Research On UAV Swarm Task Allocation Technology For Heterogeneous Target

With the rapid development of artificial intelligence technology,the intelligence level of unmanned aerial vehicle(UAV)swarm is also getting higher and higher.It can selectively and targeted implement observation,communication support and jamming attack on ground targets in a short distance.In practical application scenarios,UAV swarm are faced with the situation of concurrent execution of multiple tasks and coexistence of multiple task modes,and the attributes,values and requirements of multiple tasks are often heterogeneous.At the same time,the task execution environment is often dynamic,and there is an urgent need for dynamic optimization of UAV swarm system.Based on the above problems,this paper comprehensively considers the three aspects of heterogeneous task target selection,heterogeneous task resource allocation and heterogeneous task decision optimization,and uses game theory and reinforcement learning theory to study the distributed dynamic task allocation method of UAV swarm with strong adaptability to heterogeneous target tasks,and then verifies it in different environments.The main work and research contents of this paper are as follows:(1)Aiming at the problem of heterogeneous task target selection for multi-point reconnaissance and communication services in UAV swarm,a hierarchical Stackelberg game model is constructed,and the utility function of game participants is designed according to the characteristics of heterogeneous tasks.A distributed strategy update iterative algorithm is proposed,which converges and obtains the game equilibrium solution,and realizes the stable convergence of the UAV swarm task allocation scheme and the optimization of the total income of the system task.(2)Aiming at the problem of heterogeneous task resource allocation facing multi-mode,multi-value and multi-demand in UAV swarm,an Overlapping coalition game model is constructed.Considering the task mode,value,demand and UAV resource comprehensively,a reasonable coalition formation criterion is designed,a multi-mode heterogeneous task overlapping coalition formation algorithm is proposed,and the stability of coalition formation is proved.It is verified that the proposed method can effectively improve the system utility and task success rate,and can achieve efficient collaboration for heterogeneous task objectives in different environments.(3)Aiming at the decision optimization problem of heterogeneous tasks in UAV swarm facing the dynamics of unknown behavior patterns,a Markov game model was constructed based on the decision optimization of cluster for single target tasks.A cluster collaborative intelligent decision-making algorithm based on multi-agent joint reinforcement learning is proposed,and its convergence and effectiveness are proved.Based on the above work,a cluster collaborative intelligent decision algorithm based on distributed reinforcement learning is proposed to optimize the cluster decision for multiple target tasks,which can achieve collaborative interference of UAV swarm.Aiming at the problem of unknown behavior patterns of target objects that make it difficult to optimize cluster collaborative resource allocation,a coalition formation scheme based on state feedback of learning processes is designed by incorporating a coalition game structure into the constructed Markov game model,and a distributed collaborative reinforcement learning algorithm based on joint interference state feedback is proposed.It is verified that the proposed algorithm can effectively enhance system utility and achieve rapid recognition and task allocation of heterogeneous multi-objective tasks with unknown behavior patterns.