Toward a Framework and SUMO-based Simulation for Smart Traffic Control Using Multiagent Learning

In this paper, we propose an approach and its frame-work of our adaptive traffic control system using reinforcement learning. The proposed approach attempts to determine the best traffic light strategy from reducing pollution to reducing waiting times, managing emergency ambulances, dynamic speed managing and optimizing traffic jams. In the proposed work, a reasonable process for managing agents is utilized. By utilizing SUMO features we prepare a real-world scenario with different types of vehicles, bus stops, traffic lights, by utilizing map data from OpenStreetMap. Simulations using SUMO are implemented for higher traffic efficiency and fairness compared with the manual phase-fixed and adaptive traffic light. In this way an emergency vehicle (Ambulance, Fire service car and Police) can be gracefully handled by the traffic light system without human interaction.