Research On The Evaluation Of Traffic Condition Using Mobile Devices

Road traffic jam is definitely a major source of inefficiency and wasted fuel. Measuring and localizing the congestion is an important step towards reducing the time people spend stuck in traffic, it is also one of the important issues in intelligent transportation systems. There are mainly three kinds of traffic state estimation systems now, namely reling on the road-side fixed sensors, wireless detection and location and reling on the vehicles equipped with GPS receiver. Road-side fixed sensors (such as loop detectors, RFID readers, video cameras) and vehiclar communication module have coverage limitation and high installation and maintenance costs. The use of a wireless base station for detection and localization is severely affected by the mobile network. All of them can not achieve good performance in traffic estimation. With the rapid development of phones, smart phones can provide location estimates with GPS sensor. Using position samples from drivers’phones to monitor traffic delays opens up the possibility of road condition estimation.This paper describes a general system architecture using the GPS data obtained from mobile phones. First is to evaluate the link average speed of a single vehicle, piecewise cubic Hermite interpolation method is employed to improve the accuracy of evaluation. Concerning that the speed of the same type vehicle is relatively stable, next is to obtain this, parameter using the method of weighting. Again, use the average speed of different types of vehicles and ratio of those types as the inputs, and the link average speed as the output to construct the RBF (Radical Basis Function) neutral network. With the network, we can get the link average velocity. Last, we introduce some traffic parameters characterizing road traffic conditions, and a new road traffic estimation model based on the Greenshields velocity-density inference model.The experimental and analysis results proves that the piecewise cubic Hermite interpolation method, the fusion, method and the RBF network method all decrease the estimation error. And with low penetration rate, we saw a significant decline in the evaluated error of section average velocity and traffic density with the new road traffic estimation model.