Bayesian Network Approach for Dynamic Bus Speed Guidance on Mixed Bus–Car Traveling Lanes

Minimizing bus waiting time and stops at intersections along arterials is essential for improving bus travel efficiency. Many methods have been proposed to achieve this goal by utilizing emerging connected technologies to provide bus speed guidance. However, most studies have only focused on traffic scenarios with exclusive bus lanes. In reality, especially in urban areas of China, exclusive bus lanes have not been designed for many arterials. The mixed bus–car traveling lanes could lead to traffic flow fluctuation and bus travel time uncertainty, but these challenges have not been sufficiently addressed in existing methods. This paper introduces a hybrid Bayesian network (HBN) method to consider the uncertainties and to determine dynamic bus guiding speeds for minimizing stops and waiting time at intersections. The approach consists of two components. The first is to predict real-time bus travel time and queuing states based on historical bus travel time distributions and real-time bus locations and signal states. The second is to generate bus guiding speeds considering speed limits and average speeds of passenger cars based on the predicted results. The proposed method was evaluated in a simulated environment of a real-world urban arterial in Kunshan, China, and four volume/capacity (V/C) ratios of total traffic flows were considered to simulate different traffic conditions. Evaluation results show a noticeable decrease in bus waiting time and stops at intersections compared with the original trajectories without guidance and a deterministic method. When V/C ratios are 0.5, 0.6, and 0.7, bus waiting times are reduced by 87.36%, 86.20%, and 84.81%, respectively, while the reduction decreases to 12.97% when V/C equals 0.8 with increased traffic demands.