Deep Reinforcement Learning Approach for Smart Traffic Signal Control System

Traffic congestion is becoming a critical issue with the increasing usage of vehicles. The rapid increase in vehicles has naturally led to traffic congestion all over the world. The traffic congestion forcing people to sit idle in their vehicles, consuming more fuel, and increase of air pollution. Traffic light control policies are not optimized, leading vehicles to wait pointlessly for non- existent traffic to pass on the crossing road. The current traffic signal systems follow fixed timing patterns that are not adjusted depending up on the volume of vehicles present on the traffic junction. Traffic light control policies can be greatly improved by implementing Deep learning (DL) and Reinforcement Learning (RL) concepts. Reinforcement Learning has become a popular method for controlling traffic signals in recent times as it can learn the best control policies through practical experience. By using RL, this research work aims to enhance traffic control efficiency and minimize issues such as delays, traffic congestion, fuel consumption, and pollution. The goal of this proposed work is to reduce the cumulative waiting time of vehicles at the junctions. Waiting time, Reward, Queue Length are considered as performance metrics. Moreover, the proposed approach is designed to be cost-effective and can be implemented in real-time systems.