Abstract
The average travel speed on certain road sections is an important piece of information for the intelligent transportation system. Traditional ways for estimating travel speed usually depend on dedicated sensors or infrastructures, which is financially costly. Alternatively, as an infrastructure-free way, vehicular crowdsensing can be used to collect data including real-time locations and velocities of vehicles, which is quite low cost and effective. This paper aims to produce a fully covered distribution of average travel speeds for road sections both in time and space domains based on vehicular crowdsensing data. However, due to the uneven spatial-temporal distribution of vehicles and the variation of their data-offering intervals, vehicular crowdsensing data are usually coarse grained. This coarseness leads to missing travel speed values of vehicles on some road sections. To handle this problem, we propose an approach that exploits the spatial-temporal causality among travel speeds of road sections by a time-lagged correlation coefficient function. We use a time-lagging factor to quantify the time consumption of vehicles traveling along road sections. Then, we utilize the local stationarity of correlation coefficient to estimate the travel speeds of road sections. Experiments based on real taxi trace data show that the proposed method performs better than some methods in use.