Estimating Operating Speed on Highways using Multiple Linear Regression and Artificial Neural Network Technique

Abstract

The speed variation along the successive highway sections is one of the most important factors in assessing geometric design consistency. Therefore, it is necessary to predict operating speed on important highway geometric features involving major safety issues. The present study aims to develop operating speed on the curve and tangent section of four-lane divided highways. For this research, the data is collected on 44 highway sections in India, including 22 curved and 22 tangent sections. The geometric features and free-flow speeds of various vehicles were collected. This study analyzed the speed profiles of vehicles that follow different statistical distribution patterns other than normal distribution. Multiple linear regression (MLR) and Artificial Neural Network (ANN) techniques are adopted to develop operating speed models on curve and tangent sections. The variables like curve radius, curve length and deflection angle are identified as most significant for modelling operating speed on the curves. Similarly, the shoulder width, median width, and access density are found to influence the operating speed on tangent sections. The developed models are successfully validated with field data. The performance measures such as RMSE and MAPE are applied to check degree of accuracy of developed models. The results revealed that the ANN models perform better than MLR models in curved and tangent sections. The developed models are helpful for highway and traffic engineers in establishing posted speed limits on critical sections of highways.