Morteza Montazeri-Gh - Associated Professor, Systems Simulation and Control Lab, Department of Mechanical Engineering, Iran University of Science and Technology, Tehran, Iran
Abbas Fotouhi - PhD Student, Systems Simulation and Control Lab, Department of Mechanical Engineering, Iran University of Science and Technology, Tehran, Iran

In this study, traffic condition recognition and prediction is performed using real velocity data recorded in the city of Tehran. Data gathering was done using the global positioning systems (GPS) which saved vehicle’s velocity each second. Using the velocity time series, traffic groups are defined. Average velocity is used as a characteristic parameter of the traffic groups. The traffic groups are classified into four traffic conditions regarding to their average velocity. After driving condition classification, neural networks are used for driving condition prediction. A radial basis function (RBF) network is utilized for forecasting the traffic condition in the near future. Using the RBF network, the percent of correct predictions achieve to 95%, 82% and 65% for 1, 5 and 60 seconds ahead respectively.

Driving Pattern Recognition, GPS, Prediction, Neural Networks