Evaluating capacity of transportation operations with highway travel time reliability

Abstract

Highway traffic congestion is often considered part of routine operations and is anticipated with transportation travel times; however, severe and unexpected delays cause disruption to deliveries, schedules, operations, and other enterprise logistics. This paper includes a description of methods for changepoint detection applied to disaggregate vehicle speed and volume data to establish a transportation travel time reliability threshold. In the context of enterprise logistics and surface transportation, reliability is measured by the frequency of deviations from anticipated vehicle speeds associated with a given operating condition. In this paper, anticipated speeds are calculated as the speed values with the highest kernel density estimate (KDE) in lieu of traditional statistics of median or mean speed conditions. Specific complexities of transportation networks, such as stochastic system capacity, necessitates the identification of operational thresholds associated with rapid performance degradation. Results from a demonstration dataset (comprised of millions of observations) indicates a rapid decline in reliability near half of the capacity defined by traditional traffic models. Variability of the travel time reliability threshold was observed across geographic locations and years of observation, which conforms with concepts of stochastic system capacity based on inherent randomness of exogenous system factors.