With the institute’s drive to make contributions towards smarter mobility, it was fantastic to have some of our students attend the Mathematics Applied in Transport and Traffic Systems (MATTS 2018). This 3-day-workshop was hosted in the impressive Science Centre in Delft University of Technology (TU Delft). The feel of the city of Delft is very unique. Many of the pavements are shared by pedestrians, cyclists and cars alike. Since 2000, the city implemented car free zones in the historic city centre (despite much scepticism and opposition). In 2009, it was evaluated that the changes were received overwhelmingly positive (over 70% more of people appreciated the city more since the changes). It was really fascinating to see how Dutch urban policy has effectively promoted pedestrian and bicycle mobility.

The conference was opened by inviting everyone to participate in a study tracking our movement across Delft - investigating how people explore urban environments for the first time. This PhD project is part of the larger European Horizon 2020 project, Allegro, investigating and understanding slow traffic modes. Led by Professor Serge Hoogendoorn of Smart Urban Mobility at TU Delft, they aim to fill this lack of knowledge. Researchers develop and empirically underpin comprehensive behavioural theories, conceptual models and mathematical models to explain and predict the dynamics of pedestrian, bicycle and mixed flows within an urban context. Special attention is given to the role of ICT in learning and choice behaviour.

We gained a lot of insight into cutting-edge research in transport, from more intelligent intersections and autonomous vehicles to stochastic solutions - the talks were varied and complex… focusing on various macroscopic and microscopic models which gave some inspiration to our smarter mobility project.

Some of the keynotes included:

• Discrete choice and machine learning: two peas in a pod? - Dr. Michael Bierlaire
• Modelling Traffic Dynamics at Large Urban Scale: The Contribution of Network Macroscopic Fundamental Diagrams - Dr. Ludovic Lecierq
• Rebuilding The Traffic Dynamics From Partial Traffic Data Using Information Theory and Hybrid Markov Chain Neural Network Approach - Dr. Alexandros Sopasakis

Noémie Le Carrer presented her research on Optimising Ship Scheduling Subject to Uncertain Sea Levels which was well received, showcasing a snapshot of our work in stochastic optimisation that we undertake here and applying these methods.

The talks were fascinating and really highlighted where improvements can be made in the sector and academic literature. An interesting point from the keynote by Dr. Carolina Osorio from MIT: is that models have grown increasingly complex - but at a cost of time and computing power. Adding complexity should be adequately justified. As technology advances, data-driven approaches are leading the way... next-generation models should stay as simple as possible to compliment the data, and be calibrated efficiently (especially for high-res and physical models) in order to drive significantly improved mobility models.

We both expanded our awareness on how neural networks are trained on driver’s behaviour, predicting traffic jams before they happen, platoon forming to allow cars to pass intersections simultaneously, and much much more. It was fascinating to see how far the reach is of academic research in traffic, and looking for ways to bridge these solutions to reality. I’ve come back brimming with ideas and hopefully with some potential collaborations who are excited to apply their work.