【作者】 张立东；

【导师】 贾磊；

【作者基本信息】 山东大学 ， 检测技术与自动化装置， 2012， 博士

【摘要】 交通与人类文明的发展息息相关,人类自诞生便与交通结缘,远古时代徒步迁徙至今日上天下海,遨游太空,无处不有交通的身影。矛盾总是对立存在,交通的发展促进人类文明进步的同时,也带来一系列的负面问题,如能源浪费、废气排放、噪声污染等,尤其在当前小汽车急剧发展的中国,交通拥堵及其带来的相关社会问题,已经成为政府和老百姓都为之“皱眉”的话题。在交通拥堵的诸多表现形式中,交通溢流无疑是较为严重的一种。所谓交通溢流是指由于道路规划或渠化、交通信号配时等不利因素的影响,导致某一路段某一流向,在特定时间段内累积排队等待通行的交通流队列长度大于路段长度,且排队长度蔓延至下一路段,因其类似容器中的流体超过了容器的容积而产生的溢出现象,故称其为交通溢流。交通溢流危害巨大,在其发生时,若不加以控制,溢流会像“传染病毒”般逐渐蔓延,由单个路段扩散到多个相关路段,乃至整个城市路网,最终路网车辆在交叉口处互锁而导致大规模的交通拥堵,使整个城市的交通处于瘫痪状态,城市基本机能丧失,因此,开展关于交通溢流的机理、原因,以及控制方法的研究具现实意义。仿真是验证所提理论算法是否正确与优越的较好方式之一,欲搭建交通流仿真平台,首当建立合适的交通流模型。在详尽分析研究前人研究成果,尤其是微观跟驰模型的基础上,考虑到现实中驾驶员类型多种多样,性格脾气、驾驶习惯各异,由此引入驾驶员灵敏度系数的概率分布的思想,提出随机微观交通流跟驰模型；在OVM模型中,无论是在开放性和周期性边界条件下,基于直线运动的交通流跟驰模型已有大量的研究,而相对于现实中大量存在的弯道道路,报道较少,于是提出弯道交通流跟驰模型,引入圆周运动中的角速度,最大速度等思想,并分析得到了模型的稳定条件；研究城市交通流,交通信号灯的影响和作用是不可回避的重要因素,基于此提出信号灯作用下的微观跟驰模型,分别针对队列头车和跟驰车对信号灯的敏感度不同而建模,仿真分析验证了结果的正确性。实现交通溢流的控制,当先掌握其机理与成因,并在此基础上提出合理有效的识别算法,先识别,后控制,识别是控制的基础,因此接着重点解决交通溢流形成机理、成因及识别的问题。从关联路口的交通流量关系、信号设置关系的角度,以及交叉口延误模型三个因素两种思路情况下,探讨了溢流的形成机理及原因。在分析路段瞬态最大交通密度的基础上,挖掘传统模型中的交通溢流区域,并通过仿真实验的手段,探讨了溢流情况下路段的速度、密度、速度-密度关系以及交通流波动现象。交通溢流的识别是一项带有较强主观性的认知过程,而模糊理论在处理此类问题较传统方法更有优势,于是提出了基于模糊理论的交通溢流识别算法,建立了模糊推理器,仿真验证了其正确性。交通溢流最常见的形式即为单路段交通溢流,从路网物理结构上而言,单路段溢流也是该种拥堵形式的基础单元,相比于单路口信号控制,溢流的控制涉及溢流的消散和相邻路口多个支路交通流的延误最小等多项复合指标,因此更为复杂,而控制过程中涉及众多的经验知识和交通常识,于是以人工智能思想为指导,提出溢流智能控制器,该控制器由溢流相位差模糊推理器、相位相序设置专家系统、交通流神经网络预测器以及相位时间的模糊推理器四部分组成,能够较为全面的实现“溢流识别—相位差设置—相序设置—交通流预测—相位时间推理—执行控制—效果评价”的流程,在自主开发的仿真平台上进行仿真,并与强制控制进行仿真比较,结果表明,文中所提算法相比于强制控制方式在解决交通溢流控制方面性能更优。目前,包括Paramics、VISSIM在内的国内外流行微观交通仿真软件虽提供有二次开发接口,但在建立自主交通流模型和智能控制算法方面,还存在不足,于是在团队前期研发的交通流微观仿真系统的基础上,针对交通溢流这一具体模拟问题进行了升级改造,阐述了平台的架构、重要仿真实体及其相互关系、车辆控制逻辑和信号控制逻辑及其相互关系,以及交通溢流的仿真设置和应用实例问题,开发具有自主知识产权的仿真平台。最后对全文的研究创新点和不足之处进行了总结,并就进一步的研究工作进行了展望。更多还原

