The Impact of Artificial Intelligence on Transport Modelling

Shelly Webster, WSP Senior Principal Transport Modeller, sat down with renowned North American modelling specialist Rick Donnelly to discuss strategic transport modelling.

Specifically, they discussed how it is being impacted by Artificial Intelligence (AI) and behaviour changes in transit patterns, and what the future could hold for projects and the profession.

 

Can you explain what strategic modelling is?

SW: When we talk about strategic modelling we’re looking at a higher-level type of modelling. It’s typically considering the bigger picture and high-level policy rather than the nitty gritty detail at the operational level.

 

RD: It’s where all the policies and projects come together. The strategic model allows you to look at all their interactions and play ‘what if’ games about the future. For example, what are people doing and who are they doing it with? Down at the operational level you think about vehicle flows on a typically smaller network and their impact on network performance and reliability. In strategic models you can think about networks, technologies, and behaviours that don’t exist. What will we be doing in the future? Working differently? Acting differently? 
How are these models currently used?

 

SW: A strategic model is typically used to test long range plans and to do megaproject analyses and pricing analyses, amongst other things. It is also used for business case appraisal for large transport infrastructure projects as a starting point to justify whether to go ahead and develop it further. Some may have more advanced and more sophisticated models than others depending on their purpose, but also on data and resources available. There is a perception that models can predict the future but at the end of the day it depends on what you put into the model – it’s testing a range of possible outcomes based on a certain set of assumptions. You must be aware of that and not just treat it as a black box that can somehow predict the future with certainty.

 

RD: It’s a subtle difference. We predict what would happen if the future was to occur right now. Our current models predict what those impacts would be, based on people’s current choices, patterns, tastes, and preferences. We know they will be different in the future, but can only guess at how.

 

What is influencing change in the industry?

RD: Automation coupled with AI is rapidly changing our economies in ways never seen before. Some people believe that this shift will be no different than the industrial or information revolutions. Jobs didn’t disappear then, but rather morphed in line with those revolutions. What is really different this time is that there was no AI in the previous revolutions that could do things so much better than people can, and that will really change things. The same will hold true in transport planning and modelling. AI will provide data analyses and predictions at an order of magnitude that is faster than we can today. A quarter of the people involved today will be required, and they’ll be leveraging AI rather than competing with it.  Where Shelly can still make a difference is with her ability to use AI to inform project development, interact with the public, and make the business case for a project. AI probably won’t be able to do that for a long time.

 

img-strategic-modelling-rick-donnelly-quote 

 

SW: It’s not just the changes in technology that are affecting industry, it’s the behavioural changes that occur as a result. For example, a few years ago working from home or remotely was more the exception than the norm. People now don’t have to come into the office certain days of the week. Our transit patterns are changing – traditionally our models were focused on people getting into cars or onto public transport to go to work. Today, we have a lot more travel choices and substitutes. We can order online and have purchases delivered to our door rather than having to travel to the shop, or play video games and socially interact from the comfort of your lounge room. We don’t go to the cinema as much as we once did, preferring to stream HD video on large-screen home cinemas. All these little shifts trigger changes in travel patterns. The challenge is forecasting them and adapting to that quickly while still creating a bigger picture view in our models.

 

Where is the future of strategic modelling headed?

RD: Big data is changing the landscape significantly and we have access to far more information every year now than we ever had in the rest of our careers combined. It makes models easier and faster to build because we can base them on these big data points now.

 

SW: The increasing availability of data also assists in making the process of creating models more transparent. Traditionally a lot of assumptions and adjustments were made in the models which are not necessarily transparent or clear to stakeholders. The availability of more data makes the process of building and validating transport models as well as the understanding of travel behaviour and patterns easier than before. Stakeholders including the private sector are expecting a lot more from these models and this can add another layer of formal, robust risk and uncertainty analysis. People are making important decisions based on the outputs as well as investing significant amounts of money. Private investors are asking the right questions and are quite pragmatic about how we should grow and address transport challenges.

 

RD: When I started out in this profession we used models to look only at big megaprojects and long-range transport plans. Strategic modelling in those realms examined major policies and infrastructure investments. Now we’re shifting towards operations and maintenance to maximise the value of it. But that’s asking traditional models questions to inform analyses at levels of spatial, temporal, and behavioural levels of precision and accuracy beyond what they were designed for. Now Shelly and I ask ourselves, ‘How do we interface traditional strategic models with operational models better suited for such analyses? That’s the holy grail of transport planning models’.

 

What role will modelling play in creating future cities?

SW: We won’t so much be looking at how we operate within the city, rather we will be looking at what our vision is for the city. I don’t think I’ve ever worked anywhere with the level of transport infrastructure building that is going on in Sydney at the moment and all at the same time. Motorways are being planned and built, new train lines, as well as light rail. It is a huge amount of investment and when you’re testing so many things you have to think about how they will all work together not only in the current environment but also in the future. That’s a lot more difficult than thinking about single projects or policies in isolation.

 

RD: Shelly is right – we try to reduce the complexity of cities down to manageable levels, but we never start with a future vision of Sydney and work backwards from it. Walt Disney built Disneyland using ‘Imagineering’, working backwards from his vision to put pieces in place to incrementally bring this project to life. We are good at engineering the pieces, but don’t see the bigger distant future they fit into (or not). Politicians ask me to tell them what the future is going to look like, but in reality their collective decisions and social and technological progress will shape them in ways we can only imagine. They think we know what the future will hold, but instead we’re moving towards painting several alternate futures and what how that will shape cities and how people use transport systems to interact within them.

 

Shelly is a Principal Transport Modeller with WSP. She has over 20 years of background in the transportation planning and modelling field with key experience in strategic modelling, roads master planning, policy and strategy.

 

Rick Donnelly is recognised as a leader in travel modelling and simulation with more than 30 years of experience. Based in the US for the past ten years, he has led the design of state-wide travel forecasting models in Arizona, Maryland and North Carolina. His most recent focus has been on activity-based modelling of person and freight flows, along with agent-based simulation of defence logistics systems. Mr Donnelly is currently leading the design of a bilevel integrated economic-transport model for Ontario (Canada) as well as high speed rail forecasts for the Texas Department of Transportation. A co-chair of the Transportation Research Board’s Committee on Travel Forecasting Resources and a member of the Association for European Transport’s freight and logistics committee, Mr Donnelly has a thorough understanding of the global transport market.

 

 To stay abreast of our latest news, publications, videos and posts, please follow us on LinkedIn.