基于速度的城市快速路交通拥堵预测研究

A Traffic Congestion Prediction Model Based on Vehicle Speed of Urban Expressways

交通拥堵预测是解决交通拥堵问题的重要步骤之一。为缓解交通拥堵,选取速度这一参数建立交通拥堵预测模型。在对速度时间相关性和空间相关性分析的基础上,提出了基于时空特性和径向基神经网络的速度预测多点模型。将预测结果与决策阈值相比较,粗略地判定拥堵等级,并运用模糊算法对速度和拥堵程度进行量化,建立相应的模糊规则体系,并应用模糊逻辑推理得到定量的拥堵度指标。结合实例进行仿真和分析,与基于单一时间序列的预测方法相比较,基于时空特性的预测方法的平均绝对相对误差由7.45%下降到了3.61%,有效地提高了速度预测精度,基于速度的拥堵预测模型的识别准确率较高。利用模糊算法评判拥堵程度,可得到量化的拥堵度指标,使拥堵程度一目了然。

The prediction of traffic congestion is an important step for mitigating the issue itself.In this paper,vehicle speed is used as a parameter to develop a model for predicting traffic congestion on urban expressways.Through the analysis of the temporal and spatial relevance of vehicle speed,a multi-location prediction model of speed is proposed using radial basis function（RBF）neural network and based on the temporal and spatial distribution of vehicle speed.The level of congestion is roughly determined through comparing the predicted value with the predefined decision thresholds.Moreover,a fuzzy algorithm is used to quantify the speed and the degree of congestion,and thus construct a set of fuzzy reasoning rules is also constructed,with which is adopted to obtain quantitative measure of the level of congestion.A case study consisting of simulation and numerical analyses shows that this proposed prediction method based on the temporal and spatial distribution of vehicle speed can effectively improve the prediction accuracy.When compared with the prediction method developed using a single time series,the mean absolute relative error decreases from 7.45%to 3.61%.This proposed prediction model performs higher recognition accuracy.Moreover,the degree of congestion can be determined with the proposed fuzzy algorithm,and thus a quantized congestion index can be developed,which makes the degree of congestion much clearer.