Research And Simulation On Neural Network Predictive Control For Freeway Ramp Based On Elman

Freeway traffic control is considered as an important component of intelligent transportation system.For the drawbacks of traditional traffic control techniques,people have begun to apply Aartificial Intelligence (AI)and Computational Intelligence (CI)to freeway traffic control in recent years. Neuralnetwork predictive control (NNPC) takes full advantage of the nonlinear,self-organizing, self-learning performance of neural networks and the rolling optimizing, feedback adjusting effectiveness of predictive control. It is more suitable for this Freeway traffic control.This paper based on how to make use of an existing resources of freeway well, exaltation traffic volume for purpose, carry on thorough research of on-Ramp Meter,the freeway traffic flow forecast and controller is Designed and Simulated based on particle swarm algorithm.The following researches are involved in this paper:Firstly,the macro-dynamic model of highway traffic flow is analyzed.The Elman recurrent neural network model is built and an PSO algorithm is used to train the weights and of the neural network. compared with the standard BP algorithm,PSO which using real coding and, has more simple structure and faster convergence study, The simulation results show that the model suitable for highway traffic flow short-termforecasting.Secondly,in view of the transport system is non-linear, dynamic and strong disturbance, in order to further enhance the performance of NNPC,a non-linear optimization controller based-on PSO is presented,using PSO algorithm to achieve control of the rolling optimization, and simulation results achieved freeway traffic flow neural network control system, and the forecasting model of the system, and rolling optimization algorithm,feedback correction,simulation parameter settings problems were analyzed.Finally,Predictive Controller was simulated and verificated, with the measured data of a highway and results showed that, compared with no control and coordination based on genetic algorithms, the proposed control strategy in this paper has a better performance indicators in reducing the average waiting time on ramp and improving service performance.