Computational Methods for Sustainable Mobility - Interpretation and Prediction of Tracking Data using Graphs and Machine Learning

Abstract

Many of today’s urgent challenges, such as greenhouse gas emissions and climate change, air quality and health, or traffic and congestion, are closely linked to the movement of people and goods. A major cause of these problems is fossil fuel based individual transport, making individual mobility behavior change a requirement for solving them. Computational methods based on data collected using Information and Communication Technologies and Location-Based Services can play key a role in supporting sustainable mobility. The focus of this dissertation is the development and application of computational methods to support sustainable individual mobility in four different ways. Gathering empirical evidence on how Mobility as a Service affects the mobility behavior of individuals, developing a framework for more generalizable methods to preprocess tracking data, developing methods to support the modal shift of individuals toward more sustainable modes, and developing methods to support the sustainability of personal vehicles. A core contribution of this dissertation is the formalization of a graph-based repre- sentation of individual mobility called location graph. This representation is compact, privacy-preserving, and can be created based on a wide range of different datasets, which simplifies the development of transferable computational methods. Based on location graphs, we developed machine learning methods for identifying user groups with similar mobility behavior and for imputing missing activity labels. These methods were applied to problems related to the management of Mobility as a Service offers, a core concept to enable modal shift for individuals. To support the sustainability of personal vehicles this thesis includes work on traffic prediction, a critical component of an intelligent traffic management system. Fur- thermore, it also includes a study showing that owners of battery electric vehicles can meet most of their charging demand using power generated from their own rooftop photovoltaic system. The latter has great potential to further reduce the greenhouse gas emissions of personal vehicles.