Deep Reinforcement Learning Evolution Algorithm for Dynamic Antenna Control in Multi-Cell Configuration HAPS System

In this paper, we propose a novel Deep Reinforcement Learning Evolution Algorithm (DRLEA) method to control the antenna parameters of High-Altitude Platform Station (HAPS) mobile reduce number low-throughput users. Considering random movement HAPS caused by winds, throughput users might decrease. Therefore, that can dynamically adjust based on in coverage area improving users’ throughput. Different from other model-based reinforcement learning methods, such as Q Network (DQN), proposed combines (EA) with (RL) avoid sub-optimal solutions each state. Moreover, consider non-uniform user distribution scenarios, which are common real world, rather than ideal uniform scenarios. To evaluate method, do simulations under four different scenarios and compare conventional EA RL methods. The simulation results show effectively reduces low after moves.