City Modelling Lab at 100 (sprints 🏃♀)
We recently celebrated our 100th sprint in Arup’s City Modelling Lab. In this blog post we take a little stroll down memory lane.
Our story starts in the summer of 2019, with five (more or less) fresh-faced, enthusiastic individuals. We had just spent approximately 2 years working on building digital solutions to problems in transport. Gerry secured R&D funding for us to wrap our heads around MATSim, an open-source software tool that allows you to build transport-flavoured agent-based models.
We left Fred alone while we went on holiday. During our absence, he made a start on many of the software tools we still use today.
At the end of that summer, we started sprinting. We haven’t stopped since. We changed our pace a few times — reducing our sprint duration from 2 weeks to 1 during the pandemic to respond to all of the new behaviours we needed to capture in our models and increasing the duration to a month after the pandemic to cool off.
We made many friends along the way and worked with amazing people. They all have one thing in common — a genuine drive to make transport better for our fellow human beings.
Our first major collaboration was with Transport for London on a city-wide simulation of London. The size and complexity of London was a real challenge for us at the time. Between then and now, we have learned how to create even bigger simulations, including country-wide simulations that push the boundaries even further. We helped New Zealand Ministry of Transport build their own in-house ABM team from scratch, leading to the establishment of a sister City Modelling Lab Down Under. With Transport Infrastructure Ireland, we built a country-wide ABM that supports traditional road pricing implementations but also prompted some of our most exciting new innovations.
In more recent months, we’ve been working very hard on BERTIE with Transport East, a collective of partners in the East Anglia region of the UK. BERTIE is a shining example of all the great features and types of analyses we have implemented in the Lab.
Our knowledge and understanding have grown steadily and significantly since we started 4 and a half years ago. We quickly realised that the quality of our inputs to MATSim can make or break the model, so we dedicated years of study and hard work to capturing and representing human behaviour. The generation of agents and their daily plans alone is a huge field of study in itself. We have studied and relied on Open Street Map (OSM) data for many things. From finding agent home and work locations, building networks to figuring out what type of home an agent might be living in. OSM is a powerful resource to work with, and over the years we have also learnt of its’ limitations.
We have diversified too. Our agent-based transport models now serve as inputs to other analyses and models. We calculate carbon emissions, and analyse their impact on people. We model Electric Vehicles and their impact on carbon and energy systems. We slice and dice our models, with a particular focus on who is impacted and how. We pay special attention to differences in behaviour based on gender and endeavour to embed them into our simulations.
Throughout, we have tried to share as much as we can with our community by open-sourcing our code and writing these blogs. We believe that sharing our work can only lead to positive outcomes. We have a dedicated section of our publication here on medium, where you can find out more about each of our open-source tools.
The past four years have been a whirlwind of software development, transport modelling, expanding outwards into other domains and exploring different sources of funding. We’re very proud of what we have achieved and are looking to the future.
Here’s to the next 100 sprints! 🥂
