Autonomous Vehicles

Table of Contents

Published: May 14, 2014

This is a writeup of a shallow investigation, a brief look at an area that we use to decide how to prioritize further research.

In a nutshell

  • Why did we decide to look into this area? Widespread adoption of autonomous vehicles may be able to save roughly ten thousand lives and a hundred billion dollars a year in the U.S., making their development a topic of substantial humanitarian significance. Autonomous vehicles will also likely have major impacts on urban planning and the advisability of various infrastructure projects, and we wondered whether the prospect of autonomous vehicles was being appropriately incorporated into long-term planning in these areas.
  • What did we learn? Other actors appear to largely have appropriate incentives to invest in research, development, and advocacy for effective policies around autonomous vehicles. The implications of autonomous vehicles for infrastructure projects and other long-term plans are very uncertain at this point, and are subject to increasing attention from urban planners and others. At this point, we do not see a major asymmetry between the importance of these issues and the funding and attention currently available, though we have very limited confidence in that judgment.

1. Why did we decide to look into this area?

We decided to look into this area for two reasons:

  • Widespread use of safe autonomous vehicles may eventually substantially reduce the rate of traffic accidents, which account for more than 30,000 deaths and an estimated $300 billion of economic costs annually in the U.S.1 Fagnant and Kockelman 2013 estimate that 50% market penetration of autonomous vehicles would save roughly $100 billion/year in economic costs and roughly 10,000 lives per year, though we have not vetted these estimates.2 Although we would guess that for-profit corporations have appropriate incentives to invest in research and development in this area, we wanted to confirm that impression and to investigate whether there might be other activities necessary for eventual widespread deployment of autonomous vehicles that were being overlooked.
  • The prospect of widespread use of autonomous vehicles may have implications for major infrastructure investment projects and other long-term social plans that are not being appropriately incorporated into existing analyses.3

2. Our process

We reviewed a number of reports about the state of autonomous vehicle technology and regulation and spoke with two people with knowledge of the field:

  • Daniel Fagnant, Assistant Professor at University of Utah (starting August 2014) and author of Fagnant and Kockelman 2013.
  • Bryant Walker Smith, Fellow, the Center for Automotive Research at Stanford and the Center for Internet and Society at Stanford Law School; Lecturer, Stanford Law School.

3. What did we learn?

Our impression continues to be that for-profit corporations have appropriate incentives for pursuing most of the relevant research and development necessary for the deployment of autonomous vehicles.4 Although philanthropists may be better suited to supporting work on advocacy than research itself, many actors are already moving to regulate autonomous vehicles and appear to be doing so broadly in line with the public interest.5 The people we spoke with broadly seem to perceive the field of autonomous vehicle research to receive an appropriate amount of funding and attention,6 though Bryant Walker Smith noted a few areas that current actors may be overlooking.7 The long-term implications of autonomous vehicles currently seem to be too uncertain to warrant inclusion in planning to for major infrastructure projects and other long-term social investments, and the relevant professional communities appear to be paying increasing attention to the prospects for autonomous vehicles.8 Reports on autonomous vehicles lay out many areas for further research and policy development, and we expect that significantly more investment will be required in coming years.9 However, in our very limited review of this field, we did not see evidence of a clear asymmetry between current funding and importance, and we expect that more traditional automative research and policy funders will enter over time as these questions become more salient. Based on our current understanding, we see further investigation of other potential focus areas as a more promising use of our resources at the moment.

4. Questions for further investigation

Our research in this area has been very limited, and many important questions remain unanswered by our investigation.

Amongst other topics, further research on this cause might address:

  • How might additional funding for research accelerate development of autonomous vehicles?
  • What kinds of planning decisions should take the prospect of autonomous vehicles into account, and what role should that prospect play in decision-making? How important are the decisions that might be affected?
  • What types of research are best-suited for philanthropic, as opposed to government or for-profit, funding?

5. Sources

DOCUMENT SOURCE
Anderson et al. 2014 Source (archive)
Fagnant and Kockelman 2013 Source (archive)
Notes from a conversation with Bryant Walker Smith on April 11, 2014 Source
Notes from a conversation with Daniel Fagnant on March 18, 2014 Source
Planning for Autonomous Driving Source (archive)
The Impact of Automation on Environmental Impact Statements Source (archive)
Expand Footnotes Collapse Footnotes

1.Fagnant and Kockelman 2013, Table 1, page 4.

