Optimizing Intelligent Transportation Systems with Multi-agent Reinforcement Learning: A Socio-economic Impact Assessment

Rapid urbanization has exacerbated traffic congestion, presenting significant socio-economic and environmental challenges globally. This paper evaluates the socio-economic impact of implementing Intelligent Transportation Systems (ITS) enhanced by a novel Socio-Economic Reinforcement Learning (SERL) framework. We aim to minimize congestion and enhance overall transportation efficiency. The proposed method employs a hierarchical reinforcement learning algorithm specifically designed for complex multi-intersection urban traffic networks, considering socio-economic and environmental factors. Extensive simulations utilizing real-world traffic data assess the impact on travel time, fuel consumption, and emission levels. Experimental results indicate that our approach reduces average travel time by up to 26.67% compared to fixed-time control methods, decreases fuel consumption by 13.99%, and lowers COx/NOx emissions by 20.82% in specific scenarios. These significant improvements over traditional and existing RL-based methods underscore the potential of SERL-powered ITS in promoting sustainable urban development and improving socio-economic outcomes.