Approximation and Short-Term Prediction of Bus Dwell Time using AVL Data

Abstract

A significant proportion of bus travel time is contributed by bus dwell time for passenger boarding and alighting. This paper reports an investigation conducted on approximation and short-term prediction of bus dwell time based on historical AVL data collected from selected bus routes in Auckland, New Zealand. Three distribution functions including normal, lognormal and Wakeby distribution functions were assessed to approximate the distribution of bus dwell time. Autoregressive Integrated Moving Average (ARIMA) is assessed for the first time to make a short-term prediction of bus dwell time. Wakeby distribution outperformed the most commonly used distribution function namely lognormal distribution to approximate the dwell time for both peak and off-peak periods while ARIMA performed reasonably well for a short-term prediction of the dwell time.