Research on Dynamic Base Station Deployment Optimization Method for Parking Lot Based on Deep Q-Network

Aiming at the uneven base station coverage problem caused by the complex propagation environment in parking lots, this study proposes a dynamic base station deployment method based on an improved Deep Q-network (DQN). The impact of Non-Line-of-Sight (NLoS) propagation is quantified by building a three-dimensional channel model, and an adaptive reward function for channel state awareness is designed to incorporate multi-dimensional optimization objectives such as line-of-sight probability, coverage gain and mobile energy consumption. The dual network delay update architecture is innovatively introduced to effectively suppress the Q-value overestimation phenomenon, and combined with the prioritized empirical playback mechanism to accelerate the convergence of the algorithm. The experimental results show that, compared with the traditional genetic algorithm, the method improves the coverage by 11% (from 82% to 93%), shortens the convergence time by 34.4% (from 320 to 210 seconds), and reduces the out-of-bounds occurrence rate by 86.7% (from 15% to 2%) in a typical parking lot scenario of 200×150m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, which significantly improves the algorithmic robustness and computational efficiency. In the future, this method can be extended to multi-base station collaboration and three-dimensional space deployment, providing new ideas for the deployment of communication base stations in complex scenarios.