Modelling speed–density relationship in traffic streams

This study set out to estimate the mean speed of passenger cars based on traffic density in two traffic conditions in Tehran, Iran: a homogeneous flow regime in urban highways and a heterogeneous flow regime in suburban highways. Three established models – Greenshields, Greenberg and Underwood – were initially used. However, for higher model compatibility and better mean speed prediction, two new models were proposed. Step-wise regression was used as the statistical analysis for all models and model evaluation was made through multi-collinearity checks, Durbin–Watson tests and coefficients of determination. Results showed poor compatibility of the three established models in a heterogeneous stream and partial compatibility in a homogeneous stream. Using developments of the established models in heterogeneous flow and a new dual-regime model in homogeneous flow, the mean speed could be predicted with higher accuracy and better compatibility.