Session: 02-01-01: Product and Process Design 1

Paper Number: 166517

Robustness of Priority Metric-Based Traffic Signal Control Algorithm 

Background

Traffic signal control algorithms are essential in urban traffic systems as they can reduce traffic congestion and vehicle emissions as well as enhance overall transportation efficiency. As urban population grows and traffic patterns become more complex, these control algorithms are required to effectively manage varying traffic conditions while maintaining consistent performance. This paper presents a robustness study of a recently reported traffic signal control algorithm and insensitivity of its control parameters to diverse traffic scenarios. By focusing on both system-wide efficiency and adaptability, the traffic signal control algorithm is expected to maintain urban traffic systems resilient as traffic demands continually change.

Motivation and Purpose
Previous generations of traffic signal control systems, such as MAXBAND and TRANSYT, primarily relied on fixed-time control strategies derived from historical data. While effective under undersaturated conditions, they lacked the flexibility to accommodate real-time variations in traffic flow. Subsequent adaptive methods, including SCOOT, OPAC, and PRODYN, offered improved responsiveness by using live traffic data to optimize signal timing. However, these methods still face limitations when confronted with unpredictable and complex traffic situations often encountered in modern urban networks. A recently proposed priority metric-based traffic signal control algorithm has shown promising results in minimizing network-wide fuel consumption, making it a valuable candidate for further investigation of its resilience and capacity under diverse conditions.

Contribution to Advancing Science and Engineering
Robustness testing of traffic signal control algorithms is a critical step in algorithm development and deployment. By conducting a thorough investigation of how optimized signal parameters handle variations in traffic density, composition, and flow patterns, this work contributes to the field of intelligent transportation systems. Specifically, it adds to the growing body of research that aims to design algorithms capable of maintaining consistent performance even in the presence of significant fluctuations or atypical traffic scenarios. In doing so, the study provides a framework for evaluating algorithm adaptability, thereby supporting future developments in traffic control technologies.

Methodology
In this research, we employ a simulation-based approach to evaluate the control algorithm’s robustness and the sensitivity of its control parameters. A genetic algorithm (GA) is utilized to determine the optimal parameter weights in a priority metric that considers total vehicle speed, the number of vehicles, total vehicle waiting time, and vehicle types. These weights are first calibrated under a baseline traffic condition using a microscopic traffic simulation model called SUMO. We then systematically test these parameters across multiple traffic scenarios - ranging from peak-hour congestion to off-peak low-flow conditions - to assess the algorithm’s performance stability. Recent studies showed the benefits of using high-fidelity traffic simulation with real-world data, enabling more accurate predictions of algorithmic behavior under realistic conditions.

Preliminary Results and Conclusions
Our simulation results reveal that the control algorithm, once calibrated under a nominal traffic scenario, retains robust performance when exposed to diverse traffic conditions. Notably, network-wide fuel consumption remains consistently low across peak and off-peak traffic scenarios, showing the insensitivity of the pre-optimized control parameters to moderate fluctuations in flow and vehicle composition. While further evaluation is necessary for extreme cases - such as sudden incidents or highly unbalanced traffic distributions - this initial robustness test strongly suggests that the algorithm is well-suited to complex urban environments. By demonstrating stable performance under various conditions, the study presents the benefits of integrating such robust algorithms into future intelligent traffic management systems.

Presenting Author: Minjung Kim The University of Alabama

Presenting Author Biography: Minjung Kim received her B.S. degree in Mechanical Engineering from Incheon National University, South Korea. She is currently a Ph.D. Candidate in the Department of Mechanical Engineering at The University of Alabama. Her research interests include machine learning, data science, control, and traffic optimization.

Authors:

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