Vehicle Distributions in Large and Small Cities: Spatial Models and Applications

Abstract

Safety applications based on vehicle-to-vehicle (V2V) communication have become a major goal toward the intelligent transportation. The performance of the applications is mainly affected by the communication links, which in turn are governed by the topology characteristics of vehicular ad hoc networks (VANETs). To analyze the performance of V2V communication for VANETs, accurate spatial modeling is of great importance. There is an absence of a widely accepted two-dimensional (2D) model that well characterizes the random vehicle locations; the widely used Poisson point process (PPP) model is simple but not close to reality. In this paper, we concentrate on spatial point process modeling for random vehicle locations in large and small cities, performing empirical experiments with real location data of mobile taxi trajectories recorded by the global positioning system in Beijing city of China and Porto city of Portugal. We find that the empirical probability mass functions (PMFs) of the number of taxis in test sets in different regions of Beijing or in Porto all follow a negative binomial (NB) distribution. The spatial correlations of the points are established by comparing the results of different sampling methods. Based on the above, we show that the Log Gaussian Cox Process (LGCP) model, whose empirical PMF nicely fits the NB distribution, accurately characterizes diverse spatial point patterns of random vehicle location in both large and small cities. This is verified by the minimum contrast method. Then, we study the node degree as an important metric for the communication performance of the networks. It is shown that the connectivity of the LGCP model closely represents the connectivity found in the actual data, for both representative cities. The LGCP model is far more accurate than the widely used one-dimensional models and the 2D PPP for modeling the vehicle distribution, which is significant in V2V communication.