【作者】 臧利林；

【导师】 贾磊； Gang-Len Chang；

【作者基本信息】 山东大学 ， 控制理论与控制工程， 2007， 博士

【摘要】 交通与人类的生活息息相关。在大多数大中型城市,随着社会经济水平的逐步提高和汽车工业的不断发展,私家车的数量迅速增多,越来越多的城市道路交通拥挤造成出行者的行程时间更长,这将导致较低的社会生产力,增加噪音和环境污染。怎样有效地缓解交通拥挤、提高交通系统的效率成为交通工程师和研究者首要关注的问题。传统上,拥挤问题通过对交通路网的物理扩张来处理,但这种方法不仅造价昂贵、破坏城市环境,而且其缓解交通拥挤的效果也十分有限,因而该方法已经很少被考虑。另一种缓解拥挤的方法是提高现有交通资源的利用率,执行合理的交通控制策略。随着计算机信息处理能力的不断强大和电子技术、先进控制理论的不断发展,后者越来越受到世界各国交通部门和研究者的重视。毫无疑问,交叉口信号控制作为交通管理的主要组成单元,在整个城市交通中扮演着重要的角色,如何建立一个精确有效的智能信号控制系统对城市交通管理具有重要的意义。本文以城市交通信号为研究对象,主要包括两个方面的研究内容:一是基于混沌时间序列的城市交通流短时预测模型与方法研究,该方法将为后续的交通信号控制提供必要的交通流数据;二是城市交通信号优化控制算法的研究,主要从单交叉口信号动态控制和区域交通信号协调控制两部分进行深入的研究。本论文在广泛查阅各类文献的基础上,分别从上述两个方面开展工作,符合当前智能交通控制理论的研究方向,是当前城市智能交通研究的热点与难点。论文完成的主要任务及创新点如下:1、短时交通流的预测是城市交通诱导和信号控制的基础。预测方法的优劣对交通流预测结果产生直接的影响,从而影响交通信号控制的精度。在以往的研究中,交通流的中长期预测已取得了较好的研究结果,但是,关于短时交通流量实时预测(通常是预测未来5-15分钟的交通流量)的研究还未能取得令人满意的结果。本文第三章将混沌理论应用到短时交通流预测,首先分析了交通流的混沌特性,并从短时交通流量序列的关联维数和最大Lyapunov指数证明了交通流具有混沌特性,然后建立了基于混沌时间序列的短时交通流预测模型,采用加权局域预测方法,获得了令人满意的预测效果。该方法能够为交通信号控制提供比较准确的交通流信息。2、模糊逻辑在单交叉口信号控制中的应用主要有两个方面,一方面是对于当前相位绿灯延时模糊控制,另一方面是交通信号相位顺序的优化控制。存在的交通信号相位模糊控制方法主要有跳跃式和自动产生式。前者是根据当前交通流数据从预设的信号相位中选择下一相位,若预设的下一相位排队车辆较少,则跳过该相位选择车辆排队长度较长的相位运行,从而改变相位顺序;后者没有预设的相位,根据交通流的需求,实时地决定下一相位。这些相位优化方法容易误导行人和驾驶员,比较适用于行人和非机动车较少的交叉口信号控制。本文第四章提出了一种基于相位优化的单交叉口信号模糊控制算法。考虑行人和非机动车的因素,通过最小相位时间保证行人和非机动车安全通过交叉口。改进了存在的信号相位模糊控制方法,在保持原相位顺序不变的前提下,根据实时交通流数据,通过绿灯延时和相位的优化,使系统动态地响应交通流的变化。该方法能够有效地提高绿灯时间利用率,从而减少车辆平均延误时间。3、当涉及到多个交叉口信号控制时,合理的交通信号协调控制能够有效地提高区域交通的效率。存在的控制方法主要是建立在区域所有交叉口处于同一交通流状态的前提下,基于共同周期进行协调控制研究。在实际中,由于整个区域的交叉口交通流量不可能总是保持相同或非常相近,那么,这些方法必然会在某些交叉口因非最优周期而导致更多的延误,因此有必要根据区域交通的地理特征和实时交通流数据开发一种动态的子区域划分方法,并基于子区域的划分进行交叉口信号协调控制。本文第五章检索了已有的交通子区域划分方法,在前人的基础上,以车辆平均延误最小为优化目标,提出了基于子区域划分的交通信号优化模型。在模型中,采用HCM2000延误估计模型,并嵌入了著名的Robertson车队离散模型,从而更准确地对信号定时方案的性能进行评估。4、遗传算法是模仿自然界生物进化机制发展起来的一种随机全局搜索优化方法,被广泛地应用到最优化问题的求解。为了获得区域交通信号优化模型的最优解,本文第六章在Visual C++环境下开发了基于遗传算法的区域交通信号优化方法,并采用自适应策略确定遗传算法的交叉概率和变异概率。5、微观交通仿真程序CORSIM作为一个无偏的工具用来评估同一交通情境下的不同的信号控制方案。本文第六章分别采用干线信号和多干线交通网络信号控制为对象,将遗传算法与TRANSYT-7F优化后获得的交叉口信号周期、绿信比和相位差分别导入到微观交通仿真系统CORSIM中,运行CORSIM程序,通过比较仿真结果来证明提出模型的有效性。结果表明本文提出的区域交通信号优化方法优于TRANSYT-7F。更多还原

