An unmanned aerial vehicle (UAV) cluster self-organizing flight method, system and UAV

Abstract

The application provides a UAV cluster self-organizing flight method, system and UAV, comprising: based on the proposed boundary constraint triangular dense packing method, using an independent reinforcement learning evaluation unit structure fitness, based on the fitness applying an evolution algorithm to obtain the optimal formation unit structure of the UAV cluster, and a UAV strategy model capable of flying to reach the unit structure; transforming the UAV parameters, pre-training to construct a UAV strategy model library, determining the strategy model matched with the performance parameters of the UAV cluster from the strategy model library; based on the task information, based on the observation of the flight information of the adjacent UAVs of each UAV and the selected strategy model, determining the strategy action of each UAV; the application can be used for the UAV cluster self-organizing flight under the communication denial and local observation restriction based on the structure evolution dense packing formation and the independent reinforcement learning based on the single-agent perspective; meanwhile, the technical scheme provided by the application is easy to understand, simple to implement and has wide applicability.

Images