Speed prediction is essential for a wide range of traffic control measures, such as proactive control. Moreover, the estimation of individual vehicle speed enables more accurate traffic optimization and thus, is vital in the design of proactive traffic control method. Most of the previous studies investigated the speed prediction relying on macroscopic traffic data since there is a difficulty in obtaining realtime microscopic traffic data. This paper proposes a novel methodology bases on the kinematical characteristics of steady-state traffic flow for predicting the individual vehicle speed utilizing artificial intelligence, such as neural network and support vector machine. The microscopic and macroscopic traffic data sets used to train, test and validate were collected at a freeway stretch in the real world. The evaluation of the proposed model indicates that the proposed model is contributed to predicting individual vehicle speed.

CICTP