Abstract: Due to the traits of nonlinear, capacity of study, adaptivity and anti-interference, neural-fuzzy network is suitable for the control of ramp metering. A neuro-fuzzy network controller is developed based on GA-BP for the deficiencies of the existing freeway on-ramp control. The controller configuration is determined and the controller parameters are designed in detail. The process of this design involves three parts: the selection of the input parameter and the output parameter, the design of the structure of fuzzy neural network and the algorithm design based on GA-BP. Finally, this neural-fuzzy controller is carried out by means of MATLAB software. The simulation for this controller shows that the method developed is more helpful to level off the density of main line than the method based on BP and the other methods based on ALINEA

Key words: genetic algorithm, freeway, on-ramp control, neurofuzzy networks

摘要： 鉴于模糊神经网络具有良好的非线性特性、学习能力、自适应能力和抗干扰能力，本文将模糊神经网络技术引入到高速公路入口匝道控制中。提出一种基于GA和BP算法的模糊神经网络控制器，并对控制器进行了详细设计。设计过程主要分为三部分：输入输出参数的选择、模糊神经网络的结构设计以及基于GA-BP的学习算法设计。最后，使用MATLAB软件对其进行了仿真。仿真结果表明，本文提出的方法是有效的，较之基于BP的模糊神经网络控制和ALINEA控制，能更好地稳定主线交通流密度。

关键词: 遗传算法, 高速公路, 入口匝道控制, 模糊神经网络