1. The Future of Automobility Automated Vehicles and Public Policy
Marc Scribner
Research Fellow
Competitive Enterprise Institute
2014 Preserving the American Dream Conference

2. Defining “autonomous vehicle”
Intervening advanced driver assistance systems: automated technology has been phased-in during recent years, including adaptive cruise control, lane keeping assistance, self-parking (NHTSA Level 1/2 Automation)
In this context, “autonomous vehicle” refers to a highly or fully automated vehicle, one which can direct the core driving functions (NHTSA Level 3/4 Automation)
Broadly, autonomous vehicles are motor vehicles capable of highly or fully automated driving, which means “computer direction of a vehicle’s steering, braking, and accelerating without real-time human input”1
1. Bryant Walker Smith, “Automated Vehicles are Probably Legal in the United States,” The Center for Internet and Society, Stanford Law School, November 1, 2012.

3. NHTSA levels of automation
Automation Level
Definition
Level 0 – No-Automation
Traditional manually driven vehicles, including those with automated warning systems or automated secondary controls (e.g., headlights, turn signals).
Level 1 – Function-specific Automation
One or more independent automated primary control functions (steering, braking, throttling). These include adaptive cruise control, electronic stability control, and dynamic brake support in emergencies.
Level 2 – Combined Function Automation
Two or more automated primary control functions designed to work in unison to relieve the driver of control over these functions. Driver must be able to retake manual control of the vehicle with no warning.
Level 3 – Limited Self-Driving Automation
Driver can cede full control of the vehicle in some situations. Must have ability to retake manual control following warning and transition period.
Level 4 – Full Self-Driving Automation
Vehicle control functions fully automated for an entire trip. Driver has no expectation (or ability) to retake manual control at any point.
Automation Level
Definition
Level 0 – No-Automation
Traditional manually driven vehicles, including those with automated warning systems or automated secondary controls (e.g., headlights, turn signals).
Level 1 – Function-specific Automation
One or more independent automated primary control functions (steering, braking, throttling). These include adaptive cruise control, electronic stability control, and dynamic brake support in emergencies.
Level 2 – Combined Function Automation
Two or more automated primary control functions designed to work in unison to relieve the driver of control over these functions. Driver must be able to retake manual control of the vehicle with no warning.
Level 3 – Limited Self-Driving Automation
Driver can cede full control of the vehicle in some situations. Must have ability to retake manual control following warning and transition period.
Level 4 – Full Self-Driving Automation
Vehicle control functions fully automated for an entire trip. Driver has no expectation (or ability) to retake manual control at any point.

4. SAE levels of automation

5. Potential benefits of automated vehicles
Reduced accidents (human error is a factor in approximately 93% of crashes)
Reduced congestion (suboptimal merging maneuvers are the largest non-weather drivers of road congestion)
Improved air quality (reduced congestion, moving right along the speed-emissions U-curve)
Improved individual mobility (auto access for the disabled, elderly, and youth)

6. Consumer model release estimates
Consumer availability very uncertain given highly proprietary nature of technology
What developers have claimed publicly:
Volvo, 2017 (100-volunteer road testing)
Google, 2018
Nissan, 2020
Daimler, 2025
Continental AG, 2025
Most likely NHTSA Level 3?
Highly automated vehicles by ?
Fully automated vehicles by ?

7. Two technology models to consider
Independent Automated Vehicles
Cooperative Automated Vehicles
Google Self-Driving Car, Washington, D.C., May 2012
Safe Road Trains for the Environment (SARTRE) Project/Volvo platoon, Barcelona, Spain, May 2012

8. Two technology models to consider
Independent Automated Vehicles
Cooperative Automated Vehicles
Capable of directing driving with only onboard equipment and data
No specification mandates or infrastructure upgrades needed, at least in principle
Data are distributed; privacy and illicit manipulation risks are lower—although predictable selfsame machines can be manipulated even when unconnected
Greater potential personal mobility benefits
May contain components of highly automated vehicles, but rely in some way on vehicle-to-vehicle and/or vehicle-to-infrastructure communications to direct core driving task
Probably will depend on spec. mandates and large infrastructure investments
Large cybersecurity risks
Greater potential congestion/emissions reduction benefits

9. Recognizing legality, not legalization
Highly and fully automated vehicles are currently legal throughout the United States
The law does not address the relevant features and thus does not forbid them
Outside of 4 states plus D.C., however, this legality is not recognized by statute or regulation
Recognizing highly/fully automated vehicles’ legality is important to avoid stifling development and deployment
However, legislative and regulatory overcaution represents the most significant threat

