Bayesian Network Method of Speed Estimation from Single-Loop Outputs

Real-time and accurate traffic speed is important for a successful traffic management system. However, the most common form of the single-loop detector is incapable of providing speed measurements. This paper presents a method of speed estimation from single-loop detector data using Bayesian network method. After analyzing the causal relationship between volume, occupancy, and speed, a Bayesian network model of speed estimation is proposed using volume and occupancy from single-loop outputs. The Gaussian mixture model (GMM) and the expectation-maximization (EM) algorithm are used to represent model and train model parameters, respectively. The proposed method is implemented and evaluated using the field data from urban expressways in Beijing. Estimated speeds are compared with the observed speed data and also with results from conventional algorithm. The results show that the proposed method is robust for every kind of sampling intervals, lanes, and traffic condition. The mean absolute error holds more than 2 km/h decrease. This method can be efficiently applied in traffic management system.

摘要实时精确的车流速度对于交通管理系统来说是至关重要的。然而, 最普遍的单线圈检测器却不能输出速度参数。本文提出了一种新的单线圈检测器速度估计的贝叶斯网络方法。在分析流量及时间占有率与速度之间的因果关系基础上, 通过单线圈检测输出采样间隔内的流量和时间占用率数据, 建立了速度估计的贝叶斯网络模型, 采用高斯混合分布函数和EM算法进行模型表达及参数训练。通过北京快速路实地数据对算法进行了验证, 结果表明算法在不同采样间隔、不同车道及不同交通状态下均具有较强的鲁棒性, 与传统算法相比平均绝对误差减少2 km/h左右。这一方法可以应用于交通管理系统速度的估计。