Development of a model to evaluate capacity of urban multi-lane roads under heterogeneous traffic conditions

Road capacity is defined as the maximum sustainable hourly flow rate at which vehicles can reasonably be expected to traverse a point or uniform section of a lane during a given time period under prevailing roadway and traffic conditions in the US Highway Capacity Manual. The knowledge of capacity of a given section of a road is an important input parameter for transport planning and traffic management studies. Presently, there aren’t any up-to-date guidelines for road capacity estimation in Sri Lanka. The use of foreign guidelines is not recommended as each country has unique factors that influence capacity. Since urban multi-lane roads are typically the busiest roads, this research study focuses on developing a capacity estimation model for urban multi-lane roads in Sri Lanka. Flow and speed data were collected using manual counting methods and Google Distance Matrix API (Application Program Interface) method respectively. The heterogeneous traffic flows were converted to Passenger Car Units (PCUs) using Chandra’s method. Greenshields’ traffic flow model was used to calibrate the empirical data. Capacity values were established from the developed flow-speed model. Using this method, the capacity values of all study locations were established. The average observed lane capacity was 1829 pcu/h/l. Regression models were developed to estimate capacity of four-lane and six-lane roads. It was observed that the four-lane road capacity was influenced by the effective lane width, access point density, built environment and median type whereas the six-lane capacity was influenced by the effective lane width and access point density. The four-lane capacity model had an R-squared value of 0.81 and the six-lane capacity model had an R-squared value of 0.86. The two models were combined to create a single model that predicts both 4-lane and 6-lane roads. In addition to the capacity models, a regression model was developed to estimate the Free Flow Speed (FFS) of roads. The predictor variables of the FFS model are lateral clearance, built environment and median type. Verification of developed models were done by surveying 10 road sections. It was observed that all three models accurately predicted flow and speed from the statistical tests done (Mean Absolute Percentage Error