Finding Vacant Taxis Using Large Scale Gps Traces

In modern cities, more and more vehicles, such as taxis, have been equipped with GPS devices for localization and navigation. The GPS-equipped taxis can be viewed as pervasive sensors and the large scale traces allow us to reveal many hidden "facts" about the city dynamics. In this paper, we aim to estimate the wait time and probability of taking a vacant taxi according to time and position. Further more, we provide recommendations for passengers who want to take a vacant taxi. To achieve these objectives, firstly we preprocess the large scale taxi GPS traces data set to generate the Map Grid Based (MGB) index. Secondly, with the MGB index, we apply the nonhomogeneous Poisson process corrected by the conditions of road and weather (NPPCRW) method to perform estimation and recommendation. We build our system based on a large scale real-world GPS traces data set generated from more than 12000 taxis in Beijing over a 110 days period. Then we validate the system with extensive evaluations including in-the-field user studies.