Modelling Tomorrow's World: four perspectives

In thinking about tomorrow’s world, it is important to separate out the modelling required in operating such a future, recognising quite different levels of automation and connectivity, data and information flows, pricing and control mechanisms; as opposed to the modelling required to forecast what such futures may look like, and supporting decision-making today.  In that context it may be useful to separate out four quadrants in the modelling vs time range.

Modelling Today's World with Today's Models

There are sceptics out there that feel that today’s models are no longer good enough, even for dealing with today’s problems and today’s solutions. I disagree. They aren’t broken, but not necessarily used very intelligently. Their underpinning theory and maths is thorough, unfortunately leading to complex and time consuming applications. 

But the questions are changing, new outputs and insights are nowadays required by decision-makers and stakeholders. The need to provide more detail on local impacts increases the demands on models, probably making them even less transparent, more data hungry and slower to run. 

I agree there are improvements we should try to make in today’s models. Speed and ease of interpretation would be high on my agenda.

Modelling Tomorrow's World with Today's Models

I guess we don’t have much choice for now.  Although tomorrow’s transport environment may be quite different, that doesn’t automatically imply that the travel choice processes in our models need to change dramatically. I don’t think it unreasonable to expect people to still make rational decisions, perhaps even more rational in an information-richer environment. 

The uncertainty of future available travel options, currently unimaginable technological enhancements, already observed changes in travel behaviour trends, different urban forms and ways of living can be explored with existing models, but in scenario settings. 

Can tomorrow’s models reflect the already changing travel trends in today’s world and do these trends invalidate today’s tools?

There are exciting opportunities, too, to use existing models and model components that we have largely ignored. Data-rich futures with almost perfectly informed travellers, instantaneous system reactions to incidents and central control over pricing and routing may make socially optimised travel patterns more achievable. The price of anarchy, the difference between selfish user equilibrium and collaborative system optimum, will become more than a construct. Modelling Tomorrow’s World with today’s models should help identify where investment in new data and models is most warranted.

Modelling Tomorrow's World with Tomorrow's Models

This leads to the real question for Modelling Tomorrow’s World: what aspects of tomorrow’s world cannot be model well with today's models and what beneficial developments are happening already in modelling data, and visualisation? What can we learn from other, associated disciplines? That’s why I am excited about this event and look forward to the presentations and demonstrations. And to me it is not immediately obvious, nor is it inevitable that tomorrow’s models must be very different from today’s. Operationally, absolutely. But for forecasting, not so sure…

These are some of the challenges that I see:

• How do we adapt today’s models to represent modes that do not yet exist, even if the mode choice model’s structure may remain largely unaltered? 
• How can we reflect the impact of technological innovations such as full and continuous connectivity? What can we learn from the changing trends in trip making and travel behaviour amongst younger people, and will it invalidate most of our current tools? 
• Can we afford the representation of the additional dynamics of highly adaptive systems in models that are already slow? 
• Is it necessary to develop detailed models of the interactions between connected and autonomous vehicles on roads and at junctions, or are well-reasoned approximations enough?

There is another challenge: modelling and through modelling supporting the transition period from today to tomorrow; quoting the SBD/here paper again: “The longer-term outlook for autonomous cars is paradoxically easier to forecast than the medium-term”.  It may be possible and exciting to reflect 2040 or 2060 scenarios in which society and the transport system have changed dramatically and have entered a utopian or dystopian end state, if that is possible. However, that change will not be instantaneous and the pathway to such futures will require planning, monitoring and management. For example, technological changes may outpace the regulatory response; or societal acceptance may not emerge; or a black swan event may have a catastrophic impact. I haven’t seen enough analytical effort in this space.

Modelling Today's World with Tomorrow's Models

Some of the enhancements that tomorrow’s models promise, will happen anyway, irrespective of what tomorrow’s world looks like. What I am interested in is: what are the quick wins? How can we as transport modellers start using these improvements to make today’s models faster, better, more helpful to our end users? 

This may sound easy, but using tomorrow’s models may be as disruptive as tomorrow’s transport system itself. Today’s toolkit is essentially controlled by, in guardianship with software vendors. Academic developments enter commercial software very slowly, and are generally limited by the ability of software houses to fit it in their programs and programmes. Hardware used tends to be standard and is probably well below the current state of the possible. End users have little power over development agendas and budgets. That is why I expect that tomorrow’s models will take a while to become mainstream; but that is no reason not to start exploring now.

I worry about focusing on developing models to operate tomorrow’s world, technology driven, rather than on model development and, most importantly, application to forecast what tomorrow’s world might look like. For example, I expect that the case for HS2 or CrossRail2 will be affected less by the emergence of autonomous vehicles than by societal, economic and even climate changes. These can be reflected in standard inputs to and parameters in today’s models which therefore should not be so easily discarded when exploring Tomorrow’s World. 

Tom van Vuren is a Divisional Director at Mott MacDonald, a Visiting Professor at the University of Leeds and Chairman of Modelling World, which has been organised by Landor since 2006. Modelling World Middle East will be held in Dubai on 9 February 2017; Modelling World will be held on 8 June at the Kia Oval in London.