Grammatical Evolution for automatic design of actuated traffic signal control plans
Mahmud Keblawi, Tomer Toledo · Applied Soft Computing · 2025
In traffic networks, proper signal control design is essential to ensure a reasonable level of service. Signal control designs are becoming increasingly complex, with numerous settings that must be calibrated and set. This paper introduces a novel approach for handling this complexity by automatically generating optimal actuated signal control plans using Grammatical Evolution (GE). GE has proven its effectiveness in automating the design of different complex systems, such as neural networks and analog electronic circuits. GE’s distinctive mapping and representation capabilities make it a powerful candidate for optimizing various systems. In contrast to traditional optimization methods for actuated signal plans, which focus on specific parameters, such as green times and cycle length, the GE-based approach evolves complete plans, including phases, detector placements, and transit priority strategies. As a result, it eliminates the need for human intervention in the design process, making it more efficient and less time-consuming. The proposed approach was tested with an application to an isolated intersection in Haifa, Israel. The results showed that the automatically generated signal plan outperformed the existing plan by reducing delay times and queue lengths. Moreover, this method demonstrated its efficiency in generating reliable traffic signal plans under challenging traffic conditions.