面向服务可靠性的常规公交运营协同控制与智能化决策方法

项目名称： 面向服务可靠性的常规公交运营协同控制与智能化决策方法

项目编号： No.61203162

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： 陈维亚

作者单位： 中南大学

项目金额： 24万元

中文摘要： 优先发展公共交通是关系我国城市和交通发展的重要战略，提高常规公交服务可靠性是保证常规公交在城市公共交通中主体地位的有力手段。作为智能公交管理系统核心之一的线路智能车辆调度方法从始末站的动态调度逐步发展为全线车辆全过程的实时优化与控制，决策方法从人工经验、人机配合发展为系统自适应学习和自主式优化。本项目面向提高常规公交服务可靠性，基于当前智能公交系统所提供的信息环境，研究基于公交服务过程同步仿真优化决策的多车辆协同控制方法。首先界定同步仿真优化决策所需的智能公交信息环境和公交服务可靠性的动态评价内容；考虑现实系统中信息的完备性，构建基于实时完备信息和非完备信息的公交服务过程同步仿真模型；建立基于预防机制的群策略同步仿真优化决策模型；构建多车辆运营过程协同控制模型和算法。研究将综合运用离散系统同步仿真、智能化决策、协同控制等理论与方法，实现公交车辆服务过程的智能化调度决策和协同控制。

中文关键词： 智能公交系统；公交服务可靠性；协同控制；多智能体决策；增强型学习

英文摘要： Giving priority to the development of public transportation is an important strategy for China's urban and transportation development. Improving bus service reliability is a powerful means to ensure the dominant position of urban public transit services in urban transportation system. As the core technology of intelligent public transit management system, a progressive development of intelligent vehicle scheduling and dispatching for a single bus route service is featured by such tendency of from dynamic vehicle dispatching at the terminal stations to real-time optimization and collaborative control of all the vehicles of the full trips; While the methods of decision-making is developed from human experience, to man-machine cooperation and to self-adaptive learning and self-organizing optimization. Aiming at improving urban bus service reliability and using the information conditions of intelligent public transit system, this project proposes a method of multi-vehicle collaborative control based on decision-making by synchronous simulation optimization of bus service process. Firstly, the information conditions needed by synchronous simulation optimization and decision-making in the intelligent bus transit system will be defined and the dynamic evaluation of bus service reliability will be proposed; Considering

英文关键词： Intelligent Transit System；Bus Service Reliabilty；Collaborative Control；Multi-Agent Decision-Making；Reinforcement Learning