Block-type reinforcement learning based on matrix-based particle swarm optimization for base station deployment

Block-type reinforcement learning based on matrix-based particle swarm optimization for base station deployment

연구 분야: Software Development

학회: Cluster Computing

The base station, as a core component of wireless communication systems, provides connectivity between mobile devices and communication networks. The quality of its deployment directly affects the stability of user communications in various environments. This paper establishes a three-dimensional base station communication model to address the overall signal coverage rate in special regions with specific coverage requirements, which involves the deployment of base stations. Consequently, the block-type reinforcement learning (RL) based on the matrix-based particle swarm optimization (BRL-MPSO) algorithm is proposed for this purpose. This method leverages the global search capabilities of the matrix-based particle swarm optimization (MPSO) to improve the Q-table processing in reinforcement learning (RL) through block-type handling, thereby enhancing the efficiency of the algorithm and reducing computational costs. To more accurately reflect the specific coverage requirements of base stations, simulation experiments are performed. After 500 iterations, the overall coverage stabilized at approximately 95.6403%, peaking at 96.6407%, while simultaneously satisfying the specific coverage requirements for special regions. The feasibility and effectiveness of this method under the constraints of special regions are validated through comparative analysis with other algorithms and ablation experiments.