学术报告题 目： Experimental study of day-to-day flow dynamics 地 点： 土木实验四号楼六楼会议室
时 间： 2017年6月18日（周日） 9:00-10:30主讲人：肖峰
摘要：This study focuses on an interesting question that has not yet received sufficient attention in the research community: are the various day-to-day models proposed so far good enough to reflect the real-life situation, and if yes, what are their relative good performances? To answer our question, we conducted a virtual route choice experiment and collected the participants’ day-to-day route choice data, with which the existing day-to-day models are studied from the following aspects. First, the existing path-based day-to-day models in the literature are calibrated. Second, the nonlinear effects of path flows and path time differences on the route switching are investigated. Third, participants’ preferences to different paths, the variation of their sensitivity over time, and their learning behavior, are examined. Fourth, the assumption of “rational behavior adjustment process” is verified. Fifth, the prediction power of various day-to-day models is compared. Finally, various forms of Lyapunov functions used for stability analysis in the literature are researched. In this study, we adopted regression analysis to investigate the existing day-to-day models based on a virtual route choice experiment. We identified some issues (such as heterogeneity and autocorrelation) to which attentions should be paid in such qualitative research; we also observed some interesting properties of the day-to-day model that can be considered in both empirical and theoretical research. The heterogeneity and autocorrelation found in regression could be a hint for missing features in a day-to-day model. Modifying the model form can help tackle the heterogeneity and autocorrelation; however, it may deprive the key mathematical properties of a day-to-day model including steady states and stability. Alternatively, we adopted the WLS method that keeps the model forms intact. Various assumptions on participants’ route choice behaviors are examined.
毕业于清华大学，获得土木工程学士学位和交通工程硕士学位，并于2007年在香港科技大学获得交通工程博士学位，曾任美国加州大学戴维斯分校博士后，英国Maunsell咨询公司香港总部交通规划师。博士后期间多次获得加州大学戴维斯分校可持续发展交通研究中心基金，以及美国加州政府交通部创新项目基金。现任西南财经大学工商管理学院教授，大数据研究中心副主任，博士生导师。主持国家自然科学基金委优秀青年基金等多项国家和省部级课题，四川省青年百人计划获得者。研究方向包括人工智能算法与交通数据挖掘、交通网络建模和优化、道路拥挤收费、多模式综合交通系统研究、智能交通系统等。在交通及管理科学研究领域国际著名期刊和会议如Transportation Science，Transportation Research Part A、B、C，ISTTT等发表SCI、EI检索论文20余篇。
大连理工大学建设工程学部 地址:中国辽宁省大连市甘井子区凌工路2号 邮编:116024 联系电话:0411-84708500 联系信箱:email@example.com