RBF网络在交通流模型辨识中的应用

利用径向基函数 (RBF)人工神经网络来逼近已知的交通流非线性解析模型 ,讨论了高速公路交通流模型的辨识问题。提出了一种带反馈的 RBF网络模型 ,讨论了其训练算法。算法分两步实现 ,第一步利用一种改进的聚类分析方法确定隐层节点核函数的中心点 ,第二步用最小二乘法确定从隐层到输出层的连接权。最后将训练好的网络模型和给定的解析模型同时进行仿真计算 ,得出了当某路段出现突发性交通事故时交通流密度和平均速度的变化曲线。仿真结果说明 RBF神经网络模型的训练速度快和辨识精度还是令人满意的

Radial basis function neural network for a traffic flow model

This paper presents a traffic flow model using a radial basis function neural network (RBFNN) with feedback to approximate the well known nonlinear analytical traffic flow model. The improved cluster analysis training method is used to determine the cential kernel function of the hiden layer nodes. A least squares method is used to obtain the weights from hiden layer to the output layer. The change of density and average velocity of the traffic flow of traffic accident are calculated using an analytical model and the RBFNN model. The simulation shows that the training speed and the identification accuracy are quite satisfactory.