【Abstract】 Traffic has close relationship with the development of human beings. Traffic occurred as long as the birth of human beings, when they walked to find the proper living situation at ancicent time, till now we fly in the sky, ship in the sea, even cruised in atmos, traffic accompanies us everywhere in our life. Just as controdict always exists, traffic not only brings us the convenient living conditions, but causes a series of social problems, such as energy cost, exhaust emission, noise pollution, especially in nowadays China, many problems caused by traffic jam have become the frown topic of both government and general people.Traffic spillover is one the most serious kind of typical traffic jam. When the cumulative traffic flow queue becomes longer than the length of some direction in some link within a period of time because of the improper traffic plan or signal timing, and it is similar with the spillover of fliud in a container, we call this phenomenon traffic spillover. It casues great harm on the normal traffic order, if we do not take measuer in time, it will broadcast to other roads just like infectious virus when more and more roads occur traffic spillover, the cars in the network will be interlocked at signalized cross, and the whole city’s traffic turns paralyzed, the city’s function loses. It is clear that to carry out study on the mechanism, causes and control method of traffic spillover is meaningful and has great practical significance.Computer simulation is one the best way to verify whether the throries and algorithms we studied right and advanced. The key factor of one traffic flow simulation platform is traffic flow model.On the detailed study of former research paper, especiall the microscopic car following model, we propose random distribution traffic flow following model on the consideration of diverse driver’s type, which takes the driver’s sensitivity factor as probability distribution. In optimal velocity model, much study has been done on the straight line movement about open and close boundary condition, while we discover that in real life, curved road is also common, so we model the curved road following movement and analyzed its stability condition, the simulation result proved the model’s feasibility. Traffic signal is widely used in urban traffic network, its impact on traffic flow is an important factor in traffic flow model, we proposed microscopic car following model with signal influence, and build leading car and following car model, simulation shows our model is in accordance with regular model.To control traffic spillover, we should grasp its mechanism and causes of formation in first, then propose the reasonable and effective recognition algorithm. Recognition is before control, and identification is the base of control, so these are the key problems of our study. On consideration of correlated traffic volume at adjacent crosses, signal timing plans, as well as the intersection delay model, we discussed the formation mechanism and causes of traffic overflow. On analysis of traditional traffic flow model, the transient maximum traffic density in a link is mined, and through the simulation experiment, the relationship among velocity, density, speed vs. density and traffic flow wave phenomenon were studied on the overflow condition sections. As we know, overflow identification is a job with strong subjective cognitive processes, and fuzzy theory in dealing with such problems have more advantages than traditional methods, so we presents a fuzzy theory based traffic overflow recognition algorithm, establishes a fuzzy inference machine, and verify its correctness with simulation.Single traffic overflow is the most common form, from the view of network physical structure, single section overflow is alsothe base unit of this kind of jam form. Compared with single intersection signal control, overflow control relates to overflow much more factors than traditional control algorithm, such as spillover dissipation and the minimum traffic flow delays of adjacent intersections those two composite indexes, therefore it is more complex. Meanwhile, the control process involves numerous artificial empirical and traffic knowledge, so the artificial intelligence was thought as the main instruction, we proposed the overflow intelligent controller. The controller is composed by the spillover relief phase offset fuzzy inference machine, traffic phase number and sequence setting expert system, traffic overflow neural network predictor, and phase time fuzzy reasoning machine four parts. It can implement "overflow recognition, phase offset setting, phase time reasoning, traffic flow prediction,sequence setting,control executive and control effect evaluation logical process. On our self-developed traffic flow simulation platform, many simulations were carried out respectively, and compared with forced control method, the results show that our algorithm has better performance in solving traffic overflow than forced control and can better deal with traffic spillover problem.At present, thouth current microscopic traffic simulation software, i.e. Paramics and VISSIM, provide second development interface, is still inadequate in the study of develop independent traffic flow model and intelligent control algorithm. So we upgrade our own traffic flow simulation system based on this specific traffic overflow problem. It expounds the platform architecture, important simulation entities and their interrelations, vehicle control logic and signal control logic and its relationship to each other, as well as traffic overflow simulation set and an example of its application. The aim is to develop a kind of independent intellectual property rights simulation platform.Finally, the full text of the final research innovations and shortcomings are summarized, and the further research direction is prospected.更多还原