An empirical investigation of macroscopic lane-changing characteristics on uncongested multilane freeways

This article investigates the macroscopic lane-changing characteristics on uncongested multi-lane freeways, focusing on the interrelations between the traffic conditions and the frequency (and fraction) of vehicles changing lanes. The data available for analyses were collected from two test sites using a time-elapsed video recorder. It has been found that traffic flow variables, such as the headway mean and variance, speed, and density ratios between neighboring lanes, are highly correlated with the fraction and frequency of vehicles changing lanes. The results of exploratory analyses are further examined with two generalized linear models. The logistic and Poisson regression are used respectively for constructing the lane-changing fraction and frequency models. With maximum likelihood estimation both models provide consistent information and achieve a reasonable goodness of fit.