Research On Algorithms For UAVs Pathplanning In Hazardous Environment

With the rapid development of unmanned aerial vehicle(UAV)technology and artificial intelligence,the demand for automatic control methods increases sharply.From the perspective of modern UAV property,UAV is relatively small and very limited computing units and communication units can be equipped on it.For this reason,various flight path planning algorithms are designed for this problem.Most current UAV track planning algorithms turn the searching problem to shortest path problem or shortest time problem,and the security needs of the UAV are ignored.Under some condition the UAV will be much more prone to crash due to environmental factors,so the target of the planning problem should not only concentrate on the length of the path and the time consuming of the plan,and such target can lead to infeasible solution.According to the above problem,this article focuses on the path planning of UAVs group collaborative search in the hazardous environment.To improve the routing security of the UAVs group,two new heuristic algorithms are proposed.The main work of this article is as follows:(1)To solve the UAV searching problem of in the hazardous environment,a new objective function is created,both UAV security,search time and duplicated searched paths are considered.An improved greedy algorithm is proposed to solve the problem.This algorithm can be regarded as an improved method of Breadth-First-Search,much of time and storage cost is saved in the complex network,by greedy strategy through decompose the complex network to relatively simple network and solve step by step.So,the time complexity can be reduced by the algorithm.(2)To eliminate the duplicated searching step under the same conditions,an algorithm of path planning based on reinforcement learning has been proposed,which focuses on the opinion of the path value.The state-action matrix of each raster pixel is trained under a certain environment.During the searching process,it can be directly calculated by the task related parameters and trained matrix without traversing again.This algorithm greatly reduces the cost of double counting in the same environment.(3)The proposed algorithms are extended to the problem of UAVs group collaborative search in last part.When it comes to search efficiency,UAVs group definitely outperform a single UAV.Meanwhile,the expectation and variance in the hazardous environment of UAVs crash probability is reduced through the corresponding evaluation index constraint algorithm,which further improves the search efficiency overall of the UAVs group.Simulation experiment is performed to contrast the two algorithms with traditional method,and the results shows that the proposed algorithms in our work outperform traditional method in several aspects,including the total search time,the path of hazardous expectation and its variance.So,we can safely trust that the two algorithms have obvious advantages compared to the traditional methods.