Coverage Path Planning for Multiple UAVs Using Reinforcement Learning

초록

UAVs are often used to observe inaccessible places or to explore large areas. If fast searching for large areas is needed, it is recommended using more affordable UAVs than using one powerful UAV. To operate multiple UAV search systems effectively, an appropriate search algorithm is important. The easiest method for operating multiple UAVs in area search problem is to divide the search areas and assign them to each UAVs. In this case, it is additionally necessary to efficiently plan the route without overlapping within the allocated area. For preparing unexpected risks such as collision with other UAVs, it is necessary to be able to generate routes in real time. This paper aims to solve Coverage Path Planning (CPP) algorithm by using deep Q-learning in an unpredictable environment.