Cooperative Task Planning Method Of Multi-UAV In Dynamic Environment

Unmanned Aerial Vehicle(UAV)plays a key role in battlefield intelligence gathering,reconnaissance,electronic interference and firepower strike.In modern warfare,UAV often perform tasks in a cooperation way.The tasks planning method of multi-UAV in dynamic environment can increase the efficiency of executing tasks and make the UAV fleet fulfil complex task at a small cost.The research on task planning of multi-UAV mainly includes the establishment of the task planning model and the application of task planning algorithm.The existing task planning models lack the consideration of battlefield constraints and fail to meet the actual requirements.Moreover,the calculation cost of task planning algorithm is too expensive and the response to dynamic situation is not timely.Aiming at the above problems,we first research the description and modeling of task planning.We analyze the constraints of cooperative task assignment of multi-UAV,expend the existing cooperative multiple task assignment problem(CMTAP)model to meet battlefield requirements.When UAV fleet encounter emergency during task execution,we establish a dynamic task reassignment model and make local adjustment to task assignment scheme.Based on the model of task assignment and reassignment,we then research the cooperative task assignment and reassignment method of multi-UAV.We propose a task assignment method based on Self-organizing Feature Map(SOM)neural network.This method initializes the SOM neural network based on the battlefield information and combines learning rules of SOM neural network to solve the task assignment problem.We also propose a task reassignment method based on grouping strategy,we use a cluster algorithm based on SOM neural network to cluster the tasks that need to be reassigned and adopt an ant colony algorithm to adjust the task sequence of UAV fleet.Finally,A UAV path planning method based on Q-learning in dynamic environment is proposed.According to the characteristics of path planning problem,we establish the Markov Decision Process(MDP)model.Based on the MDP model,the combination of Q-learning algorithm and eligibility traces technology is adopted to plan a flight path for UAV that could avoid threat area.The result of simulation experiment and the comparison with other algorithms show that the cooperative task assignment and reassignment method of multi-UAV proposed in this paper can obtain a reasonable and effective task assignment scheme,and can respond dynamically to various emergencies encountered by the UAV fleet.The path planning method of UAV proposed in this paper improves the adaptability of UAV to dynamic unknown or partially unknown environment.