朱晓东

       城市快速路入口匝道作为连接地面道路与快速路的重要环节，驶入车流量大、车辆换道行为频繁，易造成快速路主线下游拥堵，并向快速路上游传播；同时入口匝道处排队往往会蔓延至相邻交叉口，甚至造成排队溢出，严重影响地面交叉口群的通行效率。因此，综合考虑快速路主线、入口匝道和毗邻交叉口群的协同控制对于协调快速路与地面道路相互作用关系，平衡快速路和地面道路的交通需求，提高整个混合道路系统的交通效率有重要作用。本文利用交通采集系统和交通控制系统提供的交通信息，深入挖掘入口匝道与快速路主线和毗邻交叉口群的流量相关性，提出交叉口群关键路径识别方法和综合考虑快速路主线、入口匝道、交叉口群的协同控制策略。主要研究内容包括以下几个方面：

     （1）以上海市南北高架东侧徐家汇入口匝道影响区域为例，首先针对采集的交通流数据存在的数据缺失与异常数据问题进行处理，再应用经验模态分解法对交通流数据进行去噪，并对快速路主线、入口匝道和交叉口群进行交通流特征分析，为后续研究打下数据基础。

     （2）应用基于动态时间弯曲距离算法的时间序列相似性度量方法，以入口匝道断面为基准，分别度量其与快速路主线各断面及交叉口群各交叉口进口道的流量变化特征相似性。基于流量变化特征相似性的度量结果及其影响因素对交叉口群内各交叉口进口道进行K-Means聚类分析，为实现交叉口群内路径的分级控制提供理论依据。

     （3）应用隐马尔可夫模型描述各交叉口进口道断面流量时间序列和隐含的入口匝道与交叉口相关状态序列之间的相互作用关系，构建了交叉口群的路径关联度识别模型。之后，提出改进的交叉口群路径关联度模型对路径关联度识别模型的初始参数进行标定。利用路径关联度识别模型预测识别交叉口与入口匝道的相关性得到交叉口群的关键路径。通过不同案例的识别结果表明此模型具有良好的实际意义与应用价值，为后续交叉口群和入口匝道协同控制提供基础。

     （4）提出了综合考虑交叉口群关键路径、入口匝道和快速路主线的协同控制策略。通过结合路段承载车辆数模型、宏观交通流METANET模型和ALINEA算法构建协同控制模型，建立相应的目标函数和评价指标。采用MATLAB软件针对不同的仿真场景进行数值仿真，验证了控制策略的可行性和有效性。

关键词：入口匝道，交叉口群，动态时间弯曲，隐马尔可夫，协同控制

    The expressway on-ramp, as an important link to connect the ground road and the expressway, is easy to cause congestion in the downstream of the expressway and spread to the upstream of the expressway because of the large traffic flow and frequent behavior of changing channels. Meanwhile, queuing at the on-ramp will often spread to adjacent intersections and even cause queuing overflow, which will seriously affect the traffic efficiency of the ground intersection group. Therefore, the coordinated control of the main line, the entrance ramp and the adjacent intersection can play an important role in coordinating the interaction between the expressway and the ground road, balancing the traffic demand of the expressway and the ground road and improving the traffic efficiency of the whole mixed traffic system. In this paper, the traffic information provided by the traffic collection system and the traffic control system were used to excavate the flow correlation between the entrance ramp, the main line of the expressway and the adjacent intersection. The key path identification method of the intersection group and the coordinated control strategy of the main line, the entrance ramp and the intersection group are put forward. The main research contents are as follows:

   (1) Taking the influence area of the Xujiahui entrance ramp on the east side of the Shanghai North& South Viaduct as an example. First of all, the missing and abnormal data problems in traffic flow were dealt with, then the traffic flow data were denoised using the empirical mode decomposition method, the characteristics of traffic flow were analysed on the main line of the expressway, the ramp and the intersection group, which laid the data foundation for follow-up research.

  (2) The time series similarity measure method based on dynamic time warp(DTW) algorithm was used to measure the similarity of flow variation characteristics of each section of the main line of the expressway and each approach of the intersections. Based on the measurement results of the similarity of flow change and its influencing factors, the K-Means clustering analysis were used to provide a theoretical basis for the hierarchical control of paths in intersection group.

  (3) The hidden Markov model (HMM) were used to describe the interrelationship between the time series of the section flow and the implicit relationship between the entrance ramp and the related state sequence of the intersections, and the path correlation recognition model of the intersection group was constructed. Then, an improved path correlation recognition model of the intersection group was proposed to calibrate the initial parameters. Using the model to predict the correlation between the intersection and the ramp then identify the critical path of the intersection group. The identification results of different cases show that the model is of good practical significance and application value, which provided a basis for the coordinated control of the intersections and entrance ramps.

  (4) A coordinated control strategy considering the critical path of intersection group, the on-ramp and the main line of the expressway was proposed. By combining the model of vehicle bearing capacity, the METANET model of macro traffic flow and the ALINEA algorithm, the cooperative control model was built, and the corresponding objective function and evaluation index were set up. MATLAB was used to carry out numerical simulation for different simulation scenarios, which verified the feasibility and effectiveness of the control strategy.

Key Words: On-ramp, intersection group, dynamic time warp(DTW), hidden Markov model (HMM), coordinated control