Research On Multi-UAV Task Assignment And Path Planning Algorithm In 3D Environment

With the rapid development of UAV technology in the military and civilian fields,the application scenarios of UAVs are becoming more and more massive and the tasks performed more and more complex,and it has become a trend to improve the collaboration capabilities of multiple UAVs.How to reasonably allocate tasks and plan paths in 3D environment,take the advantages of each UAV,reduce the overall consumption of UAV and improve the mission reward is the key to multi-UAV collaboration.At present,the existing research on the multi-UAV task allocation problem in 3D environment mainly focuses on intelligent bionic algorithms,which is simple and intuitive but easy to fall into local optimum or fail to converge prematurely.Research on multi-UAV path planning problems in 3D environments has focused on intelligent bionic algorithms and reinforcement learning,but intelligent bionic algorithms are difficult to solve real-time path planning problems in which the multi-UAV systems that obtain information through sensing devices.In view of the above problems,this thesis researches and improves the multi-UAV task assignment and path planning algorithm in 3D environment.The specific work is as follows:(1)This thesis researches the task allocation problem of multi-UAV transport delivery in 3D environment.In this scenario,this thesis establishes an unsaturated task allocation model for multi-UAVs in a 3D environment with the goal of maximizing the integrated reward.Then,this thesis proposes a genetic algorithm based on beetle antennae search to solve this task allocation model and performs simulation.The results show that the genetic algorithm based on beetle antennae search can solve the task assignment problem of multi-UAV transport delivery in 3D environment with better performance compared with other algorithms.(2)This thesis also researches the path planning problem of multiUAV fire rescue in 3D environment.This thesis builds a high precision urban high-rise fire scene based on Unreal Engine 4 and Microsoft AirSim plugin.Then this thesis converts the path planning problem of multi-UAV fire rescue in 3D environment into a partially observable Markov decision process,and proposes a Multi-Agent Proximal Policy Optimization algorithm based on Variational Auto Encoder and performs simulation.The results show that the proposed MAPPO with VAE can accomplish the path planning problem of multi-UAV fire rescue in 3D environment,the VAE method can accelerate learning and convergence,and the MAPPO with VAE can ultimately obtain better results compared with the MADDPG with VAE.