Research On A Real-time Algorithm For Hybrid Task Assignment In UAV Cluster

To the accompaniment of the development of military science and technology,the UAV technology in the aviation field develops very rapidly.In comparison with an expensive,multi-functional large-size UAV,the low-cost UAV swarm has potential advantages in operational reliability and complex environment adaptability.Therefore,how to use efficient control strategy to achieve multi-aircraft cooperation and maximize the overall performance of the tasks has become a focus in the field of modern UAV research.Due to the influence of task environment and its own resources,the problem of multi-aircraft cooperative task assignment is a multi-objective optimization.In recent years,researchers at home and abroad put forward hybrid control architecture by combining the advantages of centralized control and distributed control of UAV swarm.In the task allocation of hybrid control,the central control node allocates tasks according to the global information first,and each UAV executes tasks in accordance with the decision-making scheme of the central control node.In case of emergency,each UAV takes advantage of its own onboard processor to make decisions by itself,and the decision-making scheme completes multi-aircraft cooperation through the communication network between UAVs.However,there are still some problems in the hybrid control architecture.For instance,when the UAV encounters a pop-up mission or a UAV is no longer suitable to perform a task in the task queue due to its own reasons,there will exist the problems of a large communication overhead and the result’s tendency of falling into the local optimal solution when the task is transferred to other nodes for execution by distributed consensus,As a consequence,this paper,aiming at the issue,proposes to divide the task allocation process into two stages: : task pre-allocation and distributed consensus.Specific to the deficiency of distributed consensus’ s tendency of falling into the local optimal solution the task,in the pre-allocation stage,is firstly handed over to the central control node to use the global information for calculation.At this time,only the income and loss of each UAV executing the task independently need to be calculated,as well as the normalized income and loss value after the task is added to different UAV task sequences,a group of UAVs with the larger normalized income are selected to be added to the satisfaction set,and the subset suitable for performing tasks is calculated at a faster speed,without calculating the global optimal solution,which reduces the calculation time of the central control node.Secondly,in the distributed consensus stage,only the UAVs in the satisfaction set obtained from the pre-allocation stage participate in the distributed consensus stage,aiming at the disadvantage of large communication volume in the distributed consensus stage,which reduces the number of UAVs participating in the consensus as well as the communication overhead in the distributed consensus.Finally,tasks of different natures will be handled differently: when a task can be completed by one UAV,a UAV with the greatest benefit will be selected to perform the task through task pre-allocation and distributed consensus;when the task is complex and needs to be completed by multiple UAVs,the central control node will transform the problem of co-tasking execution into a "0-1" linear programming problem,and calculate a group of UAV collaborative tasks that meet the needs of task resources by the means of genetic algorithm,and guarantees the real-time completion of tasks by taking advantage of the idle time of task sequence at the same time.Through OPNET simulation software,different scenarios are simulated to verify the performance of the algorithm in the paper: as for a single pop-up mission,by comparing with the traditional distributed consensus scheme,the improved contract network protocol reduces the calculation time of task allocation and improves the task completion rate;as for a complex pop-up mission,the improved genetic algorithm is verified to have better convergence and running speed firstly,and then through the comparison with the MSOCFA algorithm of collaborative set built with the traditional distributed consensus,the simulation result demonstrates that the selection of UAV collaboration has a shorter calculating time when the number of UAVs is large.