avoidance algorithm of UAV cluster under mountain terrain is verified.The result shows that the scheme can realize the dynamic avoidance of UAV clusters with multi-target.