UAV Trajectory Optimization based on Predicted User Locations

Unmanned aerial vehicles (UAVs) can extend the coverage of wireless networks due to their high mobility and the favorable radio propagation characteristics. This paper studies the trajectory optimization of UAV that are acting as radio relays. The optimization is based on predicted user locations (UTO-PUL) to assist communication to the ground users who are unable to get coverage from the base station (BS). The existing work on trajectory design has considered several optimization approaches as well as reinforcement learning (RL) algorithms. All the algorithm takes into consideration the existing state of the network such as the channel conditions, initial positions and computes the destination of the UAV based on it. The proposed algorithm is designed to also consider the predicted user mobility of the future instances. The objective is to ensure the ground users are connected to the BS. The proposed UTO-PUL algorithm's performance is evaluated using simulations in a scenario with challenging terrain, where the proposed algorithm reduced the probability of users having no coverage by between 45% to 85% compared to non-predictive approaches, and achieved gains in median downlink signal power of 14 dB compared with a deep reinforcement learning (DRL) algorithm.