Trajectory design of cellular-connected UAV patrol and mobile edge computing system: A deep reinforcement learning approach

This paper investigates an Unmanned Aerial Vehicle (UAV)-based detection system deployed in an urban environment. Cellular network-connected UAVs collect data from multiple inspection points scattered in urban buildings and upload the data to a ground base station (GBS). Our goal is to minimize the energy consumption of the UAVs while accomplishing the data uploading task by designing the UAV inspection sequence, UAV path planning, and UAV correlation rate. To solve this intractable non-convex problem, we propose the EEGA-TD3 algorithm. First, an adaptive genetic algorithm is proposed to obtain the optimal inspection sequence by considering the energy consumption and throughput of the UAV and the transmission task. Subsequently, we utilize the dual-delay deep deterministic policy gradient (TD3) algorithm to optimize the UAV flight trajectory. The scheme is able to realize continuous control and guide the UAV to dynamically adjust its flight strategy according to the amount of data. Simulation results show that the proposed algorithm is able to flexibly select the trajectory scheme that accomplishes the data transmission task and saves energy compared to the traditional algorithm.