Session: 02-05-01: Data Driven Design

Paper Number: 146048

146048 - Optimal Traffic Signal Control Based on Model Predictive Control Using Real-Time Traffic Data 

Optimal traffic signal control can improve network-wide fuel efficiency as well as alleviate traffic congestion. Traditionally, traffic signals have been controlled using limited vehicle detection sensors; however, the recent advent of advanced traffic detection technologies, such as the sensor fusion approach that combines multimodal data from cameras and radars, enables the real-time collection of rich traffic information. The real-time traffic data can be used to predict future traffic flow and control traffic signals for optimization targets such as traffic throughput, total travel time, and network-wide fuel consumption. This paper proposes a Model Predictive Control (MPC) algorithm for optimal traffic signal control at real-world traffic intersections. The proposed traffic signal MPC method employs a Neural Network-based future traffic volume prediction model to minimize network-wide fuel consumption and unnecessary vehicle waiting time. Genetic Algorithm (GA) is used as a global optimization method to determine optimal control parameters including the prediction time horizon, optimal weights for road preference and waiting time of vehicles. In order to evaluate the effectiveness of the proposed method, a traffic simulation model is developed in SUMO, a high-fidelity traffic simulation environment. The traffic simulation model is based on a real-world traffic network incorporating the plausible vehicle detection range of a state-of-the-art sensor technology within the network. The incoming traffic flow in the model is simulated using actually measured traffic data from the traffic network, enabling a comprehensive assessment of the novel optimal traffic signal control method in realistic conditions. The simulation results show that the proposed traffic signal MPC method can significantly reduce network-wide fuel consumption compared to the conventional fixed-time control and comparable to recently studied priority metric-based control method. Additionally, incorporating truck priority in the control algorithm leads to further improvements in the fuel consumption reduction.

This paper presents a novel traffic signal control method based on Model Predictive Control (MPC), designed to reduce network-wide fuel consumption. The proposed algorithm was evaluated using a traffic simulation environment called SUMO. Within the simulation, three signaled intersections were implemented based on a real road, and the traffic flow was simulated based on measured traffic data from that road. To estimate future traffic flow for optimal control, a linear regression model combined with a neural network was employed. The accumulated vehicle throughput for each phase combination was calculated by the prediction model and used as primary factor to determine the next green signal. Total vehicle waiting time was also considered as an additional factor due to the discrepancy in traffic volume between main roads and side roads. A Genetic Algorithm (GA) was used to find optimal weights for waiting time, main road priority, and prediction time horizon. The simulation included passenger vehicles and oversized trucks, with multiplication factors applied to the counts and waiting times of the oversized trucks count, to emphasize their importance in reducing fuel consumption. The simulation results showed that the proposed MPC-based traffic signal control algorithm significantly reduced network-wide fuel consumption compared to a conventional fixed-time four-phase control algorithm.

Presenting Author: Minjung Kim The University of Alabama

Presenting Author Biography: Minjung Kim received her B.S. degree from Incheon National University, Incheon, South Korea in 2017 and M.S. degree in Mechanical Engineering from The University of Alabama, Tuscaloosa, AL in 2023. She is currently pursuing her Ph.D. degree in Mechanical Engineering at The University of Alabama, Tuscaloosa, AL. Her current research interests include application of machine learning to controls of transportation systems through traffic simulation and optimization.

Authors:

Minjung Kim The University of Alabama
Hwan-Sik Yoon The University of Alabama