Intelligent Learning Control Approaches For Freeway Traffic System Based On The Distributed Parameter Model

With the development of economy and the speeding up of city modernization, urban traffic congestion has become a very challenging problem worldwidely. At present, on-ramp metering is one of the most effective approaches to improve the situation of freeway traffic. This control scheme aims at limiting the traffic flow entering the freeway mainstream according to the traffic condition on the freeway mainline, so as to ensure the freeway system be on the desired state. In practice, freeway traffic system is a kind of complex large system with strong nonlinearities and uncertainties. And it is influenced by various kinds of external disturbances and interferences. How to apply advanced intelligent control methods to the control of freeway traffic system has become a hot topic for the traffic engineers and research scholars.In this paper, we introduce the distributed parameter model of the freeway traffic flow system and propose a series of intelligent learning control methods, including fuzzy control and the iterative learning control. The main work and key innovations are summarized as follows:1) Most of the freeway traffic control scheme are developed and analyzed by using the ordinary differential equation (ODE) model, which ignores the actual space-time evolution characteristics of freeway traffic system. To overcome this problem, this work introduces the distributed parameter model of freeway traffic system, which uses the two-dimensional partial differential equations of space and time to describe the traffic flow system interior characteristics. The dynamic characteristics of freeway traffic system based on distributed parameter model are not only related to the information of time dimension, but also related to the spatial dimension, which are more close to the actual freeway traffic system.2) For the practical freeway traffic control problem, a new PI boundary control method based on the freeway traffic distributed parameter system model is proposed with taking the boundary conditions into consideration. Further, to overcome the problem that the selection of PI controller parameters requires accurate model information and expert experience, a fuzzy self-tuning PI boundary control method is developed to use the fuzzy control logic to tune the parameters of PI controller just according to the system feedback information. The proposed fuzzy self-tuning PI controller has strong adaptive ability, and can effectively suppress the influence of the system disturbances. Compared with the control method based on the differential model, the simulation results illustrate that the tracking performance of the proposed method is better.3) Considering the repeatability and periodicity of freeway traffic system, the iterative learning control approach for distributed parameter system is explored firstly, and then an iterative learning ramp control method for freeway traffic system is proposed. Differently from existing iterative learning ramp metering, the error information used in the proposed ILC law is not only a function of time, but also related to the spatial location. In the other word, different from the ILC method based on macroscopic traffic flow model, the proposed method considers the space and time characteristics of freeway traffic system such that it can greatly improve the convergence speed and the control performance of the freeway traffic control system. The simulation results further confirm the effectiveness and applicability of the proposed ILC approach.