【作者】 丁恒；

【导师】 陈无畏；

【作者基本信息】 合肥工业大学 ， 车辆工程， 2011， 博士

【摘要】 信号交叉口交通流受驾驶技术、交通组成、信号控制、交通管理、交通环境等诸多因素影响，导致交通流在时间和空间内通常具有多态特征。现有的信号控制方法因未能充分考虑交叉口交通流分布特征，造成交叉口成为道路网络交通延误增长的主要区域，影响整个路网的运营效益。为了解决现有的控制方法不能适应多态交通条件下交叉口信号控制问题，本文依托国家自然科学基金项目《混合交通环境下行驶工况的仿真及控制策略研究》（No.70771036）和《平面交叉口交通流片段仿真及管理措施实证研究》（No.71071044）课题，通过分析多态交通流特性，研究交叉口通行能力模型及交通流预测方法，进一步提出多态交通条件下交叉口信号控制方法及延误分析理论。本文研究工作主要表现在以下几个方面：（1）通过分析多态交通流时间和空间分布多态特性，从人、车、路三个方面分析影响交通流多态特性影响因素；提出了多态交通流与相对稳态交通流的界定方法，标定了车头时距、交通流量两项统计指标的界限值。（2）结合稠密特性函数可以任意精度逼近任何分布特征数据的特点，引入稠密性无限混合Gamma分布函数，提出了基于极大似然算法的交通流车头时距拟合方法，并通过实际调查数据证明了拟合方法的精确性。（3）基于可穿越间隙理论，采用无限混合Erlang稠密性分布函数，分析了次要道路交通流可穿越主路交通流车头空隙的概率，建立了无信号交叉口通行能力模型。（4）基于小波分析理论，采用RBF神经网络和Markov链预测方法，建立了短时多态交通流组合预测模型。（5）根据多态交通流在日交通流量统计序列的变化特性，通过小波分析对日交通流量序列信号进行降噪处理，结合F.Webster-B.Cobber信号配时理论提出了多态交通特性信号交叉口控制时段划分方法。（6）根据交叉口多态交通流分布特征和时空资源条件，分定时控制、感应控制和预测控制三种类型提出了信号控制方法：①针对非对称多态交通流交叉口，根据交通流量通过调整交叉口空间资源，利用相位优化技术提出了交叉口信号伴随相位定时控制方法。②针对交通流量多变、进口车道较少的小型交叉口，基于交叉口车辆排队理论建立了交叉口每相位主流向通行综合效益PI函数，依据交叉口流向冲突关系，提出了相位自组织优化的交叉口感应信号控制方法。③针对进口车道较多的多态交通交叉口，通过在交叉口进口设置多功能进口车道，依据排队理论和多态交通流预测结果，建立以交叉口总运行延误最小为优化目标的多功能车道流向选取模型。在相位有效绿灯时间满足最大排队长度通行条件下，滚动优化每相位并逐步实施，提出了设置多功能车道的交叉口时空资源动态优化的预测控制方法。（7）基于随机过程理论，提出了多态交通流量稠密特性函数概率统计方法；结合定时控制、感应控制与预测控制不同策略下信号相位变化状态，建立了多态交通流信号控制延误模型。（8）使用Vissim及相关软件对多态交通流信号控制方法和延误分析理论进行仿真，证明了控制方法的有效性和延误分析理论的精确性。更多还原

【Abstract】 The traffic flow in signalized intersection always embodies multi-state characteristic in terms oftime and space, influenced by a variety of factors, such as driving skill, traffic composition, signalcontrol, traffic management, traffic environment, etc. The existing signal control methods, whichhaven’t fully taken the distribution characteristic of traffic flow in intersection into account，causeintersections becoming the main areas of delay increasing in the road network and affect theoperation benefit of entire network. In order to solve the problem that the existing control methodcould not adapt to the intersection signal control in multi-state traffic conditions, the characteristic ofmulti-state traffic flow has been analyzed, traffic flow prediction method and capacity model ofintersection have been established, supported by the National Natural Science Foundation projects“Study on simulation and control strategy of driving cycle in mixed traffic environment”(No.70771036) and “The simulation of traffic flow segment and management measures empiricalresearch in plane intersection”(No.71071044), Furthermore, signal control methods and delayanalysis theory have been proposed. The research has been basically reflected in the followingaspects:(1) By analyzing the multi-state characteristic of traffic flow in terms of time and space, theinfluences have been analyzed from three aspects: driver, vehicle and highway. Decision methods ofmulti-state traffic flow and relatively steady-state one have been put forward, and the thresholdvalues of headway and traffic flow have been calibrated.(2) Combined with the feature that denseness function could approximate any distributioncharacteristic data with arbitrary degree of accuracy, dense infinite mixture Gamma distributionfunction has been introduced to establish the headway fitting method based on the maximumlikelihood algorithm, and then the accuracy has been verified by the investigation results.(3) Based on gap acceptance theory, the capacity model of unsignalized intersection has beenestablished, from analyzing the probability that minor road traffic flow passes through the headspaces of major road one, by the infinite mixture Erlang denseness distribution function.(4) Based on wavelet analysis theory, the combination forecasting model of short-timemulti-state traffic flow has been established with RBF neural network and Markov chains predictionmethods.(5) According to the variation feature of multi-state traffic flow in the daily traffic statisticalseries, denoise process of daily traffic series signal has been implemented via wavelet analyzing. Combined with the signal timing theory by F.Webster-B.Cobber, the control time division method ofmulti-state signalized intersection has been proposed.(6) According to the distribution characteristic of multi-state traffic flow and time-spaceresources in intersections, control methods have been established in three types: timing control,inductive control and predictive control.①For unparallel multi-state traffic flow intersections, the adjoint phase timing control methodhas been set up after adjusting the space resources and using phase optimization technique.②On account of small intersections with fewer entrance lanes and changeable traffic flow, thesynthesis benefit PI function with main traffic permitting in each phase has been established basedon the queuing theory. According to the conflict relationship of each direction in intersection,inductive signal control method with phase self-organization optimizing has been proposed.③With regard to multi-state traffic intersection where there are many entrance lanes,multifunctional lanes should be installed. According to the queuing theory and predicting outcome ofmulti-state traffic flow, the selecting model of multifunctional lanes direction has been presented,which takes the minimum total operational delay as optimization objective. Under the condition thateffective green time could satisfy maximum queue length passing through, every phase has beenrolling optimized. And in this case, the predictive control method on space resources dynamicoptimization in intersection installing multifunctional lanes has been proposed.(7) The denseness function probabilistic statistics of multi-state traffic flow has beenestablished on the foundation of random process theory. Combined with the phase changes ofdifferent strategies, such as timing control, inductive control and predictive control, the signalcontrol delay model of multi-state traffic flow has been proposed.(8) The effectiveness of control strategy of multi-state traffic flow and the accuracy of delayanalysis theory have been confirmed by the simulation using VISSIM and related software.更多还原