Potential Predictability of vehicles´ visiting duration in different areas for large scale urban environment

The newly emerged vehicular communication network is seen as a key technology for solving the increasingly serious vehicular traffic congestion as well as improving road safety, and applications of vehicular networks are also emerging at the same time. Predicting vehicular visiting time is vital to both solving the vehicular system problem and building efficient vehicular networking. It is an open and unsolved problem that how much the vehicular staying duration of visits to different areas can be predicted. In this paper, we use real vehicular traces in Beijing and Shanghai and an area partition model to explore the limit of predictability of the vehicular visiting time in different areas in large scale cities and analyze the effect of different precision and slot time on predictability. We conclude that using a proper time slot is an efficient way of prediction and a higher predictability can be achieved if the requirement for precision is reduced. Among all the cases we study, we find the highest potential predictability in Beijing and Shanghai in the case with smallest slot time and lowest requirement for precision.