It's time to clarify autonomous-vehicle laws
The likelihood that you'll encounter autonomous vehicles on public roads will increase this year. That's due in part to the technical success Google has had with its autonomous car program and in part due to the efforts of legislators to promote—or at least define the rules of the road for—the new vehicles.
Update (January 4): Toyota and Audi will demonstrate autonomous-driving features at the Consumer Electronics Show, according to the Wall Street Journal.
The Car Connection last fall reported that Google has racked up more than 300,000 miles in its autonomous vehicles, with the only at-fault accidents occurring when a human driver was in control. That technical feat has been accompanied by legislatures in three states explicitly allowing the vehicles to operate on public roads. Nevada acted first, followed by Florida and, last fall, California. The Center for Internet and Society at Stanford Law School maintains a list of relevant bills under consideration in various states. Bills have been under consideration in jurisdictions including the District of Columbia, Hawaii, New Jersey, and Oklahoma.
The need for such legislation and the legal status of autonomous vehicles in states that have not acted are unclear. Google co-founder Sergey Brin seems to think legislation is not required. CNN reported that when Gov. Edmund “Jerry” Brown signed the California autonomous-vehicles bill into law, Brin was asked who would get a ticket if a driverless car runs a red light. Brin replied, “Self-driving cars do not run red lights.”
However, Bryant Walker Smith, a fellow at the Center for Internet and Society and a fellow at the Center for Automotive Research at Stanford, sums up the ambiguity surrounding the status of autonomous vehicles in the title of a 98-page paper published November 1: “Automated Vehicles are Probably Legal in the United States.” He takes as his starting point “…common law’s presumption of legality.” He goes on to quote a ruling in “United States v. Gourde: “There is no principle more essential to liberty, or more deeply imbued in our law, than that what is not prohibited, is permitted.”
Smith notes that three key legal regimes might be implicated in the question of autonomous-vehicle legal status: the 1949 Geneva Convention on Road Traffic, National Highway Traffic Safety Administration regulations, and relevant state codes. He describes in detail the difficulties of determining the applicability of such codes. For example, California expressly states that a bicycle is not a vehicle, yet a bicyclist may be convicted of violating a law that imposes speed limits on vehicles.
The picture becomes even murkier when it comes to the Geneva Convention on Road Traffic, one of the goals of which is to promote uniform rules to help drivers from one country operate safely in another. It addresses vehicles, convoys, and even draught, pack, and saddle animals and notes that drivers must at all times be able to control their vehicles or animals. Smith concludes that the requirement for driver control refers to animals, perhaps with carts in tow—no one at the time anticipated truly autonomous motor vehicles. As Smith writes, “In 1949…deliberately requiring a motor vehicle to have a driver would have seemed as important as deliberately requiring that vehicle to maintain contact with the ground.” He emphasizes, “Even the visionary behind the General Motors Futurama at the 1939-40 World’s Exhibition stopped short of predicting a truly driverless car.” At that event, the theatrical and industrial designer Norman Bel Geddes proposed an automated highway system that would “correct the faults of human beings as drivers” but not replace them.
Smith concludes as well that federal standards do not prohibit automated driving. Federal rules, he writes, assume but do not require the presence of a driver, and they do not prohibit drive-by-wire systems, although requirements regarding driver activation of hazard lights in emergency situations may present legal ambiguities.
Smith then surveys state codes and concludes that no state expressly requires a vehicle to have a driver, although various codes do impose obligations on a vehicle's driver, operator, or owner. An expansive view of responsibility, he writes, “suggests that various persons could be deemed to operate an automated vehicle”—even nonhuman persons such as corporations, partnerships, and other legal entities. (Then also of course there are the New Zealand SPCA driving dogs.)
Despite various possibilities as to whom should be considered the “driver” or “operator” of any vehicle, state codes could be interpreted to require a human presence. The process of your autonomous vehicle chauffeuring you to work, dropping you off, and then going on its own to find a parking place might run afoul of codes requiring that unattended vehicles have their brakes set, motors shut off, and keys removed. It is possible, Smith writes, that automated vehicles might be considered to be permanently attended by computers or remote operators. In addition, there remain questions regarding serious accidents involving unattended autonomously operating vehicles—who would meet requirements for drivers to remain at the accident scene, report to authorities, and render aid to the injured?
Because of the ambiguity in applying motor-vehicle codes written before autonomous vehicles were envisioned, Smith suggests that states may wish to clarify their laws regarding autonomous vehicles, and he proposes draft language that would address some of the issues he raises in his paper. He notes that while a bill adopting his language would be incomplete (not addressing insurance, liability, and environmental impact, for example), the language does distinguish between an automated vehicle's ordinary and virtual driver. “The natural person occupying or otherwise using an automated vehicle is subject to existing rules of the road unless the manufacturer or insurer of the vehicle has assumed these responsibilities by registering as a virtual driver,” he concludes.
Legislative clarification can be helpful, and effective, nonrestrictive legislation would be welcome. Human “drivers” have already ceded control of their vehicles' braking, steering, adaptive-cruise-control, collision-avoidance, and other systems to computers.
Autonomous vehicles will evolve rapidly. The New York Times reports that Richard Wallace, director for transportation systems analysis at the Center for Automotive Research (CAR), predicts the likelihood of a safe left-lane commute by the end of the decade, with full autonomous driveway-to-destination capability being achieved within 15 to 20 years.
A report prepared by CAR and KPMG examines market dynamics, convergence, adoption, and investment implications. The report authors, including Wallace at CAR and Gary Silberg, a partner and automotive national sector leader at KPMG, touch on legal issues. For example, they ask, “If the driver, by design, is no longer in control, what happens if the vehicle crashes?” They add, “Insurance underwriting will be another thorny issue. Interviews with insurance risk firms indicate that the entire underwriting process will need to be revamped, and a greater portion of the liability could transfer to manufacturers and infrastructure providers (federal and state).”
But Wallace and Silberg mainly focus on technology. Currently, they note, sensor-based approaches cannot adequately mimic human senses and are not sufficiently cost-effective for mass-market adoption. Connected-vehicle approaches offer alternatives, but DSRC (dedicated short-range communication) does not work with pedestrians and bicyclists, DSRC-based V2I (vehicle to infrastructure) might require significant infrastructure investment, and V2V (vehicle to vehicle) would require high market penetration to operate effectively. They suggest that a converged solution would facilitate adequate mimicking of human senses and reduce the need for an expensive mix of sensors and expensive infrastructure investment. They conclude, “We believe convergence of sensor-based and connected-vehicle technologies will happen and will have a positive effect on the adoption of both systems. We think drivers will take the leap. Convergence will bring enhanced mobility and safety and reduced environmental impacts. It may also have far-reaching implications for the traditional automotive value chain and beyond.”
Like it or not, the vehicles you encounter on public thoroughfares will be increasingly operating on their own, whether or not a human sits behind the steering wheel. The Times article quotes Anthony Levandowski, one of the lead engineers for Google’s self-driving-car project, as saying that driving cars “is the most important thing that computers are going to do in the next 10 years.”