Autonomous vehicles (AVs) are a rapidly evolving mode of travel, and one of several emerging megatrends in transport planning that will profoundly change travel behavior and patterns over the coming decades. There are several AV definitions in the literature and press. We follow the SAE definitions, which include six levels of automation ranging from none at all to fully connected and autonomous vehicles, with the latter including a high level of system-level control of the transport network. The discussion that follows assumes a fully automated future (SAE level 5), although some of the concepts discussed may be applicable to lower levels of automation.

Simulating the effect of AVs will require fundamentally rethinking how most transport planning models are structured and applied. There are some effects that can be represented using traditional models, but the majority of likely behavioral changes will affect all traveler preferences, choices, and costs. Many of these will be second or higher-order effects that will be realized through feedback loops or replanning. Discussions about the policy and planning context, major behavioral changes, and modeling frameworks are described in the sections linked to below. Case studies of several evolving models are also presented. The reader should be able to grasp the appropriate modeling approaches in light of their requirements.

Autonomous vehicles: Planning and policy context

A wide variety of questions about AV futures are being posed by policymakers. A number of policy papers have been published on the topic, but they describe the problem space much better than specifying forecasting approaches. A number of policy issues and strategies identified in NCHRP Report 456 are summarized on the [planning and policy context] page. Other issues identified in other reports have been added to the discussion.

Autonomous vehicles: Behavioral considerations

The availability of AVs, either as privately-owned, shared, or hired vehicles, will open a range of travel possibilities not available today, and change many aspects of current modes. Time spent traveling in AVs will be spent on other activities, reducing the disutility and cost of travel. Sharing an AV or forgoing private auto ownership will likely become commonplace, further changing mobility patterns and creating new travel opportunities for young, elderly, and disabled persons. These and other [behavioral considerations] will form the basis for model specifications that include the full range of AV impacts on travel behavior and network performance.

Autonomous vehicles: Modeling frameworks

The likely effects of AVs will influence all aspects of travel behavior encompassed by travel forecasting models. This will necessitate a revolution in [modeling frameworks] suitable for measuring their extent and impacts. While the range of policy issues and strategies is long and varied there are few AV futures that are unlikely to influence all aspects of modeled behavior, to include long-term location choices, short-term travel choices (to include daily activity patterns and activity scheduling), and network assignment. While some can be accommodated within the traditional four-step modeling paradigm most will be better addressed using visioning and activity-based travel modeling frameworks.

Further reading

A number of policymakers and planners have written white papers and reports describing how AVs might affect land use and travel behavior over the coming decades. Many of these were reviewed in the preparation of this and related pages in TFResource. Some cover the policy, planning, and legal aspects of AVs, to include:

+ Lauren Issac
+ Rachel Skinner
+ Johanna Zmud
+ Townsend

There is considerably less definitive guidance on how to include AVs in travel forecasting models. One emerging resource is the on-going NCHRP Project <?>, which seeks to provide guidance on how trip-based, activity-based, and visioning models might evolve in order to incorporate travel changes due to AVs. A large number of academic papers and research reports have been recently published or in progress, but most of them can be quickly implemented and capable of addressing the issues described in the pages linked above.


This category has only the following subcategory.