10. Current legislative status

11. Current regulatory status
States
Nevada: Regulations took effect on March 1, 2012
California: Manufacturer testing regulations issued May 19, 2014; post-testing operation/licensing regulations due in late December 2014
D.C.: Regulations proposed April 4, 2014
Federal
National Highway Traffic Safety Administration (NHTSA) released its Preliminary Statement of Policy Concerning Automated Vehicles on May 30, 2013

12. Issues and challenges
Google’s safety record? Unblemished, but test data are still too sparse to make statistical comparisons with human drivers (need 725,000 crash-free miles of representative driving)
Federal vs. state regulation? NHTSA at least 3 years away; application impacts of existing state law are unclear
Cybersecurity risks? Connected vehicles face biggest hurdles, but data security (wireless software updates), manipulation, and privacy concerns will be raised even with “fully autonomous,” unconnected vehicles
Crash ethics—how will cars be programmed to crash?
Products liability? Automakers, OEMs, software providers, operators, etc. will face civil liability
Bad legislation/regulations? Has already occurred, unfortunately

13. Demonstrating automated vehicle safety benefits
All Crashes
Fatal Crashes
Vehicle-miles Traveled (VMT)
2.954 x 1012
Vehicles Involved in Crashes
18,705,600
45,435
VMT per Crash
160,000
65,000,000
Crash-free VMT Required for Benefit*
725,000
300,000,000
*Poisson distribution, P-value < 0.01, using 2009 data (Smith, Goodall, Census, NHTSA).

14. Federal vs. state regulation
State/local
NHTSA FMVSSs are safety regulations, dictate design
FMCSA FMCSRs similarly cover motor carrier design (49 C.F.R. Part 393), but also operator-machine interactions (e.g., HOS rules) and licensing req’s
Responsible for licensing, operations, testing
State courts are not clearly preempted by FMVSSs/FMCSRs on tort liability (see Williamson v. Mazda, 562 U.S. __ (2011))

15. Cybersecurity risks
Amplified by connected vehicle technology (spoofing sensors, etc.)
But given that machines are deterministic, even a single “rogue” vehicle on a roadway of unconnected automated vehicles could manipulate thousands of cars (e.g., suboptimal following speed that causes other cars to react predictably)

16. Crash ethics
An automated vehicle must be able to determine the possible outcomes of a trajectory choice, severity of a specific outcome, and the conditional probability of an outcome occurring given the vehicle’s trajectory choice.

17. Crash ethics: Goodall’s three-phase strategy
Begin with a rational moral system designed to minimize crash impacts based on broad principles, such as injuries being preferable to fatalities;
Introduce machine learning techniques to observe human behavior in actual crashes to determine common values; and
Enable the automated vehicle to express its decision logic using natural language to allow humans to understand and correct its ethical processes.

18. Products liability under increasing proximity
Massive increase in information wealth
More operator data will heighten providers’ foreseeability standards, increasing litigation risk
In response, providers will likely demand even more robust data collection
Yet, this then again increases their foreseeability standards

19. How not to regulate autonomous vehicles
Case: Washington, D.C. (2012)2
Original bill included provisions that:
Required autonomous vehicles be powered by “alternative fuels”
Imposed a new mileage-based tax on all autonomous vehicles
Mandated a licensed driver be in the driver seat during autonomous operation
The first two provisions were removed before final passage after harsh criticism
2. See, e.g., Marc Scribner, “Driverless cars are on the way. Here’s how not to regulate them,” The Washington Post, November 2, 2012.

20. Transit and urban development implications?
Potential to resolve first/last-mile challenges
However, improved private transportation services could render most public transit obsolete
PULL! “Self-driving taxis” could make costly auto ownership less attractive and compact cores more attractive
PUSH! However, if technology significantly shifts value of time by making “driving” less physically and cognitively taxing, expect increased outward growth
These two phenomena are not mutually exclusive!
Regardless of eventual outcome, expect large VMT increase!

21. Upshot for policymakers: proceed with caution!
Many unanswered questions
Great uncertainty, particularly with respect to technological evolution and liability
Great risk of overcautious regulation—could delay rollout for consumers and greatly increase prices
Unnecessary delay/price increase if highly and fully automated vehicles are demonstrated to be safer necessarily means increased property damage, congestion, injury, and death

22. Six guiding principles for sound autonomous vehicle policy
Recognize and promote the huge potential benefits
Push back against the precautionary principle
Don’t presume to know how the technology and products liability will evolve
Minimize legislative/regulatory intervention
Focus on developing clear and simple rules that preserve technology neutrality
Make clear legislative/regulatory distinction between highway and non-highway vehicles

23. The Future of Automobility Automated Vehicles and Public Policy
Marc Scribner
Research Fellow
Competitive Enterprise Institute
2014 Preserving the American Dream Conference