Estimating dynamic transport population for official statistics based on GPS/GSM

Transport and traffic data collection methods have developed along two main directions. One is the off-line method using pen-and-paper surveys or face-to-face interviews, and the other is the on-line method using internet surveys or XML messages from companies' administrations. The transport and traffic statistics based on these data are rather static representing an average, macro-view of transportation. In order to further enhance transport and traffic statistics, linking traffic to social economic fields, real-time collected data based on GPS/GSM should be available. Marked advantages of GPS and GSM data collection are that transport and traffic data are captured automatically at highly frequent rates, which provide dynamic information and offers opportunities to reduce administrative burdens. And also with the help of GPS information, the individual vehicles can be identified by connecting with other data source. However, though GPS/GSM data collection sounds a promising technology, its adoption is seriously hampered by the fact that few vehicles are equipped with GPS transponders and that not all drivers use a GSM. GPS/GSM-collected data thus constitute only a limited part of the whole transport and traffic data for a delineated area and specific time slots. In this paper, we aim to validate the usage of GPS information for traffic statistics. Traffic density is theoretically applied to integrate traffic engineering, social economy and statistics methods. Then with the limited number of GPS/GSM vehicles, we take their advantages of doing the timing and identification issues for carrying out density estimation. Further, statistic method of Horvitz-Thompson is used to up-scale the dynamic density to the whole network and time period for official statistics.