Traffic Forecasting and Autonomous Vehicles
Luis Willumsen, Kineo Mobility Analytics, Serbjeet Kohli, Steer Davies Gleave
Autonomous vehicles will affect many aspects of travel from the next decade and our forecasting efforts must consider their most likely impact on aspects like congestion, trip induction, incidents and values of time.
We are often asked to forecast traffic (and sometimes concession revenues) or travel demand 20 or 30 years into the future. Delivering a reasonably accurate forecast is an important component to estimate the feasibility of a particular project or the viability of a comprehensive transport plan. In most cities, the benefits of the plan or project will depend on an estimate of travel times on alternative modes (where relevant), routes and modal shares. We have been doing this for many years (with variable accuracy) under a broad assumption of “business as usual”; that is travel trends, capacities and mode preferences (defined by the parameters in our models) will be retained into the future.
Over the last couple of years we have learnt that this is unlikely to be the case. Several trends that were weak in the past are becoming increasingly significant: “peak car”, very variable fuel prices, internet shopping, distant presence. The most disruptive of these future events is likely to be the introduction of autonomous vehicles or self driving cars.
Some of the world’s largest commercial organisations, such as Google, Apple , BMW, Volvo and Uber, are sdeveloping self-driving vehicles and the technologies, systems and processes associated with them. Government agencies, innovation centres and other institutions have also taken interest in the matter and are helping this development. In addition, a number of countries (the UK, the Netherlands and certain US states among the foremost group) are leading the way to establish and clarify the regulatory environment in which self-driving vehicles can be trialled and deployed.
The paper will focus on how to deal with this particular disruption in our models and traffic forecasts as it is no longer possible to ignore it. The paper will try to summarise what we know about the possible impact of autonomous vehicles and also what we cannot yet know at this stage. It is expected that autonomous vehicles will have a number of impacts including:
• Reduction of accidents and therefore reduced number of incidents and improvements on travel time reliability;
• Increases in capacity with no investment in infrastructure; the impact will be different depending on the road type and context;
• Trip induction or dilution depending on whether they are rented or owned and how they are used;
• Disruption to the concept of value of in-vehicle time that may well become almost zero as the time will no longer be wasted.
The paper will attempt to summarise what seems to be the state of knowledge of these effects and where there is significant uncertainty about their magnitute. Of course, the size of the effect will depend on the degree of penetration of autonomous vehicles in the fleet and there is also uncertainty about this rate.
The authors have tried to grapple with some of these issues in our traffic forecasts; however, we recognise that we have only dealt with the most obvious of these impacts and that much more needs to be done.
The paper will try to propose some guidelines to deal with these issues on our models and it is expected that this will trigger some interesting discussions
Association for European Transport