A new approach for drone tracking with drone using Proximal Policy Optimization based distributed deep reinforcement learning

Abstract

In this paper, a distributed deep reinforcement learning algorithm based on Proximal Policy Optimization (PPO) is proposed for an unmanned aerial vehicle (UAV) to autonomously track another UAV. Accordingly, this paper makes three important contributions to the literature. The first one is the development of an efficient UAV tracking algorithm, the second one is the presentation of a deep reinforcement learning approach that can be adapted to the problem, and the third one is the introduction of a generalized distributed deep reinforcement learning platform that can be used in various problems such as tracking, control and mission coordination of UAVs. In order to validate the developed approaches, the PPO algorithm is simulated with the deep reinforcement learning algorithm in a distributed and non-distributed manner, a follower UAV is trained in different scenarios and the distributed and non-distributed performances of the training using CPU are obtained, scenarios using general and adaptive learning algorithms are given, and finally, the performances of the algorithms developed in the paper are presented explicitly. (c) 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).