A Simulation-based Approach for Constructing All-day Travel Chains from Mobile Phone Data

The purpose of this work is to investigate replacing travel diaries with sets of call detail records (CDRs) as inputs for an agent-oriented traffic simulation. We propose constructing an agent population directly from a CDR dataset and fusing it with link volume counts to reduce spatio-temporal uncertainty and correct for underrepresented traffic segments. The problem of finding a set of travel plans which realizes a set of CDR trajectories and is consistent with a set of link volume counts is rephrased in terms of calibrating a choice model. This enables us to make use of an existing calibration scheme for agent-oriented simulations. We demonstrate our approach by an illustrative scenario with synthetic data.