Research On Path Planning For Multi-UAVs Collaborative Coverage Search Under Two Types Of Prior Information Conditions

The unmanned aerial vehicle(UAV)has advantages of small size,high flexibility,and low cost.Thus,it plays more and more important role in battlefield reconnaissance,radiation zone search and rescue,and forest fire control because it can avoid workers injured with lower cost.Therefore,many real-world scenarios have begun to use UAVs to assist in task execution.Currently,majority of successful studies focus on single UAV path planning and obstacle avoidance.On the contrary,there are still many shortcomings in the research of multi-UAV systems,mainly in the following three aspects.How to assign tasks and plan paths for complex and discrete terrain? How to effectively coordinate the collaborative search among multiple UAVs to achieve safety and low cost? How to develop search strategies and evaluation indicators based on different levels of information? Therefore,this thesis studies the multi-UAV cooperative coverage search problem under two different information conditions with the goal of finding a reasonable and efficient path planning solution while ensuring the safety of UAVs.The specific research contents are as follows:(1)A basic model for multi-UAV path planning is established based on the distributed multi-robot cooperative framework and model predictive control time-domain rolling optimization method.At the same time,the search area is modeled using a grid method,in which the search area is discretized as a two-dimensional matrix carrying some required information.On this basis,the differential evolution algorithm(DE)is used to solve different optimization objectives in coverage search problems.Meanwhile,the traditional DE is simplified based on measurements of UAV performance and population richness,which significantly reduces the complexity of the multi-UAV path planning problem.(2)For the case where only the terrain of the search area is known,an optimization model is established.In the model,the coverage rate and energy consumption are two optimization objectives,while turning angle and minimum inter-UAV distance are two constraints.Furthermore,a dynamic reward function is innovatively proposed.According to the function,weights of the energy optimization function and coverage optimization function can be adjusted based on the current location of the UAV.Consquently,the behavior tendency of the UAV in the search and non-search areas can be adapted.Simulation experiments show that using the dynamic reward function has a significant ability to avoid getting stuck in local optima in non-convex terrain.(3)For the case where there are dynamic targets and fuzzy information about the target area on a known map,a method using the coverage probability map update is proposed for map modeling.The map describes the state of the grid using coverage rate and grid target existence probability.Considering the significant impact of individual rules on group behavior,a repulsion and attraction model(R-A)is introduced to specify local rules for UAVs.The problem is optimized by maximizing the number of dynamic targets captured within a specified time.The reward function considers coverage rate,target existence probability,collision avoidance,and R-A model rules.Simulation experiments show that the coverage probability map method has a significant improvement in guiding the uniform search ability of the UAV swarm compared to traditional probability map methods.Moreover,the UAV swarm guided by the local rules based on the R-A model shows significant aggregating and dispersing search behavior,and more targets are found in the same time compared to traditional probability map methods.