Research On Key Technologies For Deployment Optimization Of Air-ground Collaborative Unmanned System

Today,unmanned systems technology is evolving rapidly,the application areas are expanding and the tasks performed are becoming more complex and diverse.Unmanned systems are required to perform continuous tasks in complex environments,and traditional single aircraft or formations are often unable to respond effectively to such scenarios where information changes dynamically.By enabling individuals within a cluster to collaborate efficiently,intelligent unmanned clusters form a self-organizing,highly stable distributed system,and by sharing information about the environment within the cluster,they expand their awareness of the environment and effectively improve the group’s ability to perform complex tasks.How to enable intelligent unmanned clusters to maximize the benefits of the unmanned systems themselves,and improve overall information processing and instant decision capabilities,is currently a pressing issue.This thesis considers a small air-ground collaborative unmanned intelligent cluster system,focusing on the following elements.(1)To address the problem of optimizing the deployment and energy efficiency of UAV swarms performing reconnaissance missions,a multi-UAV reconnaissance scenario is constructed as a single coverage scenario,and the utility function and energy efficiency of UAV swarm reconnaissance coverage are given by considering various parameters of the UAVs and the environment,decomposing the optimized reconnaissance coverage into UAV swarm location deployment and energy efficiency maximization,and constructing these two processes as a potential game to prove the existence of a Nash equilibrium solution.An efficient distributed autonomous optimization algorithm is designed and the UAVs are able to optimize the current strategy until convergence to the Nash equilibrium point.Simulation results verify the convergence and stability of the algorithm.(2)A reinforcement learning based dynamic path planning algorithm for unmanned vehicles is proposed for the complex environment faced by unmanned vehicles in the process of providing communication services and power supply to a swarm of drones,and the continuous movement of drones needs to be considered.The communication service and power supply from the unmanned vehicle to the UAV swarm are modeled,the optimization objective is set,and an improved Q-value table with greater dynamic adaptability is proposed to improve the stability of the algorithm applied to the dynamic environment.Simulations show that the algorithm can effectively plan the path of the unmanned vehicle and provide services efficiently while avoiding obstacles.(3)To address the problems of more iterations required and longer convergence time of the algorithm proposed in(2),a fast path planning algorithm based on migration learning is proposed to better apply unmanned vehicle path planning to real-life application scenarios.A similarity estimation process is added to the migration,and a joint similarity estimation method is proposed to select the scene with the highest similarity as the source scene,which improves the reliability of the migration.Learning evaluation metrics are designed for the migration learning path planning application in practice.Simulations show that the proposed algorithm can effectively shorten the time of path planning and has good robustness in different scenarios.