2.Fagnant and Kockelman 2013, Table 2, page 8. They also estimate “comprehensive” cost savings of roughly $200 billion/year, which “include[s] indirect economic factors like the statistical value of life and willingness-to-pay to avoid pain and suffering, with values recommended by the USDOT” (page 9).

3.See, e.g., Planning for Autonomous Driving.

4.

“Auto manufacturers are attempting to rapidly develop autonomous vehicles. Manufacturers including General Motors, Nissan, Volvo, Audi, and Mercedes-Benz aim to release vehicles with self-driving capabilities by 2020. These vehicles will probably be autonomous in most situations, such as on freeways and in low-speed areas, but not in intermediate-speed urban situations, where there is more uncertainty. Google aims to have autonomous vehicles on the market by 2017 or 2018, though they would still need licensed drivers at all times. There is some uncertainty among the tech community as to whether Google will be able to accomplish this.” Notes from a conversation with Daniel Fagnant on March 18, 2014

“Technical research in this space is generally well-funded by universities and corporations” Notes from a conversation with Bryant Walker Smith on April 11, 2014

5.

“Powerful interests are already engaging in the regulatory space around autonomous vehicles. For instance, Google successfully lobbied for laws in Nevada and California that expressly permit the operation/testing of autonomous vehicles, and automakers have also been active in shaping legislation and regulation in several states Plaintiff attorneys, insurance companies, and consumer groups (particularly those with privacy interests) are also starting to engage in this space. Privacy concerns are frequently voiced in conversations about regulating vehicle automation and connectivity. There is generally uncertainty about (1) how safe these systems must be and (2) how this level of safety can be demonstrated. Quickly deploying systems might save lives but could also lead to a backlash when a crash does occur. Several states have adopted legislation on autonomous vehicles, and some of those states have moved on to developing regulation. There are differences between states’ regulatory regimes, but so far none of them conflict in important ways. Auto manufacturers and Google are wary about a patchwork of regulatory regimes, and, for the same reason, automotive industry groups have been cautious and particular about the legislation they support. Many groups are working to develop unified regulatory proposals involving, for example, model legislation or coordination among state motor vehicle administrators. The National Highway Traffic Safety Administration (NHTSA) is unlikely to develop regulations in the near term, though some states have expressed interest in federal guidance. In the longer term, NHTSA will play a leading role in defining performance standards.” Notes from a conversation with Bryant Walker Smith on April 11, 2014

“In 2013, the National Highway Traffic Safety Administration (NHTSA) issued a policy statement saying that autonomous vehicles are not yet safe enough for commercial introduction and that more research and technological development is needed. The statement established guidelines for testing and identified areas for more research. The best and most likely scenario for regulation in the long run is one in which autonomous vehicles are regulated at the state level based on a national framework. Alternatively, requirements could be federally mandated or determined state by state. Groups currently involved in regulatory efforts include NHTSA, which is studying regulatory systems in states that are considering commercial licensing or have allowed testing on public roads, and the Transportation Research Board (TRB), which is working to develop national guidelines for regulation, in addition to various state governments. Other groups working in this area include:

  • Google, which advocates extensively on the topic
  • RAND Corporation, which recently published a report on autonomous vehicles
  • KPMG and the Center for Automotive Research (CAR) in Michigan, which collaborated on a report on vehicle connectivity technologies.
  • A number of universities

… Some people, including NHTSA, are concerned that introducing autonomous vehicles before the technology is sufficiently advanced could result in fatal accidents that could set back the successful roll-out of the technology. It is not clear at what point autonomous vehicles will be safe enough to alleviate this concern; some people have suggested that autonomous vehicles should be twice as safe as the average human driver before being sold commercially. There are some privacy concerns with regard to vehicle-to-vehicle and vehicle-to- infrastructure connectivity. It would be important to make this data anonymous so that individual cars could not be tracked. There is no significant ideological or structural opposition to autonomous vehicles. Google is not planning to compete with manufacturers—instead, it will likely build and sell technology to manufacturers—so manufacturers are unlikely to try to obstruct their efforts.” Notes from a conversation with Daniel Fagnant on March 18, 2014

6.

