Intelligent Learning Control Method For Freeway Traffic On-Ramp

As we all know, freeway traffic system has strong nonlinear, fuzziness and uncertainty, and it is also vulnerable to a variety of outside interference, such as weather, road, and drivers traffic behavior. Thus, how advanced intelligent control technology, information fusion technology and intelligent information processing technology are applied to freeway traffic system control has become a hot research for scholars now. In this research, in order to further enrich and improve the application of fuzzy control and iterative learning control for on-ramp control, the intelligent learning control method for freeway traffic on-ramp was focused on. The main contents and key innovations are summarized as follows:1) The models of freeway traffic on-ramp were derived in detail and summarized (such as differential form, differential form, distributed parameter form), especially the extended macroscopic traffic flow model, where the variable of queue length of on-ramp is considered and whose form is differential. In addition, the freeway traffic distributed parameter system model also is also given. Taking the extended macroscopic traffic flow model for example, whose form is differential, this thesis gives the control objectives and performance evaluation (average travel speed) of freeway traffic on-ramp.2) Based on the differential equation model of the ramp system, a new fuzzy self-tuning PI controller for freeway ramp metering is designed in this project. By considering the upstream and downstream traffic flow information sufficiently, the designed PI controller has a better tracking desired density, and it can not only effectively suppress the overshoot and influence of the system disturbances, but also achieve soften control. Furthermore, a fuzzy logic is designed to tune of the PI controller parameters automatically. Compared with the traditional PI controller, ALINEA controller, the simulation results illustrate the validity and efficiency of the presented method intensively. Meanwhile, it has better tracking characteristics, and the steady-state error tends to zero faster.3) Based on the extended macroscopic traffic flow model, where the variable of queue length of on-ramp is considered and whose form is differential, this research presents a new iterative learning control approach for the mainline traffic flow density via ramp metering by virtue of the repeatability feature of the traffic system. Making full use of the errors of traffic flow density and on-ramp queue length in the previous conduction, the proposed approach is able to realize the tracking of desired density and queue effectively. And the convergence is proved with rigorous mathematic analysis. The simulation results further confirms the effectiveness and applicability of the proposed approach considering the queue length factors. Consequently, this method has a strong ability to suppress disturbance, as well as better transient response performance and traffic capacity.