基于分层递阶与多智能体理论的城市交通协调控制研究
文献类型：学位论文
| 作者 | 刘小明 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2004-06-15 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 王飞跃 |
| 关键词 | 分层递阶 多智能体系统 人工交通系统 XML 车辆代理模型 面向Agent 协调控制 Coordinated hierarchical system Multi-agent system Manual traffic system XML Vehicle agent module Agent-oriented Coordination c |
| 其他题名 | Study of City Traffic Coordination Control based on Coordinated Hierarchical and Multi-Agent Theory |
| 学位专业 | 控制理论与控制工程 |
| 中文摘要 | 随着现代城市交通量的增长和智能交通研究的发展，在对城市交通控制的 研究中，以整个城市或者区域作为研究对象的区域控制方法越来越受到研究人员的重视。 分层递阶控制做到了集中控制与分散控制的合理结合、互补长短，是复杂规模系统控制的首选腔制方式。而城市交通系统是一个典型的非线性、时变、 复杂的大系统，因此采用分层递阶控制理论来解决城市交通协调控制是一个较好的研究方法。但值得提出的是分层递阶结构形式受协调器的性能、工作状况、通讯带宽的限制，荠且各局部智能控制器间的协作完全依赖协调器而进行，自主性较差，上层全局协调器可能由于计算量过重而使计算效率变低，不能适应实时智能交通控制的需要。多Agent系统是Agent的集合，是当今人工智能的前沿学科，是分布式人工智能研究中的一个重要分支，其目标是将大的复杂系统建造成小的、彼此相互通信及协调的、易于管理的系统。多Agent系统中每个Agent都是一个具有相同问题求解方法的自治系统，能利用局部信息进行自主规划，并能通过规划推理解决局部冲突实现协作，从而完成与自身相关的局部目标。因此将多Agent技术与分层递阶结构相结合正是解决城市交通控制的一个可行的策略。 交通系统除了其多学科、跨领域、规模巨大等复杂特性之外，还具有实验试行成本极高甚至无法进行的特征。国际上已经有多种大规模的交通规划、管理、优化和控制计算机仿真模拟系统，但随着计算机容量和速度的提高，随着并行、分布和网格计算结构和方法的不断普及，目前的交通仿真模拟思想的局限性也同益明显，与计算机所能提供的性能脱节，没有最大可能地发挥计算机的作用。因此，本文的前一部分首先对人工交通系统的实现作了一定的研究，在此基础上，在后一部分由点到面逐步地对城市交通协调控制策略进行了研究，并在实现的人工交通系统上对控制策略进行了仿真研究。 本文的具体研究内容如下： 1)提出了基于Agent的车辆代理模型 在本文所提出的车辆代理模型中，一方面应用智能体理论，首先将跟车和换道现象通过模糊规则的方式分别获得加速度；另一方面将二者的加速度值通过对不同类型的驾驶员运用偏好系数相乘之后，再进行对比的方式，获得驾驶员最终的决定：跟车还是换道。在模型建立的后期，考虑到了天气因素的影响，对最终得出的加速度又进行了加权。最后，作者应用Agent技术对无灯控路口的车辆 |
| 英文摘要 | With the city traffic flow volume increasing and the study of intelligent traffic system developing, the area control methods whose studied object are the whole city or an area traffic are recognized more and more by people. Coordinated hierarchical control system has the advantages of both centralized control system and dispersed control system. With that reason, the coordinated hierarchical control system has become the chief control way for the complex system. The city traffic system is a typical complex system, so the coordinated hierarchical control method is an appropriate approach for resolving city traffic coordination control. But coordinated hierarchical configuration has its disadvantage. Coordinated hierarchical configuration is limited by the capability and communication bandwidth of cooperation unit, and not able to adapt to real time intelligent traffic control. Multi-agent system is the collection of agent, and is the front subject of artificial intelligence. The object of Multi-age, into system is to decompose the large and complex system into small, corresponding and easily managing systems. Every agent in multi-agent system is autonomic system which has the same question solving method. Every agent can realize correspond and solve conflict by reasoning and layout. So we can speak that it is a good and feasible strategy to combining coordinated hierarchical method and multi-agent theorv. Traffic system has another character. The experiment cost of traffic system is high and sometime even it is difficult to get experiment. There is much simulation software now, but these simulation software can't utilize computer resource fully. In this paper, the manual traffic system is realized in the first part. Then city traffic control strategies are studied gradually, which are simulated in manual traffic system. The material study contents are as follows: Firstly, the vehicle agent model is put forward in this paper. With multi-agent theory, at first, the acceleration of car-followed and lane-changed is got by fuzzy rule, then the acceleration is amended by the action coefficient of driver action. With the finally result, the action of car-followed or lane-changed is decided. In the module, the climate factors are considered, then the acceleration can be amended again. At last, with the agent theory, the driving action of vehicle agent in intersection which isn't controlled by lamp is studied. All the work can provide elements for latter work. Secondly, the manual traffic software is developed. On the basis of vehicle agent module, the design and realize of several key part of manual traffic system is described, and the manual traffic software is realized. Thirdly, the intersection control method on the basis of agent is studied. At first, intersection agent module and phase agent module are built. Then the mutual between different agents is studied. A |
| 语种 | 中文 |
| 其他标识符 | 828 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/5825]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 刘小明. 基于分层递阶与多智能体理论的城市交通协调控制研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2004. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。