Traffic speed information is generally estimated using widely available taxi GPS data. However, the collected taxi GPS data typically have limited spatial coverage because taxi use is preponderantly restricted to areas of high population density. Moreover, the correlation between taxi speeds and the speeds of other vehicles is not altogether clear. The present study addresses these issues by proposing an improved method of traffic speed estimation involving taxi, bus, and mobile phone GPS data. Here, the traffic speeds of the different types of vehicles represented in the GPS data are analyzed to generate a mixed distribution of the traffic speeds of different vehicle types. The generated mixed distribution is validated using license plate recognition data. Keywords: Traffic speed, GPS data, Data fusion, Trajectory data

CICTP