Neural Network Control Of On-Ramp On Freeway

Transportation is a general and guiding basic industry in national economy, as the fundamental facilities of modern transportation, freeway has become the skeleton transport mode and important channels with its characteristic of big capacity and high velocity. The traffic control of freeway is the important sign of improvement level of freeway. Besides, the research on ramp-metering control system can improve the freeway control system of our country.In recent years, the traffic volume of freeway has increased promptly, and this results in the increase of traffic jam and accident. These days, many freeways adopt ITS to decrease jam and guarantee the safe, fast and high-efficient operation.In this paper, firstly, we introduce the significance of research about ramp-metering, and show the internal and external status of ramp control algorithm. Secondly, it is a detailed introduction about several macroscopical models of freeway traffic flow. In the third chapter, BP neutral networks and RBF neutral networks are introduced.Lastly, due to traffic system is a heavily nonlinear, stochastic, time-variant and uncertain system, it is too difficult to construct an exactly mathematic model, Simultaneously, the neutral networks has widely applied to traffic flow control of freeway. We construct a model-free control system of traffic flow that includes a neutral network. The output of neural network is the transit rate that is control measure, and we detailed show the frame of neural network, here the parameter estimation is completed by simultaneous perturbation stochastic approximation (SPSA). Analysis and simulation show that this algorithm is more robust, self-adaptive, self-learning,and can solve the traffic control problem effectively and it is better real time control than other neutral networks.