MADRL-Based 3D Deployment and User Association of Cooperative mmWave Aerial Base Stations for Capacity Enhancement
Yikun Zhao, Fanqin Zhou, Lei Feng, Wenjing Li, Peng Yu
Abstract
Although millimeter-wave aerial base station (mAeBS) gains rich wireless capacity, it is technically difficult for deploying several mAeBSs to solve the surge of data traffic in hotspots when considering the amount of interference from neighboring mAeBS. This paper introduces coordinated multiple points transmission (CoMP) into the mAeBS-assisted network for capacity enhancement and designs a two-timescale approach for three-dimensional (3D) deployment and user association of cooperative mAeBSs. Specially, an affinity propagation clustering based mAeBS-user cooperative association scheme is conducted on a large timescale followed by modeling the capacity evaluation, and a deployment algorithm based on multi-agent (MA) deep deterministic policy gradient (MADDPG) is designed on the small timescale to obtain the 3D position of mAeBS in a distributed manner. Simulation results show that the proposed approach has significant throughput gains over conventional schemes without CoMP, and the MADDPG is more efficient than centralized deep reinforcement learning (DRL) algorithms in deriving the solution.