Data-driven Optimization of Transport Systems

点击翻译按钮获取中文摘要

Advances in Information and Communication Technologies (ICT) alongside advances in vehicle technologies are profoundly transforming the way we transport people and freight. ICT create opportunities for new transport services and they generate tremendous amount of disaggregate data capturing people's mobility preferences and detailed freight movements. Emerging transport services (e.g., shared mobility, e-hailing for people and freight) impact the short and long-term behavior of people and industry. Predicting these behavioral changes and designing transport systems to encourage sustainable development constitutes one of the major challenges of today's society. The use of these emerging transport services results in integrated transport systems (TS) where the historical separations between private and public transport modes as well between people and freight are blurred. While such integrated TSs have a huge potential to improve efficiency and reduce emissions, the design, planning and optimization of the overall system is complex. The proposed Research Program (RP) focuses on this challenging topic. It is important so that central authorities (e.g., public agencies or other stakeholders) can plan infrastructure, define policies and regulations that are aligned with sustainable development. Fundamental in this context is the prediction of demand and the optimization of TS supply taking demand response into account. The expected outcome of this RP is methodologies that can be integrated in data-driven decision-support tools of central authorities to analyze questions that are not possible with the tools of today. More precisely (i) provide quantitative support for policy-making (ii) design and plan sustainable integrated TS: providing the right incentives to encourage sustainable behavior of people and other actors. In this context, several important challenges arise: First, there are different types of users with heterogeneous, sometimes even conflicting, preferences that interact when sharing the capacity of the system. Second, different transport modes for both people and goods are integrated so multiple modes can be used within one trip. While integration has the potential to decrease inefficiencies, the resulting systems are large scale. Third, TSs are dynamic as the state of the system evolves over time. The originality of the proposed RP lies in the scope and complexity (optimizing supply taking a stochastic demand response into account) of the problems we tackle and the innovative methodologies we propose to address them by combining methods from several subfields of computer science and applied mathematics (machine learning, mathematical programming, approximate dynamic programming).