Research On Simultaneous Coverage And Tracking With UAV Swarm

Inspired by the collective behaviors in biotic community,the individual in the swarm system with simple local interaction rules can emerge collective intelligence that is far beyond the individual ability.This decentralized,self-organized and autonomous swarm system is capable of strong stability,flexibility and self-healing.With the continuous progress of Unmanned Aerial Vehicle(UAV)technology,UAV is developing towards lightweight,intelligent and swarm,and the UAV swarm has become a research hotspot in recent years.Selecting UAV swarm as the research object,this paper focuses on the problem of simultaneous coverage and tracking(SCAT)with UAV swarm.The paper models for the SCAT problem and proposes the corresponding methods for it.What's more,the proposed methods in this paper are corroborated by multiple experiments.The main work and contribution of this paper are as follows:(1)A frame and model in velocity space for simultaneous coverage and trackingThe paper designs a frame for simultaneous coverage and tracking with UAV swarm,which couples three subtasks: area coverage,tracking targets and tasks assignment.Using this SCAT frame,the UAV swarm system can get a better performance than the traditional decoupled-planning frame,where SCAT frame can maximize its sensor ability while keeping targets in the view of UAV swarm.Then,an optimization model is designed for swarm SCAT,which directly modeling in velocity model space.Compared with traditional configuration space,it has advantages in computation and scalability.(2)A self-organized swarm coverage method with reciprocityThe paper proposes a reciprocal decision method for swarm coverage tasks,which is based on the designed optimization model and considers the reciprocity of neighbor UAVs.The proposed method determines an optimal reciprocal coverage velocity(ORCV)space,which is proved to be collision-free.What's more,according to the special situations in the construction of ORCV space,the paper designs a random probability approach based on the Monte Carlo method for searching the optimal velocity in ORCV space.Compared with Voronoi method and Virtual Force method,the approach has a better performance,such as higher coverage rate,less convergence and computation time.(3)A decentralized reciprocal SCAT approach for UAV swarmThe paper proposes a reciprocal control method for UAV swarm SCAT,which allows the UAV swarm to cover the selected area to search more potential targets and track multiple detected targets simultaneously.What's more,the paper presents some algorithms for two particular conditions in the process of decision and uses the Robot Operating System(ROS)and Gazebo to test the proposed method.(4)A swarm SCAT method based on the deep reinforcement learningThe paper presents a swarm SCAT method based on the deep reinforcement learning,which uses the proposed reciprocal decision method to generate mass data of two-UAV SCAT problem.Initialized by the generated data,the value network learns how to accomplish the SCAT task and estimate the finish time.After trained,the value network has a great performance in two-UAV SCAT problem and it can be extended to the UAV swarm scenarios.Experiments shows that the proposed method averagely takes 7.3ms to get an optimal action and it has improvement in the UAV swarm SCAT problem,where the generated trajectories are more smooth than the proposed reciprocal decision method and the finish time is cut by 31.51%.In addition,the proposed method is verified by the Hardware in loop simulation(HILS).