A Bayesian approach for estimating link navel time on urban arterial road network

Abstract

Recently, Korean Ministry of Construction and Transportation selects three cities as the Intelligent Transport Model City to build a test bed for research in Intelligent Transportation System (ITS). One of the most sought-after information in any ITS project is to provide arterial travel speed forecasts to travellers. To estimate the arterial travel speed, one needs to apply a mathematical model supplied with sensor data generated by roadside sensors and in-vehicle sensors. In this research effort, we develop a simple Bayesian estimator and an expanded neural network model to estimate arterial link travel speed. Input data used are from dual-loop detectors and probe vehicles with DSRC(Dedicated Short-range Communication) device. Data from one of model city, Jeonju, are used to generate test data for the simulation where the probe vehicle's speed is random sampled from observed vehicles' speed. Initial run shows that the neural network model developed can provide accurate estimates of arterial link speed using only probe vehicle's speed data.