基于分层贝叶斯网络的交通密度估测模型

Road network density estimation model based on hierarchical Bayesian network

为准确、快速地判别交通状态,以宏观基本图为基础,针对交通流参数时、空二维分布的特征,建立了分层贝叶斯网络下的路网密度估测模型。采用分层贝叶斯网络描述了影响交通状态变量之间的相关性,并结合概率理论和路段排队模型进行了公式的推导,引入期望最大化扩展卡尔曼滤波法(EM-EKF)进行了参数估计和未知变量的迭代计算。以玉溪市龙马路等相关路段为例,借助VISSIM软件进行了仿真实验。通过COM接口采集密度等数据,将这些数据与模型计算结果进行了对比分析。研究结果表明:提出的模型估测结果十分接近真实情况(平均绝对误差百分比为6.98%和7.72%),验证了该模型的有效性和可靠性。

In order to accurately and quickly estimate traffic state, a density estimation model based on the hierarchical Bayesian network is proposed by studying the fundamental diagrams and traffic flow. The hierarchical Bayesian network is used to describe the correlation among the random variables. The probability theory and the link queue model are used to infer in-flux and out-flux. The expectation- maximization extended Kalman filter (EM-EKF) is introduced to estimate unknown parameter and variables. Taking Longma Road sections in Yuxi City as an example, the experiment is implemented. The state variables are compared in the simulation model, which acts as ground truth, and the state variables derived from EM-EKF. The results show that the model data is very close to the real situation (the average absolute error percentage is 6.98% and 7.72%), which verifies the validity and reliability of the proposed model.