A Model Analysis of Travel Behavior Changes by Advanced Traveler Information Systems and Their Implications to Transportation Planning

本研究考察了旅游信息提供对微观层面（个体出行行为）和宏观层面（都市圈交通网络）的影响。首先，我们建立了一种方法，定量地评估了考虑交易成本的旅行者出行前信息对旅行行为变化的影响。将交易成本定义为出行行为改变前后所观察到的期望效用之差，并采用离散选择模型对交易成本进行估计，该模型分别表示有出行时间信息和无出行时间信息时的出行方式选择行为。然后，我们开发了一个离散选择模型，该模型结合了非补偿决策规则和属性切断，以明确考虑个人的信息处理能力。将该模型应用于在途旅客信息条件下的出行方式选择行为，结果表明，该模型在可重复性和可预测性上都优于假设效用最大化的现有模型。在此基础上，利用均衡分配模型研究了中京都市圈先进出行信息系统对交通网络需求的影响。提出了两种不同的放宽驾驶员完全信息假设的方法。从实际路线需求不是在旅行时间水平上均衡，而是在效用水平上均衡的观点出发，我们首先提出了一种修正线路性能函数的方法，以纳入旅行时间和成本以外的许多不可观察因素。该方法经实证应用后，拟合优度显著提高，尤其是在非峰期。然后，我们开发了一种考虑网络感知异质性的均衡分配方法。更具体地说，根据司机对网络配置和出行时间的认知水平将其划分为不同的类别，并使用不同的原则进行分配。通过改变各类别的构成比例，应用所提出的模型，结果表明，当约20%的驾驶员选择信息不完全的路线时，拟合优度最高。最后，对提供最短路径信息的先进出行者信息系统的用户效益进行了评估。少

This research investigated the impact of travel information provision on micro level, that is, individual's travel behavior, and macro level, transportation network of metropolitan area. Firstly, we developed a methodology to quantitatively evaluate the impact of traveler's pre-trip information considering transaction cost for travel behavior changes. The transaction cost was defined as the difference of expected utilities which are observed before and after travel behavior changes, and was estimated using a discrete choice model which represents mode choice behavior with and without the information on travel time respectively. We, then, developed a discrete choice model that incorporates non-compensatory decision making rules and attribute cutoff to explicitly consider the individual's information-processing capability. The model was empirically applied to the travel mode choice behavior under on-route traveler's information, and the result showed that the proposed model was superior … More in both reproducibility and prediction to the existing model in which an utility maximizing decision maker is assumed.Next, we investigated the impact of advanced traveler information system on transportation network demand of Chukyo metropolitan area using equilibrium assignment model. Two different approaches were proposed to relax the assumption of driver's perfect information. From the viewpoint that the actual route demand is equilibrate not in travel time level but in utility level, we first developed a methodology to amend link performance function in order to incorporate many unobservable factors other than travel time and cost. This method was empirically applied and the goodness-of-fit, especially in off-peak period, was dramatically increased. We, then, developed an equilibrium assignment method considering heterogeneity of network perception. More specifically, drivers were divided into different categories according to his/her cognitive level on the network configuration and travel time, and were assigned using different principle. The proposed model was applied changing the composition ratio of each categories, and the result showed that the goodness-of fit was the highest in the situation where about 20% of drivers chose his/her route with imperfect information. Finally, we estimated the user benefit of the advanced traveler information system that provides the shortest path information to the drivers. Less