Reinforcement Learning based Approach for Multi-UAV Cooperative Searching in Unknown Environments

摘要

In this paper, an important topic of cooperative search for multi-dynamic targets in unknown sea area by unmanned aerial vehicles (UAVs) is studied based on reinforcement learning (RL) algorithm. First of all, considering the model of environmental, UAVs dynamics, target dynamics and sensor detection, the multi-UAVs sea area search map is established, then, the search map is updated by using the concept of “Territory awareness information map and the original search map is expanded. Finally, according to the search efficiency function, a reward and punishment function is designed, and RL method is used to generate a multi-UAVs cooperative search path on-line. The simulation results show that, according to the proposed algorithm, UAVs can effectively perform the search task in the sea area with no prior information.