A novel online wave-net based model for motorway traffic modeling

Abstract

In this paper, a novel method of wave-net modeling for motorway traffic system is proposed. Due to the highly nonlinear behavior of the traffic system, and the capabilities of wave-nets in multi-resolution analysis and nonlinear function approximation, the use of the wave-net as a modeling tool for traffic system seems to be very promising. We have also utilized the principle component analysis technique to process the inputs of the network, in order to reduce the network dimension while preserving its accuracy. Because of the time varying nature of the traffic system, the need of an algorithm for updating the model is vital in order to guarantee the performance of the system and the control strategies. Hence, in this work we have used online L2 learning algorithms to update the model. Simulation results, which are based on the real traffic data, demonstrate the increased efficiency of the online model compared to the offline model without update.