“Autonomous vehicles are a popular emerging research topic in transportation and engineering, and Mr. Fagnant is confident that he could have continued his research at a non-profit or a research institute, had he not chosen to go into academia.” Notes from a conversation with Daniel Fagnant on March 18, 2014

“Technical research in this space is generally well-funded by universities and corporations, and many institutions, ranging from the World Economic Forum to the Transportation Research Board, are discussing the barriers to and impacts of autonomous vehicles (with an emphasis on vehicles that resemble today’s cars and trucks).” Notes from a conversation with Bryant Walker Smith on April 11, 2014

7.“The best opportunities for philanthropy in this space probably lie in areas where the large tech companies and auto manufacturers do not have incentives to act. These opportunities might include:

  • Advocating for regulation that allows smaller companies to compete. The Nevada Department of Motor Vehicles (DMV) imposed barriers to entry that effectively limited the autonomous vehicle market to larger companies—the rationale being that larger companies would be more cautious because they had more at stake. This regulatory approach was appropriate given the level of uncertainty about the safety of these technologies, but it could shut out smaller innovators.
  • Showcasing autonomous transit system technologies. Philanthropists or companies could build functioning demonstration projects, such as driverless shuttle systems in amusement parks or downtown areas, to demonstrate the viability of autonomous transit systems. Though the U.S. military is currently moving toward showcasing some technologies, localities are less likely to build demonstration projects on their own, due to a lack of funding and political will.
  • Developing regulation for alternative vehicle concepts. Existing legislation is ill-equipped to deal with alternative vehicle concepts. For example, federal and state agencies have long struggled to define and regulate low-speed neighborhood vehicles such as golf carts in retirement communities. Nobody is advocating for changes to allow automation on this front.”

Notes from a conversation with Bryant Walker Smith on April 11, 2014

8.

“Although they were slow in beginning to discuss the impacts that autonomous vehicles might have, in the last twelve to eighteen months, urban planners have become much more focused on the topic. Autonomous vehicles are expected to affect many aspects of urban planning, including city administration, highways, transit systems, zoning, and building design. Some people claim that autonomous vehicles will help reduce congestion and emissions, while others caution that they might induce more travel or cause other behavioral changes that would negate those benefits. Mass transit Automation offers opportunities to re-imagine mass transit systems, which, in the U.S., tend to be expensive, inconvenient, and relatively fuel inefficient. For instance, driverless taxis could potentially become an affordable option for people of all income levels. Transit agencies may not have sufficient funding or interest to pursue these projects, however. Automation could enhance, compete with, or detract from these conventional systems. Near-term investments in physical infrastructure Mr. Smith is concerned that worthwhile infrastructure projects, such as a high-speed rail in California, could be blocked by premature arguments about autonomous vehicles. There is currently too much uncertainty about the effects of such technologies to argue against otherwise necessary physical infrastructure projects in the near term.” Notes from a conversation with Bryant Walker Smith on April 11, 2014

“The impact of autonomous vehicles on urban planning is a major emerging research area. Florida is one of the leading states in the effort to anticipate and respond to these implications. Because there is a lot of uncertainty around autonomous vehicles, the exact considerations that urban planners should take into account in their planning remain very much in question. For instance, it is not clear whether urban planners should expect autonomous vehicles to lead to more or less urban density.” Notes from a conversation with Daniel Fagnant on March 18, 2014

“Since the 1950s, the Long Beach Freeway has linked the massive Ports of Long Beach and Los Angeles to, roughly, the rest of the continental United States. Because much has changed in trade and traffic since then, California’s relevant transportation authorities have decided that perhaps this freeway should change as well. The resulting Draft Environmental Impact Statement (EIS), released in 2012, includes several project alternatives that feature a dedicated four-lane freight corridor for the many trucks that service the ports. In two of these alternatives, all of the trucks on the corridor are assumed to have automated steering, braking, and acceleration that enables them to travel in closely spaced platoons of six to eight vehicles. Smoother flows and lower headways mean higher vehicular capacity. Automation–or at least automation-related litigation–is coming to an EIS near you. For a transportation project, automation may be relevant to many of the project alternatives, including the no-build. Potential highway expansions typically use a planning horizon of at least twenty years, and yet several automakers now forecast that they will market vehicles with some kind of advanced automation within a decade. (To put this in slightly more skeptical terms, the self-driving cars that have been twenty years away since the 1930s are now just ten years away.)” The Impact of Automation on Environmental Impact Statements

9.

