【作者】 黄辉先；

【导师】 史忠科；

【作者基本信息】 西北工业大学 ， 控制理论与控制工程， 2000， 博士

【摘要】 交通信号控制是管理运输网络的一种基本手段，人们普遍意识到交通信号控制系统的作用还未得到完全发挥。随着智能运输系统（Intelligent TransportationSystem，简记为ITS）的发展，交通信号控制的研究远未完善，交通信号控制依然是最重要的研究和发展项目之一。到目前为止，已经研究出许多交通信号控制模型。尽管这些模型差别很大，但主要可分成两大类。一类主要针对交通流未饱和的情形进行研究；另一类主要是针对拥挤与过饱和交通流条件进行研究，其中排队一直存在且不能完全清除，排队形成和离散的动力学研究变得极为重要。本文针对交通信号优化控制问题进行了智能优化的理论分析和应用方法的研究，得出了一些有益的结论，解决了一些实际问题： 1.本文首先提出了基于智能优化的交通信号控制原理，主要介绍了遗传算法优化方法和混沌优化方法的基本思想。通过对城市单交叉路口的交通流分析，建立了单交叉口以车辆排队长为目标函数的四相位交通信号方案实时控制模型。采用遗传算法优化方法对该模型中相位绿灯持续时间进行优化仿真试验，通过与传统优化方法的仿真试验进行比较，证明这种遗传算法优化方法的优越性。遗传算法搜索适应于单交叉路口配时优化，可获得最优性能指标。 2.本文提出了改善遗传算法搜索方法的一种混合搜索算法，并从理论上证明了这种混合算法的收敛性。这种混合算法把梯度法的区间最优搜索以及遗传算法的遍历性结合起来，既利用了遗传算法全局优化的特点，也利用了梯度法的最速下降的特点，从而保证系统既避免了遗传算法要遍历几乎所有的状态、搜索时间过长的缺点，又避免了梯度法的易于陷入局部最优的缺点，提高了优化效率。仿真试验表明这种新的混合搜索方法的有效性。 3.本文建立了单交叉路口以车辆平均延误为性能指标的实时交通模型，提出了基于混沌优化的周期与绿信比综合优化方法。在一定的绿信比下，交叉口的延误指标随周期变化具有单峰性。对于周期的优化采用了变步长单向搜索方法；对于绿信比的优化，采用了混沌搜索方法；整个优化过程采用了周期、绿信比顺序优化的方法。仿真试验表明本文方法优于文献[12]及OSCADY的方法。此外，还进行了多个时间段的性能指标优化方法的研究。 4.本文针对网络细胞模型建立了基于细胞传播模型（CTM）的交通动力学模型，推导了交通信号控制的一般公式，提出了基于混沌优化的周期与绿信比综合优化方法。以具有多个交叉口的主干道为对象，针对轻度、中度、重度三种不同的交通需求进行仿真试验，仿真结果表明本文方法优于文献[5]的方 2 西北工业大学博士学位论文一法。获得了较好的性能指标。 5．本文最后讨论了智能交通信号机的研制，参考目前国内外交通信号机 的功能特点，介绍了所研制的智能交通信号机的结构原理及程序设计实现方法， 并着重讨论了交通信号机的感应控制方式的控制原理、联网方式下的通讯协议 以及联网控制的实现方法。”最后，简耍介绍了新的交通信号机的特点。更多还原

【Abstract】 Traffic signal is an essential element to manage the transportation network. Nevertheless, it is widely accepted that the benefits of traffic control signal systems are not being fully realized. Along with the movement of Intelligent Transportation System (ITS), traffic signal control remains one of the most heavily funded research and developmeiii. iL~ins. i ne research on traffic signal control is by no means complete. A number of models of the traffic signal control have been developed in the past. Despite substantial differences among them, they can be roughly classified by two approaches. One is developed mainly for unsaturated traffic flow. The other is developed mainly for congested or over-saturated traffic conditions where queues exist in the whole period and can not be cleared. The dynamics of queue formation and dissipation becomes very important. In this dissertation, the principle of the traffic signal control based on intelligent optimization, the basic optimal theories and the methods based on Genetic Algorithm as well as Chaos theory are proposed. Through the study and the analysis of traffic streams in an intersection, the real-time control model of four-phase signal plan according to the objective function of vehicle queue length in an intersection is established. The simulation tests of phase green duration in the model are conducted using Genetic Algorithm method. The simulation results show that the optimal method of Genetic Algorithm is superior to tradition methods, and it is suitable for determining the green split in a single intersection. Based on Genetic Algorithm, a hybrid optimal method is provided. It combines the advantage of global optimization in Genetic Algorithm with the advantage of the fastest dropping in Gradient method. It can also ensure to avoid much time in Genetic Algorithm and to escape from local optimization of Gradient method. The simulation results indicate that this new method increases the search efficiency. With regard to performance index of the average vehicle delay in an intersection, the traffic stream model and a new optimal method based on Chaos theory according to cycle length and green split are developed. Under a given green split, the performance index of average delay in an intersection varies with the cycle in a convex curve. During optimizing, the search step is changed according to the cycle length, an optimal method based on Chaos theory is used according to green split, and a sequential optimal method is used according to cycle length and green split. The simulation results show that the method is superior to those in OSCADY and reference 12. In addition, the optimization of performance index for a series of period is also conducted. A traffic stream dynamics based on Cell Transmission Model (CTM) is established. An optimal method based on Chaos theory according to cycle length and green split for CTM of a network is developed. The simulation tests are conducted in three cases of light, moderate and heavy traffic streams with arterial road of several 4 intersections. The simulation results show that this method is superior to the method of reference 5. The better performance index can be obtained in the new method. After the function characteristics of domestic and foreign traffic signal control machines being known, a new intelligent machine of traffic signal control is studied and manufactured. The constitution of the machine of traffic signal control and its method of programming are更多还原