项目名称: 网络化环境下交通信号系统的按需控制方法研究
项目编号: No.61203079
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 自动化学科
项目作者: 李镇江
作者单位: 中国科学院自动化研究所
项目金额: 24万元
中文摘要: 交通环境的复杂多变使得交通信号系统的控制需求呈现多样性和不确定性,本项目利用代理控制技术将集中式控制算法转变为面向具体任务的简单分散式控制代理,并根据运行条件和需求的不同,通过网络移动到交通信号控制装置和系统上运行,实现网络化环境下交通信号系统的按需控制。首先,从交通环境和需求入手,面向海量交通数据处理,采用混合高斯模型拟合交通流基本参数,基于MapReduce架构,将EM算法并行化,对模型参数进行学习求解,得到局部模型,并通过模型合并得到交通流预测模型,实现对交通流量的在线预测;然后,以交通流预测结果为基础,建立基于内容的推荐和协作过滤推荐相结合的混合推荐系统,实现控制代理的优化配置和自动推荐;最后,采用硬件在环内的交通仿真方法,搭建实验系统进行控制实验,验证该方法的可行性和有效性。通过本项目的研究,拟建立一种适应联网分布式特点、具备开放智能化特性的交通信号系统的按需控制方法。
中文关键词: 按需控制;交通信号控制;代理控制;移动代理;
英文摘要: Control demands of traffic signal systems are diversified and uncertain because of the complexity and variety of traffic environments. In this project, we use agent-based control technology to convert integrated control algorithms into simple distributed control agents according to specific tasks. These control agents can move through network to traffic signal control devices and systems according to different operation requirements and control demands. In this way, we can realize control on demand of traffic signal systems in a networked environment. Firstly, we start with traffic environments and demands and use Gaussian mixture models (GMM) to represent the traffic flow data for mass traffic data process. Based on MapReduce framework, we implement parallelization of EM algorithm and use it to learn parameters of local models. After getting local models, we find out a reasonable global traffic flow forecasting model for one intersection and realize online traffic flow forecasting. Secondly, based on traffic flow forecasting results, we build a kind of hybrid recommendation system which is consisted by one content-based recommendation system and one collaborative filtering recommendation system. Through this system, optimizing deployment and automatic recommendation of control agents are achieved. Lastly, ado
英文关键词: control on demand;traffic signal control;agent-based control;mobile agent;