[B-1A] Antennas and Propagation A (Electromagnetic Wave Propagation and Applications Except for Communications)

4:00 PM - 4:15 PM JST(7:00 AM - 7:15 AM UTC)Candidate for Academic Incentive AwardCareer Explorer

[B-1A-17] A Study on Optimal Drone Placement Using Deep Reinforcement Learning in Mobile Communication Drone Relay Systems

◎△Yuta Tamura¹, Tetsuro Imai¹ (1. Tokyo Denki University)

Keywords:

Drone,Mobile commuication system,Deep Reinforcement learning,Millimeter-wave,UAV

This paper proposes an optimal relay UAV placement method using Deep Q-Network (DQN) for mobile communication drone relay systems. In the proposed method, the relay UAV learns an optimal placement policy based on its position and channel capacity, while the reward function considers the distance to the mobile station (MS), the variation in channel capacity, and the ratio of the instantaneous channel capacity to the maximum capacity achieved so far. Performance is evaluated using 30 GHz propagation data obtained by ray-tracing simulations. The results show that the proposed method successfully explores areas with high channel capacity as the learning progresses. However, fluctuations caused by fading increase the variability of the learning performance.