【Abstract】 Traffic is very close to people’s life. In most of large and secondary cities, the number of private vehicles is increasing rapidly with the stepwise enhancement of social economy and the continuous development of automotive industry. So more and more traffic congestion has appeared and caused more traveling time for passengers, this will lead to inferior social productivity, increasing noises and environmental pollution. Therefore, it is an urgent problem for traffic engineers and researchers to study how to alleviate traffic jams and improve the efficiency of transportation system effectively. Traditionally, physical expansion of traffic network is a main means to solve the congestion and jams. But this method will cause expensive cost and destroy urban environment, furthermore, the effect is not very obvious. So it is considered rarely now. The other way to alleviate traffic jams is to enhance the efficiency of traffic facilities and implement rational traffic control strategies. With the incessant improvement of data processing capability of computer and the ceaseless development of electronic techniques and advanced control theory, more and more emphasis is placed on the second method by traffic departments and researchers in the world. Without question, signal control at intersections is a primary component of traffic management and it plays an important role in the whole urban traffic. Thus, it possesses very important significance for urban traffic management to develop an accurate and effective signal control system.The research object is urban traffic signal and two aspects are included in this dissertation. The first aspect is study on the prediction model and method of short-time traffic flow based on chaos time series, which will provide necessary traffic data for signal controllers. The second one is study on the optimization control algorithms of urban traffic signal, which include the signal control algorithm at an isolated intersection and the signal coordinated control method at reginal intersections. Referring to correlative literatures widely, the dissertation finished two aspects of work above mentioned. Its contents accord with the current research direction of traffic control theory and are hot and difficult topics in the field of urban intelligent traffic. In the dissertation, the main contents and innovation points are as follows.(1) The prediction on short-time traffic flow is the foundation of urban traffic guidance and signal control. Good or bad prediction methods have direct impacts on the prediction results of traffic flow. Hence it will affect the precision of traffic signal control. In the past study, middle-term and long-term prediction on traffic flow have made better study outcomes, but the study about the prediction on short-time traffic flow(usually predicting traffic flow in the coming five to fifteen minutes) has not gained satisfactory results. The Chapter 3 in this dissertation does research on how to apply chaos theory to the prediction on short-time traffic flow. Firstly, it analyzes the chaos characteristics of traffic flow, and proves that chaos characteristics exist in the time series of short-time traffic flow based on the correlation dimension and maximum of Lyapunov exponents. Then a prediction model of short-time traffic flow is built based on chaos time series. By using the adding-weight local-region method, the prediction results are satisfactory. Therefore, the proposed method can provide more accurate traffic information for signal controllers.(2) Fuzzy logic has been developed to control traffic signal at an isolated intersection, which is used mainly for two aspects. One is fuzzy control of green extensive time for current phase and the other is optimization control of phases’ sequence. Existing fuzzy control methods of phases’ sequence include the skipping method and the auto-generating method. The former is to select next phase from preset phases according to current traffic flow. If there are few vehicles in the preset phase, then the system will skip it and select another phase where there are more queuing vehicles. The latter is to auto-generate the phase without preset phases based on the real-time traffic flow. These methods to optimize the phases’ sequence tend to mislead passengers and drivers. They are suitable for those intersections with few passengers and non-motors. A fuzzy signal control algorithm for an isolated intersection is proposed based on the optimization of phases’ sequence in the Chapter 4. Considering of passengers and non-motors, this algorithm can guarantee passengers and non-motor to pass the intersection safely by using minimal phase time. The method improves the existing traffic signal fuzzy control algorithms. Based on the real-time traffic data, this method responds to the variety of traffic flow dynamically by optimizing green extension and phases’ sequence. The method can effectively enhance the utilization ratio of phase time and reduce average delay of all vehicles.(3) For traffic signal control at many intersections, rational coordinated control between neighboring intersections can improve the efficiency of traffic network. Most existing control methods were developed under the assumption that all intersections of the whole area were in the same traffic state and traffic signals were coordinated based on a common cycle length. In the fact, traffic volumes of all intersections are not always same or very close, so these methods will lead to more delay at some intersections with non-optimal cycles. Therefore, it is necessary to develop a method to partition the whole traffic area into several sub-areas dynamically according to the geographical characters of traffic network and real-time traffic data at intersections and optimize the signal control parameters based on the partition of sub-areas. Chapter 5 in the dissertation firstly searched existing traffic sub-area partition methods, under the foundation of previous research, and then proposed a signal optimization control model to minimize average delay of vehicles for traffic network based on partition of sub-areas. In the proposed model, HCM2000 delay model and Robertson’s dispersion formula are used, hereby it is more accurate to estimate the performances of signal control plans.(4) Genetic algorithm is a stochastic and global search technique coming from an analogy with biology in nature, which was used in computing to find exact or approximate solutions of optimization problems widely. In order to gain the optimal solution of area traffic signal optimization model, Chapter 6 developed an optimization method based on genetic algorithm by using the Visual C++ software. Self-adapting strategies were used to calculate the crossover probability and the mutation probability.(5) Microscopic traffic simulation system CORSIM, as a without bias tool is used to estimate different signal control methods in the same traffic scenarios. Chapter 6 takes an arterial and a multi-arterial traffic network as simulation objects. Input signal control parameters, such as cycles, splits and offsets from genetic algorithm program and TRANSYT-7F into CORSIM respectively, then run CORSIM program and export the simulation results. The validity of proposed model can be verified by comparing the simulation results. It proved that the proposed optimization method for area traffic signals is prior to TRANSYT-7F.更多还原