Research On Transportation Detection Based On The AGPS-abled Cellphone

This article presents a methodology for identifying travelers’ transportation modes by tracking Assisted Global Positioning System(AGPS)-equipped mobile devices in the traffic stream. Various mobile phone service providers have location-based services(LBS) that track the locations of their mobile phones. One major concern in using mobile phones for traffic monitoring is that the phones are not necessarily in passenger vehicles. The mobile device can be in a car, bus, or other modes that have distinct speed and acceleration profiles. In addition, querying the mobile device has monetary cost implications, and the higher the number of location queries from the server the higher the associated cost. This article focuses on the feasibility of using thecharacteristics of the trail of AGPS data stream to identify the mode on which the mobile device is located. Because of the sampling limitation, a GPS data logger software is used to collect the trip data and the logged data is sampled at varying frequencies as if they are coming from the mobile phones.The analysis is conducted using BP-neural networks to determine the transportation mode. The analysis also examinesthe impact of varying sampling rates (number of pings per unit time) and monitoring duration (time length of data trail) on mode classification accuracy. In total,18h of AGPS data were collected while traveling on various transportation modes throughout the2nd ring road,Beijing. Results confirm the potential of neural networks to successfully detect transportationmodes from AGPS data, both in peak and nonpeak periods. The results indicate that higher sampling frequency and longe r monitoring duration result in higher mode detection rates. In addition,the BP neural networks has better performance when detecting non-auto(bus and walk) than the auto transportation modes,and it is interesting to note that the neural networks find it easier to detect modes during peak periods...