Toward Devising A Multi-Objective Traffic Signal Scheduling Approach for Non-Lane-Based Heterogeneous Traffic

Abstract

Even though traffic signal scheduling is one of the fundamental problems addressed in Intelligent Transportation Systems, most of the related existing studies solved this problem in the contexts of lane-based and homogeneous traffic. These studies ignored various distinctive contexts prevalent in developing countries, such as non-lane-based (i.e., designated lanes are not followed) and heterogeneous (i.e., motorized and non-motorized) traffic. Therefore, in this study, we propose a new methodology incorporating chronological implementations of multi-objective optimization and appropriate simulations to develop an efficient traffic signal scheduling system appropriate for non-lane-based and heterogeneous traffic. First, we explore NSGA-II: Non-dominated Sorting Genetic Algorithm with binary and real encoding-based populations. We propose two novel objective functions considering the non-lane-based heterogeneous nature of traffic. Among the two encoding-based implementations, we find that real-encoded NSGA-II performs the best in providing an optimal traffic signal schedule on a numerical simulator. Further, we integrate the real-encoded NSGA-II in DhakaSim, which is a microscopic traffic simulator that realizes distinctive contexts of developing countries. Afterward, we conduct extensive simulations over three road networks from Dhaka, Miami, and Riyadh. Our simulation results show that the proposed method can lead to a 163.1% increase in average speed and a 56.1% reduction in average waiting time across Dhaka city’s road network during peak traffic hours, compared to traditional fixed-time signal scheduling currently used in some parts of the city. Besides, our proposed methodology confirms improved average speed and average waiting time over the road networks of Miami and Riyadh, underscoring the approach’s generalizability. These results highlight the effectiveness of the proposed multi-objective signal optimization framework for non-lane-based heterogeneous traffic and its potential to improve urban traffic flow in diverse, congested settings significantly.