An averaging approach to estimate urban traffic speed using large-scale origin-destination data

Abstract

Knowledge about the driving condition can be exploited by the energy management system of plug-in hybrid electric vehicle (PHEV) to optimise its fuel economy. However, estimating urban traffic speed can be a challenging problem, since the complexity of the road network requires a large amount of sensing coverage. GPS equipped vehicles have been used to address this problem, despite the drawback of high transmission cost due to the periodic locations reporting during the entire trip. This paper presents an alternative method to estimate urban traffic speed when only the trip distance, origin, destination and trip time are known, without the detailed information about the vehicle trajectory. Taxicab fleet data is used due to its wide area-coverage and large data availability. One month (15 million) observations of taxi origin-destination data in New York City is processed using cluster computers to create a traffic speed model of Manhattan at different times of day. The developed average traffic speed model is able to accurately predict the recorded trip duration from the taxicab data, and therefore, can be used to estimate energy consumption and optimise PHEV control.