Expectations are high, the rhetoric is abundant, the dollar signs astronomical and in all of that hype transport modellers are expected to forecast what the impacts of connected and autonomous vehicles (CAVs) may be, for transport strategy purposes but also for the return on investment of large and costly infrastructure projects. Is there anything sensible that we can contribute to the debate, any reasonable and defensible numbers that we can produce that will add structure to decision-making? I think so, of course I think so, and we can do a lot with our existing tools, even though they were developed for a quite different transport environment and have been estimated based on past trends and behaviours.
The key is to decompose this forecasting problem into three elements:
· Modelling the direct impacts of CAVs on road traffic and transport system operations
· Modelling the indirect impacts of CAVs on wider travel behaviour, transport financing and transport strategy
· Modelling the non-transport impacts on urban structure, socio-economic development, and in different energy and environmental futures for example.
Apart from different levels of uncertainty and complexity, each of these will also have different time dimensions, from very short to very long term. Let’s explore each in turn.
To model the direct impacts of CAVs we probably need to develop new tools that reflect the operational characteristics of the new technology (and I am aware that the traditional transport modelling software houses are busy doing so). For example, connectivity and automation are expected to lead to reduced road space requirements per vehicle, faster acceleration, accessing and providing huge amounts of improved real-time information that can be used in operations. Existing micro-simulation models are being refined to reflect these direct effects. But shared vehicle ownership and use requires more fundamental changes to our models that were developed for a system in which cars are used and controlled by single families. Logistics-based approaches such as ptv-MAAS modeller consider people as packages to reflect ride-sharing. Interesting agent-based model developments are now able to find a balance between the number of vehicles in the system, the wait time for users after requesting a ride and the amount of empty miles travelled by CAVs looking for their next pick-up. But this doesn’t mean that there is no longer a use for our existing tools.
To model the indirect impacts of CAVs we can and we should use our existing tools better, to improve our understanding of what might happen when CAVs become part of the transport system. The free-to-use travel time when cars drive themselves, combined with the expected overall speed-up of traffic that this enables, would almost certainly reduce the value of time of CAV travel and increase overall demand as well as CAV demand. Existing strategic transport demand models are eminently suited to estimate what these likely effects might be, firmly embedded in economic theory and rational, utility-maximising decision-making. The overall impacts on revenue, from reduced fuel tax income from electric vehicles, from decreasing patronage of more traditional public transport, or from falling parking fees because of automation, can also be derived from traditional models that calculate origins, choice of modes and destinations and the vehicle miles travelled in-between. Finally, any intervention to counteract such negatives, such as (GPS-enabled?) road user charging, can easily be tested and its effectiveness quantified using traditional, strategic transport models that are widely available and well-understood.
To model the non-transport impacts of CAVs being part of the overall transport system in a Mobility as a Service world, it will be necessary to develop narratives about potential futures in which autonomous vehicles operate. Such futures could partly be created by the new technology itself but partly they will develop exogenously – think of the sharing economy or the Internet of Things. These futures should encompass at least alternative energy and environmental futures, alternative social and demographic structures and alternative urban forms, emerging as a result of and enabling new, intelligent transport options. Given the uncertainty around these futures, I don’t think it is necessary or even worthwhile developing whole new models for these, although it would be interesting to investigate whether land use transport interaction approaches will come into their own at last. For me, such scenarios can be developed and tested by developing and agreeing appropriate inputs to traditional, existing model systems. In the end, it will all be about story-telling and sensitivity testing.
Many of the questions around the advent of autonomous vehicles will not require modelling at all; they are fundamental questions around the role of transport in society, and how we as transport planners can use the emergence of CAVs and other technologies intelligently to achieve a fairer and more sustainable transport system. Some have said that, despite its connectivity, autonomy and probably electricity, a self-driving car is still a car, and I understand their concerns. I have heard others say that the distribution of impacts of CAVs will most likely not be progressive - many early adopters are already well-catered for by the existing transport system. CAVs cannibalising public transport may increase transport poverty; some of the benefits for the most vulnerable travellers (such as the elderly) may only be realised in a fully automated system in a distant future. Perhaps the greatest challenge in this arena are the unknown unknowns. Models, and particularly complex models, are not the answer.
Despite that and returning to the title: there are big questions associated with the advent of autonomous vehicles that can benefit from model support. Rather than losing ourselves in more advanced, more complex, less understandable models and forecasting approaches, we as transport modellers can contribute to the debate, cut through some of the hype, illustrate the uncertainties and interdependencies, using existing tools intelligently and be willing to help develop the story supported by explicit and well-explained assumptions. It is the assumptions rather than the models that matter – and it is understanding and challenging these assumptions in scenarios and sensitivity tests that should help shape how CAVs are introduced into the overall transport system to achieve a fairer and more sustainable transport system, also considering goals around social inclusion, health and well-being, improvements in air quality and safety.
Tom van Vuren is a Divisional Director in Mott MacDonald and a Visiting Professor at the Institute for Transport Studies at the University of Leeds