Vision-based multivariate statistical modeling for powered two-wheelers maneuverability during overtaking in urban arterials

Abstract

The present paper employs advanced statistical modeling in order to assess the spatial factors that may influence the maneuverability of Powered Two-Wheelers (PTW) in urban corridors during overtaking. Using trajectory data from video recordings, two types of maneuvers are examined: those that occur with a lane change and those conducted without speed change and/or lane change (on the fly). The modeling approach has two parts: First, the probability to conduct an overtake is statistically modeled. Second, a Multiple Indicators-Multiple Causes (MIMIC) latent variable model is developed to assess the microscopic traffic characteristics and other critical factors that may affect the maneuverability of PTW. Results show that during overtaking, speed difference from the preceding vehicle is the most significant factor. Further results from the MIMIC model indicate that the PTW maneuverability is related to the type of overtake, the distance between the PTW and the vehicle being overtaken, as well as speeds.