Real-Time Traffic Signal Prediction and Control using Deep Q-Network

As cities grow and populations rise, an increasing number of people are taking to the roads in their personal vehicles. However, traditional traffic systems are failing to cope with this change as the infrastructure has not kept pace with the growth. Therefore, Intelligent Traffic Signals(ITS) are becoming a crucial area of research in order to address this issue. This paper proposes an ITS that utilizes YOLO and DeepSORT to detect and track each vehicle, enabling the system to calculate the traffic density. This information is then fed into a Fuzzy Inference System that selects the appropriate mode of operation for the system. These modes include Normal mode (for typical traffic conditions), Priority mode (for high traffic conditions), and Emergency mode (for the presence of emergency vehicles). Normal and Priority modes can be further divided into two sub-modes to handle intermediate conditions. DQN is used to determine the appropriate phase of the signal based on the traffic density and the mode of operation selected by the Fuzzy Inference System.