“While AVs may be commercially available within five years, related research lags in many regards. Much of this is due to the uncertainty inherent in new contexts: with the exception of a few test vehicles, AVs are not yet present in traffic streams and it is difficult to reliably predict the future fol- lowing such disruptive paradigm shifts. Moreover, technical developments along with relevant policy actions, will effect outcomes and create greater uncertainty. With these caveats in mind, it is useful to identify the critical gaps in existing investigations to better prepare for AVs’ arrival. One of the most pressing needs is a comprehensive market penetration evaluation. As KPMG and CAR,72 Google,73 Nissan 74 and Volvo 75 make clear, AVs probably will be driving on our streets and highways within the next decade, but it is uncertain when they will comprise a substantial share of the U.S. fleet. More meaningful market penetration estimates should attach dates and percentages to aggressive, likely, and conservative AV-adoption scenarios. This would provide transportation planners and policy-makers with a reasonable range of outcomes for evaluating competing infrastructure investments, AV policies, and other decisions. Other important research gaps have been identified, with broad topic areas outlined at the 2013 Road Vehicle Automation Workshop,76 as follows:

  • Automated commercial vehicle operations
  • Cyber security and resiliency
  • Data ownership, access, protection, and discovery
  • Energy and environment
  • Human factors and human-machine interaction
  • Infrastructure and operations
  • Liability, risk, and insurance
  • Shared mobility and transit
  • Testing, certification, and licensing
  • V2X communication and architecture

Many important, and frequently crosscutting, questions arise from within each of these topic areas. For example, if driverless taxis become legal and commercially and technologically viable, they could serve many trips currently served by privately owned vehicles. This would reduce parking and ownership needs, and have impacts that cut across the automated commercial vehicle operations, energy and environment, infrastructure and operations, and shared mobility and transit focus areas. Furthermore, this list does not make explicit the need for new transportation planning efforts, with most major public investment decisions planned using a 20- to 30-year design horizon. As long as these and other crucial questions go unanswered, the nation will be hampered in its ability to successfully plan for and introduce AVs into the transportation system.” Fagnant and Kockelman 2013 pg 14.

“AV technology will be disruptive and crosscutting, and several research tasks arise:

  • Develop more precise estimates of the costs and benefits of these technologies and determine whether they accrue to the operator of the vehicle or the public more broadly.
  • Develop better estimates of the distributional consequences of AV technology. (What groups are likely to gain and what groups are likely to lose?)
  • What are lessons learned from the introduction of other vehicle technologies that can prepare NHTSA and EPA for this transition?
  • What capabilities, enabled by both human capital and statutory authority, do NHTSA and EPA require to effectively serve the public interest and facilitate technology development in a rapidly evolving field?
  • How will future fuel economy standards account for AV technology? And how will private and social costs and benefits be estimated?
  • Further develop model legislation concerning AVs to avoid the “50-state patchwork” of laws that has been described by OEMs and other stakeholders as a serious concern for development and deployment of AVs.
  • Analyze the advantages and disadvantages of explicit or implicit regulatory preemption (requiring manufacturers to incorporate the most-promising forms of AV technology by regulatory mandate but simultaneously exempting the manufacturers from state court liability).
  • Analyze existing state “distracted driving” laws and whether they will need to be amended to accommodate AVs.
  • Investigate the potential impact of AVs on travel modes, and how these changes may affect planners at all levels, especially state and federal DOTs.
  • Identify, define, and examine existing models for transportation data management, as well as potential data needs for automated road vehicles. For each model identified, explore whether the model provides insight regarding how automated road vehicle data might be handled. Issues to be explored could include what parties may access personal location information, personally identifiable information, vehicle operation, etc., and how they can and cannot use these data. The latter question should include data access, sharing, and security. The research should then address how these issues would be resolved in the context of different stakeholders (e.g., vehicle manufacturers, data aggregators, government regulators, law enforcement, insurance, vehicle owners and users). The research should highlight best practices and recommend how those might apply to regulations for automated road vehicles.”

Anderson et al. 2014 pgs 146-8.