【作者】 路加；

【导师】 曹丽；

【作者基本信息】 清华大学 ， 控制科学与工程， 2003， 硕士

【摘要】 随着社会经济的迅速发展，许多城市道路的通行能力已无法满足日益增长的交通需求，交通拥挤堵塞现象日趋严重。智能交通系统的关键功能，就是要能够管理和控制交通流，避免交通堵塞现象的发生；一旦发生交通拥挤和堵塞，能够及时快速地提供有效的解决方案，疏导交通流。拥挤度的检测与评估，是正确审视交通拥挤的前提性工作，具有着重要的现实意义。本文从讨论交通拥挤的成因入手，结合国内外交通拥挤的评估方法，提出了交通拥挤度的概念，建立了能够连续描述交通拥挤程度，并可以体现人们对交通拥挤感觉的指标。通过分析交通流基本参数与交通拥挤现象之间的关系，本文运用模糊数学的方法对交通参数（交通流量和平均速度）以及拥挤度进行描述，建立了交通拥挤度的模糊推理系统。为验证交通拥挤度和模糊推理系统的可行性，本文进行了软件仿真实验。利用软件获取的仿真数据，以及人们对仿真实验所呈现的交通拥挤程度的评价数据，文章对模糊推理系统参数进行了辨识。辨识采用了两种方法，即基于最小二乘法的模糊辨识方法和基于自适应神经网络的训练。后者较好地完成了训练任务。通过分析模糊推理系统的推理曲面，验证了模糊推理系统的可行性，它能够较好地支持交通拥挤度这个概念。然后，本文提出利用浮动车检测系统作为交通参数获取手段的方法，阐述了浮动车检测系统的框架及检测流程，着重介绍了利用GPS数据计算交通参数的方法。针对交通拥挤度，文章提出了利用浮动车检测系统判别常发性和偶发性拥挤类型的方法，并根据上述推理系统框架，给出在非偶发性拥挤下的拥挤度计算步骤。更多还原

【Abstract】 Traffic congestion becomes more and more serious with the rapid development in modern society. Traffic management and control are aimed to avoid traffic jam and give effective solution of releasing transportation pressure, which is the key function of Intelligent Transportation System (ITS). Therefore, it is important to detect and estimate the traffic congestion. With the evaluation result of traffic congestion in advanced traffic information system, traveler may find the best way against traffic jam, and traffic managers can be benefited by such historical data analysis in layout of city road nets.In the thesis, with the analyzing of traffic congestion causes, many methods to evaluate traffic jam are reviewed, from which, Level of Congestion (LOC) is defined as an index to describe a continuous traffic processing from free flow to traffic jam. LOC should also be of coherence with human perception on traffic congestion. According to the relationship between characteristics of the traffic flow and traffic congestion, fuzzy logic is adopted to partition the basic parameters (mean velocity and flow volume) into several fuzzy subsets, then fuzzy inference system is established to generate LOC.To test the feasibility of LOC and fuzzy inference system, a simulation tool is used to collect the data of traffic flow and visually evaluate the traffic congestion by people. Two methodologies are used to obtain the system model. One is fuzzy identification based on the least mean square method and the other is based on adaptive neuro-fuzzy inference system. As a result from fuzzy inference surface analysis, LOC can be supported by the fuzzy inference system and can recover human’s general perception on judging congestion.In the last section, probe vehicle system is discussed as a means of data collection. The framework and detection process of probe vehicle system are depicted. Particularly, GPS data processing and traffic parameter evaluation<WP=4>based on probe vehicle reports are developed. A method to judge the type of traffic congestion is studied, and according to the previous result, the step to calculate LOC in the type of non-recurrence congestion is finally presented.更多还原