Deep Reinforcement Learning for Urban Air Quality Management: Multi-Objective Optimization of Pollution Mitigation Booth Placement in Metropolitan Environments

Urban air pollution remains a significant public health concern in densely populated cities such as Delhi, where high pollutant concentrations stem from vehicular emissions, industrial activity, and construction dust. Conventional mitigation strategies often suffer from suboptimal placement and lack adaptability to complex urban dynamics. This study introduces a deep reinforcement learning (DRL) framework for optimizing the spatial deployment of air purification booths to improve air quality. The proposed method models a $50\times 50$ spatial grid using real-world data layers, including air quality indices (AQI), population density, traffic volume, green spaces, and industrial zones. A Proximal Policy Optimization (PPO) agent is trained to maximize pollution reduction while accounting for spatial fairness and deployment constraints. The framework is evaluated against random and greedy placement baselines using metrics such as population-weighted AQI improvement, spatial entropy, traffic impact, and coverage efficiency. Results indicate that the DRL-based approach achieves a balanced trade-off between pollution reduction and equitable distribution across high-impact zones. The findings demonstrate the potential of reinforcement learning in guiding data-driven, adaptive environmental interventions within smart city infrastructures.