Research On Multi-UVA Cooperative Task Allocation Based On Alliance Game

With the continuous advancement of technology,drones have become an important tool widely used in various scenarios,and their characteristics such as strong mobility,convenient deployment,and low cost have been widely recognized.However,with the increase in the number of tasks and constraints,a single UAV can no longer meet the task requirements.Therefore,multi-UAV cooperative execution has become a necessary mode.In the collaborative execution of multiple UAVs,task allocation is a crucial link,and various constraints and goals of UAVs and tasks need to be considered.However,the existing task assignment methods lack comprehensive research on the complete multi-UAV multi-task assignment system,and the assignment efficiency is low and it is difficult to handle large-scale tasks.In addition,these methods only consider the initial constraints of the task,and do not consider the impact of factors such as random task assignment and time constraints on the task completion,which may lead to the problem of low task completion.For this reason,aiming at the problems existing in the multi-UAV multi-task assignment research,this paper conducts in-depth research from two aspects:(1)Aiming at the UAV task allocation problem,a task allocation algorithm of hybrid differential evolution algorithm(Differential Evolution,DE)and particle swarm optimization algorithm(Particle Swarm Optimization,PSO)is proposed to solve the optimization problem.Both the DE algorithm and the PSO algorithm have certain limitations.The PSO algorithm converges prematurely when solving complex optimization problems,and it is easy to fall into a local optimum.The DE algorithm heavily depends on the test vector generation strategy and the parameter values used.In order to solve these problems,the proposed Game Based DE and PSO(GBDEPSO)algorithm uses the DE and PSO optimization algorithm to cooperate based on Nash bargaining theory.The Nash bargaining theorem tries to maximize the profits obtained by the two participants,avoid local optimum,and greatly improve the search ability of the algorithm.The experimental results show that although there is a slight disadvantage in terms of training time,the total revenue obtained by the algorithm is greatly improved compared with PSO and DE,and the distance cost generated during the task execution process is minimized.(2)Aiming at the problems of various emergencies in the actual process,the task assignment completed by the static alliance of UAVs cannot be well adapted.A real-time alliance reorganization algorithm based on reinforcement learning is proposed.First,overlapping alliances are used to form the game is used to assign the tasks of the heterogeneous UAV group.When various emergencies occur in the task assignment process,this paper introduces the DQN algorithm that belongs to reinforcement learning to dynamically adjust the alliance assignment of UAV migration,and finally forms A stable alliance has a higher average income and the fastest convergence speed and can find the most suitable alliance members for the alliance leader.The practicability and efficiency of the algorithm are proved by the simulation of the whole process of dynamic task assignment of UAV swarm.