Research On Cooperative Task Assignment Of Swarm Of Drones

With the development of science and technology,the cost of high-precision weapons is getting higher and higher,and the growth of military spending in various countries can not keep up with the cost growth brought about by the progress of weapon technology,which also makes the UAV bee colony,a low-cost saturated strike method,attracted the attention of various countries,and the UAV bee colony collaborative task allocation is an indispensable part.This paper investigates the relevant UAV task assignment models and the solving algorithms of many task assignment models currently used.Based on the characteristics of UAV swarms performing one-time strike tasks,a long-range task assignment model and a short-range task assignment model are created for UAV swarms.The improved multi-population evolutionary algorithm is used to solve the remote task assignment model.In the improved multi-population evolutionary algorithm,an improved encoding method is used to encode the number of unmanned aerial vehicle,which can meet the subsequent algorithm calculation when the number of UAV swarms is larger,smaller or equal to the number needed to perform the task;and the crossover operator in the traditional evolutionary algorithm is removed and the local one is used.The method of gene location interchange supplements the ability to generate new solutions that are weakened by the removal of crossover operators;a simulated annealing algorithm is added to the genetic exchange among multiple populations to improve the convergence speed of the algorithm.Task assignment tests are carried out in the algorithm performance test to verify that the simulated annealing algorithm improves the stability and convergence speed of the algorithm.Repeated running tests were carried out in the algorithm performance test to test the effects of gene number,population number and evolutionary algebra on the benefit value and time cost,which provided the basis for the selection of algorithm parameters.The discrete particle swarm optimization algorithm is used to solve the near-range task assignment model.In the improved discrete particle swarm optimization,the genetic coding method is used to discrete coding UAV colony,so that particle swarm optimization can be applied to the discrete task assignment.The method of updating learning from individual extremum and group extremum is adopted,which makes the algorithm search for optimization quickly.In the performance test,the task assignment experiment is carried out.When the number of drone swarm increases from lack of Strike ability to saturation strike,the algebraic increase of convergence of improved discrete particle swarm optimization is not obvious.It is verified that the improved algorithm can converge rapidly in the case of insufficient Strike ability,matching ability and surplus Strike ability.The experiment of the algorithm performance is repeated.The control variable method is used to compare the influence of population number,population particle number and iteration times on the performance of the algorithm,which provides the basis for the selection of the parameters of the algorithm.Finally,the continuous scene simulation experiment is carried out by using c++multithreading programming.In the simulation experiment of the scene for improving many swarm evolution algorithms,the effectiveness of the task assignment of UAV colony is verified by using multiple swarm optimization algorithms.In the scene simulation experiment of the improved discrete particle swarm optimization algorithm,the effectiveness of the improved discrete particle swarm optimization algorithm for the task assignment of UAV swarms is verified.In the scene simulation experiment for the combination of the two algorithms,the task assignment in the long distance situation adopts improved many kinds of swarm optimization algorithms,while the task redistribution in the near distance case adopts the improved discrete particle swarm optimization algorithm.Compared with the single algorithm,the combined use improves the overall performance of the algorithm.