Short-term travel time prediction on urban road networks using massive ERI data

Providing real-time and accurate travel time can help road users plan their trips to minimize travel time. Existing methods of travel time prediction mainly employ the data of floating car and loop detectors, which are limited to the sources. Those data only provide information of some vehicle and reflects partial traffic information. With the development of RFID technology, Electronic registration identification (ERI), which employ RFID technology to track individual vehicle, is the electronic ID card of motor vehicle. The collecting ERI data can reflect whole traffic information. The purpose of this paper is to develop short-term travel time prediction model based on ERI data. Firstly, the paper introduces the collection principle of ERI data and the travel time of travel segments using the ERI data. Then the short-term travel time prediction model on urban road networks based on KNN algorithm is constructed, including steps such as constructing eigenvectors, cross-validation method to determine K value and local estimation method. Compared with the historical average model and the autoregressive moving average model, experimental results reveal the KNN outperforms the other two models.