UAV Trajectory Planning for Data Collection in Internet of Vehicles: A Reinforcement Learning Approach

Unmanned aerial vehicle (UAV) assisted data transmission is widely adopted in various internet of vehicles (IoV) applications. This paper is dedicated to the study of UAV-assisted data uploading in IoV with trajectory planning by considering factors including the energy consumption of the UAV, probability of reaching the destination, and the amount of collected data. In particular, we first present a system architecture for UAV-assisted vehicular data uploading. Next, we formulate the UAV trajectory planning problem with the goal of maximizing the amount of collected data. On this basis, we propose a deep reinforcement learning (DRL)-based algorithm, named as MACD-TD (maximizing the amount of collected data through trajectory design). Further, we transform the problem into maximizing the accumulative reward. In order to verify the effectiveness of the proposed architecture and algorithm, we built a 3D simulation platform for conducting experimental simulations. Finally, we give an extensive performance assessment and confirmed the efficacy of the proposed algorithm.

I. Introduction

With the rapid development of Unmanned Aerial Vehicle (UAV) control technologies and prevalence of adopting UAV in various applications[1], [2], great efforts have been devoted to UAV-assited data collection in Internet of Vehicles (IoV) [3]. Specifically, with the assistance of UAV, various of information generated or sensed by vehicles could be spread out over a wider area with faster speed.