Abstract
In order to acquire information about human mobility, person trip surveys have been conducted. Additionally, disaggregate behavioral models such as activity-based approach have been developed. The activity-based simulation can represent details of entire daily activities. On the other hand, observation data can grasp dynamic traffic condition. This study develops a possible method to integrate activity-based simulation and observation data, so that human mobilities according to current situation can be estimated. The proposed method is based on state space model. In the model, locations of people are represented as state vector, and zone population is defined as observation vector. The activity-based simulation PCATS is adopted as system model, and similarity between state and observation vectors is set as observation model. Since the model is non-linear and non-Gaussian, particle filter is utilized for the filtering. The proposed method is applied to the residents in center of Tokyo. Through the applications, the significance of the proposed method is confirmed.
