Vacancies

Postdoc in Design Generation and Evaluation

The global urbanization and growth of cities lead to a significant increase in mobility and logistics while the amount of space available is scarce. IT technology and mobility services in cities are evolving rapidly, and there are many public and private stakeholders involved in the design, and management of mobility and transport systems. To keep areas accessible and liveable there is a pressing need for advanced design methodologies that are capable of dealing with increasingly complex design problems such as location choice of mobility hubs, the number of shared vehicles offered at hubs and the number of parking spaces on different locations in a city.

Current urban mobility planning and design makes use of predictive digital twins based on classical traffic and transport models, in which a group of stakeholders interacts with a visual dashboard. In this approach, the stakeholders propose several designs, and transport models are used to assess the expected impacts of those designs on the modal-split, network performance, accessibility and other indicators. A substantial disadvantage of this approach is that only a limited number of designs can be considered. Another disadvantage is that only a small group of stakeholders decides about the designs, creating the risk that they unconsciously introduce bias towards specific population groups, areas and/or types of solutions.

The Mobility DesAIgn Lab focuses on novel ‘prescriptive’ digital twins that are based on AI and optimisation techniques and studies the ethical, legal, societal, and economic implications of using these prescriptive digital twins for urban mobility and spatial system design and management, enabling stakeholders to start using these prescriptive digital twins responsibly.

In this role you will be responsible for the generation and evaluation of the designs and management strategies that can be explored in the digital twin. The process of generating designs is steered by AI techniques, while the evaluation of designs is performed using existing transport models. These models simulate behaviour of travellers to compute a variety of indicators, such as modality choices, travel times, and accessibility measures, for different population groups. The indicators can be used to explore, compare and eventually choose designs in the use cases. You will collaborate closely with a fellow postdoc as well as with researchers from TNO, the university of Leiden, and with several public and private partners.

More information can be found on the careers website of TU Delft.

Postdoc in Reinforcement Learning from Human Feedback applied to Mobility System Design

Human-centred AI techniques, such as Reinforcement Learning from Human Feedback (RLHF), hold great potential for supporting design methodologies in urban planning. This PostDoc position focuses on AI-based recommender systems to guide stakeholders in choosing a mobility system design. Specifically, the research builds on RLHF techniques, considering algorithmic challenges such as multiple objectives, the robustness of designs and the transparency of the design recommendation process.

The position is offered in the context of the Mobility DesAIgn Lab, a project that investigates how digital twins can be used responsibly for design issues in the field of mobility, transport and logistics in cities. The PostDoc will be embedded in the Sequential Decision Making group within TU Delft’s Intelligent Systems department, with close collaborations with project partners TNO and TU Delft’s Transport & Planning department.

More information can be found on the careers website of TU Delft.