Adaptive Traffic Signal Control Based on Maximum Weighted Traffic Flow

The paper proposes a method for adaptive traffic signal control. The proposed method includes two stages, each of which is responsible for a separate algorithm. At the first step, the predicted “flow” of vehicles through the intersection along the given lane is estimated at the allowed traffic light signal. At the second step, a “weighted” flow estimate is formed, which takes into account the waiting time for vehicles at the intersection. The next phase of the traffic signal is determined by the criterion of maximizing the weighted traffic flow through the intersection. An experimental study of the proposed algorithm was conducted both on synthetic and real-world traffic scenarios, including an isolated intersection, a highway, and a city area. Based on experimental results, we can conclude that the proposed algorithm outperforms baseline classical and reinforcement learning methods for traffic signal control in terms of average waiting time and average travel time.