EXPLORATION OF THE EFFECTS OF HETEROGENOUS VEHICLE COMPOSITION ON URBAN ARTERIAL TRAFFIC FLOW ATTRIBUTES

The streets of Dhaka experience significant congestion due to the substantial volume of vehicles traversing them daily. The diverse mix of vehicle types, or heterogeneity, increases the complexity of traffic flow and contributes to the broader issue of traffic congestion. Given the saturation of the surrounding land, there is limited potential for expanding road capacity through infrastructure development. Therefore, optimization techniques for the existing road network need to be implemented to enhance the overall traffic flow condition. In light of this, the principal aim of this paper is to identify the vehicle types that have the most significant influence in adversely impacting the traffic flow attributes including relative delay, density, and average speed of vehicles on the urban arterial roads of Dhaka. The Abdul Gani road beside the Bangladesh Secretariat was chosen as the study area, and one hour of traffic video data in the peak period was collected to develop a microsimulation model in PTV VISSIM. The model was calibrated by tuning the Weidemann 99 car-following model parameters along with lane-changing behaviour, lateral movement, and similar parameters. The model was validated by comparing the simulated and real-time traffic flow and generating GEH Statistic for each of the approaching links of the study network. The GEH values were determined to be under 5% which deemed the simulation model appropriate for further analysis. Hypothetical combinations of vehicle composition were developed by using the Latin Hypercube Sampling (LHS) method. The LHS Python code was tasked to create 500 combinations of relative traffic flow percentages of the vehicle types. Each of the combination was simulated in the base model to obtain the aforementioned traffic flow attributes. A Random Forest algorithm was applied to the output dataset acquired from VISSIM simulations to generate feature importance plots and to rank the vehicle types according to their contribution to the values of traffic flow attributes. The Random Forest Regressor technique was also utilized to create linear relationship equations between the relative percentage flow of vehicle types and the output variables. The results suggested that buses had the most significant impact on traffic density and relative delay of the road network. On the other hand, rickshaws presented the most threat to the average speed as they had the most impact in decreasing the overall speed of the simulated links. Besides these, correlation heatmap and partial dependence plots were also generated to better understand the traffic dynamics and the relationship between the types of vehicles. The findings of this study will help transportation planners and decision-makers understand the impact of different types of vehicles on traffic flow performance and in making crucial decisions regarding congestion management strategies.