Improving the Accuracy of Real-Time Traffic Data Gathered by the Floating Car Data Method

Fajar Yoseph Chandra
Binus Graduate Programs
Bina Nusantara University
Fergyanto E Gunawan
Binus Graduate Programs
Bina Nusantara University (f.e.gunawan@gmail.com)
Garry Glann
School of Computer Science
Bina Nusantara University

Abstract: Despite the concern of privacy, the method of Floating Car Data (FCD) is clearly one of the cheapest methods to provide real-time or near real-time traffic information. The method has become more affordable with the proliferation of smart-phones and with the existing infrastructure of the wireless network. In conjunction with the virtual trip lines (VTL), the FCD method can mimic the traditional traffic monitoring method on the basis of loop detectors. In addition, the use of VTL also helps the FCD method in preventing the potential of tracking the probe vehicle, which is used to gather and report the traffic information. Although vast publications regarding the FCD method are available, the issues of the optimal length of the VTL, the timeliness of the data, and the accuracy of the geo-location data have not been discussed. This article focuses and addresses these issues empirically using data collected by a probe vehicle traveling along Jakarta Inner Road Highway in Jakarta, Indonesia. Those data are collected using ten smart-phones of the same type. As the result, the optimal trip line length can be established as a function of the level of accuracy of the geo-location data. In the case where the level of accuracy is 32~m, we determine a VTL length of 26 m should provide 95% chance that the probe vehicle will cross the line. In addition, the currently developed system can also provide about 80% of the traffic data in less than 1 min, and the remaining 20% data within 1--5 min interval. Finally, by applying a simple moving average filter, the prediction of the traffic velocity can be increased significantly, and the geo-location data error can be reduced up to 20%.

Table of contents:
1. Introduction
2. Description of floating car data system
3. Method
4. Results and analysis
5. Conclusion