Leader-follower identification methodology for non-lane disciplined heterogeneous traffic using steady state features

Abstract

Road traffic in developing countries is characterized by heterogeneous vehicle types and weak lane discipline, with motorized and non-motorized vehicles of varying sizes and maneuverability leading to diverse driver behaviors. Identifying leader-follower (LF) pairs and analyzing vehicle-following (VF) behavior under such conditions is challenging, as proximity alone may not capture a leader vehicle's (LV) influence on a subject vehicle’s (SV) behavior. Non-following episodes, even with similar gaps or time headways, highlight the limitations of fixed longitudinal clearance thresholds. This study addresses these challenges by combining the k-v fundamental diagram to estimate desirable longitudinal gaps and wavelet transforms (WT) to match LV and SV speed profiles. The proposed methodology improves accuracy over heuristic methods, increasing the R-squared value from 0.268 to 0.349 and reducing RMSE from 0.764 to 0.652, offering a robust framework for LF pair identification in heterogeneous traffic conditions.