Bayesian Markov Chain Monte Carlo-Based Simultaneous Speed Prediction Model for Heterogeneous Traffic on Urban Primary Roads

Abstract

Objectives: The study designs a simultaneous speed prediction model based on Bayesian Markov Chain Monte Carlo (MCMC) methods for heterogeneous, non-lane-based traffic on urban primary roads, with the aim of determining the effects of vehicle-specific traffic densities on the operating speeds of the six vehicle types. Methods: The traffic data were obtained from videos recorded by unmanned aerial vehicles on five main urban roads in Dhaka, Bangladesh, and were then processed by a deep learning approach in order to derive information on vehicle trajectories, traffic volumes, densities, and space-mean speeds. Six simultaneous equations for speed prediction were developed using a Bayesian MCMC framework, covering standard cars, utility vehicles, heavy vehicles, three-wheelers, two-wheelers, and non-motorized vehicles. The performance of the models was assessed by means of goodness-of-fit measures. Findings: The coefficients of determination for the models varied from 0.922 to 0.956 and the root mean square errors were between 0.50 and 1.56 km/h for the six types of vehicle. Operating speeds decreased steadily as traffic volume increased, although traffic composition had a significant effect on speed when the demand was moderate or high. Operating speeds increased and congestion was delayed when non-motorized vehicles were excluded. The transferability study showed that the models had a reliable ability to make predictions on separate road sections.