Dynamically Signal Timing Optimization of Isolated Intersection Traffic Lights Based on a Dual-Layer Framework

Abstract

Intersections are vital components of urban road traffic management, frequently facing persistent congestion challenges. Existing studies rarely combine multiobjective optimization with dynamic adjustment methods. This study introduces an innovative dual-layer framework for traffic signal optimization. The first layer involves multiobjective optimization, addressing critical performance metrics such as delay, the number of stops, and fuel consumption. In the second layer, we propose a method that uses a fuzzy neural network to learn the correspondence between queue lengths and signal timings. This two-tiered approach enables real-time adjustments, achieving dynamic signal optimization. Applying this framework with real traffic flow data to a specific road intersection allows us to determine optimal signal timings dynamically. Extensive simulations using the SUMO software validate the efficacy of our approach in enhancing intersection performance. The timing strategy implemented within this framework leads to a substantial reduction in delay, ranging from 11.1% to 29.0%. The dual-layer framework presented in this study contributes valuable theoretical insights into future research initiatives in this domain.