Accurate and speedy model of freeway traffic flow is the key for intelligent traffic control, but the results of existing freeway traffic flow modeling are not satisfactory. The wavelet transform is used to eliminate traffic noise and disturbance, and the traffic flow model is built based on a recurrent neural network in this paper. First, the freeway macroscopic traffic flow model is analyzed. Then the noise elimination method of wavelet transform is formulated, and Elman recurrent network is used for traffic flow modeling. The weights of the Elman network are obtained with an improved algorithm. Finally, a freeway with five segments, an on-ramp and an off-ramp is simulated. BP and RBF neural networks are chosen in contrast to the Elman network. The results show that the Elman network has the fewest training epochs, the smallest error and the best generalization ability. The fast learning ability and high dynamic performance of this recurrent network provide a novel and practical way to realize on-line modeling and control of traffic flow.
Freeway Traffic Flow Modeling Based on Recurrent Neural Network and Wavelet Transform
First International Conference on Transportation Engineering ; 2007 ; Southwest Jiaotong University, Chengdu, China
2007-07-09
Conference paper
Electronic Resource
English
Freeway Traffic Flow Modeling Based on Recurrent Neural Network and Wavelet Transform
British Library Conference Proceedings | 2007
|Freeway Traffic Flow Modeling Based on Neural Network
British Library Conference Proceedings | 2003
|Freeway traffic flow modeling based on neural network
IEEE | 2003
|