【作者】 杨祖元；

【导师】 黄席樾；

【作者基本信息】 重庆大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 城市交通拥堵的加剧导致了车辆延误增加、交通事故频发、环境恶化等诸多问题,已成为世界各国城市发展共同面临的社会问题。智能交通系统是解决交通拥堵问题的重要途径,城市交通信号控制是当前控制领域和交通工程领域的研究热点之一,也是智能交通系统的重要组成部分。由于城市交通系统本身的强非线性、随机性、时变性特点,以精确数学模型为基础的传统控制方法效果并不理想,因此人工智能的方法愈来愈受到人们的重视。本文从控制科学的角度出发,融合了模糊逻辑和遗传算法、人工神经网络等人工智能理论,对城市交通信号系统的智能控制展开了深入研究,以缓解城市交通拥堵带来的危害。论文取得了以下主要成果:①提出了基于交通强度的单交叉口交通信号控制控制算法该算法突破了以红灯相位和绿灯相位上下游交叉口之间车辆数为模糊控制器输入变量的局限性,还综合了绿灯期间的流率比、红灯持续时间,因此更客观地反映了交叉口各相位车流通行需求的紧急程度。仿真结果表明,本文控制方法比仅考虑车辆排队长度的控制方法更有效地减小了交叉口车辆平均延误。②提出了一种相邻交叉口交通信号协调控制算法该算法以中间路段车流量作为协调变量,综合交叉口各相位的交通流状况,确定候选绿灯相位,通过模糊推理确定是否延长当前绿灯相位。仿真结果表明该算法能有效减小交叉口的车辆平均延误。③提出了一种干道交通信号两级协调控制算法该算法以相邻交叉口中间路段上的车流量为协调变量,对各相位绿灯延长时间进行调整。该算法突破了绿波带控制方法需要统一各交叉口信号周期长度的局限性。仿真结果表明,该方法比绿波带控制方法更有效减小交叉口的车辆平均延误。④提出了一种区域交通信号分布式协调控制算法该算法根据分布式控制原则,不需要统一区域内交叉口的信号周期,而是根据相邻交叉口中间路段的车流量和本交叉口的交通流情况进行协调控制,提高了控制的灵活性和实时性。⑤提出了基于遗传神经网络的短时交通流预测方法该预测方法以均方误差作为预测性能评价指标,同时考虑了交通流的时间相关性和空间相关性,并利用遗传算法优化网络的初始权值和阈值。预测结果表明了该方法的有效性。总之,本文在有机结合模糊逻辑、遗传算法和神经网络等人工智能理论的基础上,对城市交通信号控制进行了深入的研究,提高了控制的灵活性和实时性,有效减小了车辆平均延误。更多还原

【Abstract】 The aggravation of urban traffic congestion becomes a social issue for the development of cities over the world, as it causes increased vehicle delay, frequent traffic accidents, and worsening environment. ITS (Intelligent Transportation System) is one importation solution to traffic congestion. Intelligent control of urban traffic signal is not only a hot issue of both automatic control and traffic engineering, but an important element of ITS. As a complex system of heavy nonlinearity, randomness, time-variability, urban traffic system is beyond the capability of traditional control method which is based on precise mathematical model, so increased importance is attached to artificial-intelligence method.In this thesis, in order to relieve traffic congestion, in-depth study on intelligent control of urban traffic signal is carried out based on the fusion of fuzzy logic, genetic algorithm, and artificial neural network.The main achievements in this thesis include:①A traffic intensity-based control algorithm for isolated intersections is presented.This algorithm not only takes into account and the number of vehicles between the two detectors, but also the flow ratio, red signal duration. It breaks through the limitation that the inputs of fuzzy controller only include the number of vehicles between the two detectors. It objectively reflects the urgency degree of traffic flow in each phase. Simulation result shows that this algorithm is more effective in reducing the average delay.②A coordination control algorithm is presented for two adjacent intersections.In this algorithm, traffic volume in the link of two adjacent intersections is the coordination variable. The next green phase is determined according to the traffic condition of each red phase. Fuzzy inference is carried out to determine whether to stop the current green phase. Simulation result shows that this algorithm is effective in reducing the average delay.③A hierarchical coordination control algorithm for arterials is put forward.According to the concept of decomposition-coordination, traffic volume in the link of two adjacent intersections is the coordination variable. The coordination element modifies the extension of green signal according to coordination variable. This algorithm breaks through the limitation that all intersections are controlled with a same cycle which is applied in green-wave method. It improves the adaptability and real-time operation. Simulation result demonstrates that this algorithm is more effective in reducing vehicle delay.④A distributed coordination control algorithm is presented for area traffic.According to the distributed control, it’s unnecessary to determine a common signal cycle for all intersections. The coordination control is carried out according to the traffic volume in the link of two adjacent intersections and the traffic condition of each intersection. This algorithm improves the adaptability and real-time operation.⑤A control-orientated prediction algorithm for short-term traffic flow is presented.This algorithm takes into account the temporal and spatial relevancy of traffic flow, with mean-square error being the performance index. Genetic algorithm is applied to optimizing the initial weight and threshold. The prediction result indicates that this method outperforms the traditional BP neural networks in precision.In short, an in-depth study for urban traffic signal control is carried out by sufficiently synthesized fuzzy logic, GA and artificial neural network. The control adaptability and real-time property are both improved and average delay is reduced.更多还原