基于Multi-Agent的过饱和交叉口群信号控制

发布时间：2018-05-15 23:29

本文选题：Multi-Agent + 交叉口群　；参考：《兰州理工大学》2015年硕士论文

【摘要】：交叉口信号控制是解决交通拥堵的重要控制方式,传统的基于精确数学模型的交通控制方法,在实际应用中由于其计算复杂往往无法达到预期的控制效果。Multi-Agent技术作为人工智能领域中的研究热点,不仅可以独立自主的控制对象,且Agent之间可以通过互相通讯进行协调,从而智能的控制整个区域交通系统。本文对Multi-Agent区域交通控制模型进行探讨,将自适应遗传算法引入到Multi-Agent交通控制系统中,采用模糊控制原理对交叉口群进行动态划分,充分考虑5种影响划分的因子,最后针对过饱和交叉口群进行信号配时优化研究。本文根据交通系统复杂多变的特点,建立了基于Multi-Agent的分布式区域交通控制模型,模型分为交叉口群Agent、区域控制中心Agent、主控制中心Agent三层,各级Agent除了进行同级间的协调外,还可以同上一级通信协调。同时将自适应遗传算法引入到交叉口群Agent中,克服了遗传算法早熟收敛的弊端,通过感知模块,知识库,规则库,通信模块之间的协调合作,最终选择最优的决策方式。仿真试验表明,这种控制方式,优化绿信比,得到最佳配时时间,使得交叉口平均延误时间变短。针对交通道路网中交叉口群的划分问题,采用模糊推理理论对交叉口群进行动态划分。综合考虑了交叉口间距、周期、交通流、交通流影响因子以及排队长度5种影响因子,通过仿真实验获得每种影响因子的影响权重,并将其体现在模糊规则中,最终得到相邻交叉口协调系数表示相邻交叉口的关联程度,选择路网进行仿真,验证划分方法的有效性。最后将改进NSGA算法应用到过饱和交叉口群信号配时方案中,改进NSGA算法在NSGA的基础上增加了精英策略、密度估计策略和快速非支配排序策略,克服了NSGA算法计算耗时大、容易丢失满意解的缺点。选择最大通过数和最小排队长度作为优化目标,通过Matlab仿真实验,验证了该算法在过饱和交叉口群信号配时中的有效性。
[Abstract]:Intersection signal control is an important control method to solve traffic congestion. In practical application, because of the complexity of calculation, the expected control effect can not be achieved. Multi-Agent technology, as a research hotspot in the field of artificial intelligence, not only can control objects independently, but also can coordinate between Agent through mutual communication. So that intelligent control of the entire regional traffic system. In this paper, the Multi-Agent traffic control model is discussed, and the adaptive genetic algorithm is introduced into the Multi-Agent traffic control system. The fuzzy control principle is used to divide the intersection groups dynamically, and five factors affecting the division are fully considered. Finally, the signal timing optimization for supersaturated intersection group is studied. According to the complex and changeable characteristics of traffic system, a distributed regional traffic control model based on Multi-Agent is established in this paper. The model is divided into three layers: intersection group, regional control center and main control center Agent. You can also communicate with the next level of coordination. At the same time, the adaptive genetic algorithm is introduced into the intersection group Agent, which overcomes the disadvantage of premature convergence of genetic algorithm. Through the coordination and cooperation among perception module, knowledge base, rule base and communication module, the optimal decision mode is finally selected. The simulation results show that this control method optimizes the green signal ratio and obtains the best timing time which makes the average delay time of intersection shorter. To solve the problem of intersection group division in traffic road network, fuzzy inference theory is used to divide intersection group dynamically. The influence factors of intersection spacing, period, traffic flow influence factor and queue length are considered synthetically. The influence weight of each influence factor is obtained by simulation experiment, and it is embodied in fuzzy rules. Finally, the coordination coefficient of adjacent intersections is obtained to represent the correlation degree of adjacent intersections, and the road network is selected for simulation to verify the validity of the partition method. Finally, the improved NSGA algorithm is applied to the signal timing scheme of supersaturated intersection group. The improved NSGA algorithm adds elite strategy, density estimation strategy and fast non-dominated sorting strategy on the basis of NSGA, which overcomes the time-consuming computation of NSGA algorithm. It is easy to lose satisfactory solution. The maximum pass number and minimum queue length are chosen as the optimization targets. The effectiveness of the algorithm in signal timing of supersaturated intersection is verified by Matlab simulation.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.54;TP18

【相似文献】