Empirical Analysis of Toll-Lane Processing Times Using Proportional Odds Augmented MARS

Abstract

A large deal of research has been conducted on toll roads to better comprehend the characteristics of the tolling operations. Several researchers indicated that toll-lane processing time (or service time) is the most significant key parameter in calibrating simulated toll plazas and one of the central variables affecting toll-lane capacity. This paper analyzes the effects of various factors on toll-lane processing times including traffic characteristics, vehicle characteristics, and toll plaza characteristics. The proportional odds augmented multivariate adaptive regression splines (MARS) model outperformed the proportional odds model and was used as the final model in interpreting the results. Results indicate that plazas charging higher tolls and plazas requiring drivers to pay with inexact bills have larger processing times. Increased demand and increased interarrival times at the toll plaza’s outermost right lane also have higher processing times. Finally, results indicate that processing times during the p.m. peak hour are higher than the a.m. peak hour.