I’m delighted to be spending this week committing overt acts in furtherance of the Volokh Conspiracy. Since joining Stanford in 2011, I’ve been studying the increasing automation, connectivity, and capability that promise to dramatically change our lives, institutions, and laws. My posts this week will focus on one key example: self-driving vehicles (or whatever you want to call them). The timing is fortuitous, since any remaining legal or technical issues that we fail to collectively solve in the comments section of this blog can be remedied at next week’s U.S. House hearing on “How Autonomous Vehicles Will Shape the Future of Surface Transportation.”
A number of other government bodies are already shaping the legal future of autonomous driving. Nevada, Florida, California, and the District of Columbia have enacted laws expressly regulating these vehicles, California’s Department of Motor Vehicles is currently developing more detailed rules, and a number of other states have considered bills. The U.S. National Highway Traffic Safety Administration (NHTSA) released a preliminary policy statement earlier this year, and Germany, Japan, the United Kingdom, and the European Union have also taken initial domestic steps. Meanwhile, parties to the 1949 Geneva and 1968 Vienna Conventions on Road Traffic are discussing how to reconcile language in these treaties with advanced driver assistance systems.
These efforts tend to view vehicle automation as an incremental process in which driving functions are gradually shifted from human drivers to automated driving systems. The taxonomies developed by the German Federal Highway Research Institute, NHTSA, and SAE International’s On-Road Automated Vehicle Standards Committee (on which I serve) are consistent with this view, even if they are not yet entirely consistent with each other. Yesterday’s cars have antilock brakes and electronic stability control, today’s cars are getting adaptive cruise control and lane-departure detection, and tomorrow’s cars will have highway autopilot and automatic valet parking. A fully automated car capable of driving itself from your suburban home to your downtown office while you are asleep in the back, however, is still a long way from the market.
This incremental view, while correct for the kinds of cars and trucks we regularly see on our roads, risks missing or even thwarting a parallel development that holds significant promise and some peril: low-speed, low-mass, geographically restricted vehicles that are truly driverless. Automated shuttles that are capable of carrying passengers along predefined local routes at bicycle speeds are already here today. These smart carts are explicitly considered in SAE’s draft taxonomy but largely absent from other discussions. This is unfortunate.
Automated shuttles are particularly well-suited for business and university campuses, central business districts, airports, military bases, retirement communities, and transit feeder routes. As compared to conventional motor vehicles, these platforms may be more accessible to developers (credible or otherwise), more likely to be centrally managed and maintained, and able to deal with more situations by simply coming to a stop. Moreover, they foreshadow even smaller systems, like personal delivery robots, that may someday seek their share of the public roadways.
Public roadways, however, are not shared easily or happily, and automated shuttles will need to navigate a complex and probably uncertain legal environment. I’ll briefly consider existing law generally applicable to all vehicles, existing law specific to low-speed vehicles and neighborhood electric vehicles, and recent law specific to automated vehicles.
In general, motor vehicles operated on public roads must meet federal safety standards, must be registered in accordance with state law, and must be driven only by persons licensed in accordance with state law. Even before turning to automated vehicles, this gets surprisingly tricky rather quickly.
Under the federal Vehicle Safety Act, a motor vehicle is one that is “manufactured primarily for use on the public streets, roads, and highways.” NHTSA has in the past used maximum speed as something of a proxy for whether a vehicle meets this condition and must therefore satisfy the agency’s safety standards. Curiously, NHTSA has also looked to whether that vehicle could be used on public roads without violating state law.
States, meanwhile, have not been entirely clear or consistent about either what constitutes a public road or where their licensing and registration requirements apply–two similar but not necessarily identical inquiries. This means that, for example, a university campus hoping to deploy an automated shuttle system would need to carefully consider the status of the facilities that these vehicles would travel along and across.
Federal, state, and municipal governments have tried to clarify the legal status of some lower-speed vehicles. NHTSA has enacted a special safety standard (FMVSS 500) that applies to passenger-carrying low-speed vehicles (LSVs) and neighborhood electric vehicles (NEVs) with top speeds between 20 and 25 miles per hour–but not to golf carts, which are manufactured to travel below 20 mph. States, for their part, have imposed a variety of licensing requirements and operational restrictions for LSVs and NEVs as well as golf carts (PDF) and Segway-like personal transporters (which NHTSA does not regulate at all).
This mixed approach means that an automated shuttle with a maximum speed of only 15 mph might be considered something other than a “motor vehicle,” a “motor vehicle” to which FMVSS 500 applies, or a “motor vehicle” to which the more demanding regular standards apply–and that it might either fall into any number of state buckets or simply miss them entirely.
Recent state autonomous driving laws have thickened this mix. In general, these laws contemplate the possibility of unmanned vehicles superficially but not substantively, and they fail to distinguish among different classes of such vehicles. California’s law also establishes additional regulatory gates through which vehicles must pass if they have no driver inside, and the Department of Motor Vehicles might not even reach these vehicles in its first round of rulemaking.
As I’ve discussed elsewhere, no state has sufficiently considered how existing vehicle codes should apply to any automated vehicle, much less to one that has no human operator in the conventional sense. Moreover, although these state laws distinguish between research-and-development testing and general operation, early shuttle systems that operate as pilot or demonstration projects may bridge this divide–a point to which I will return in an upcoming post.
NHTSA’s policy statement also focuses on conventional vehicles to the exclusion of low-speed, low-mass, geographically restricted ones. The agency “is not aware of any prototype automated vehicle systems that are capable of operating on public roads without the presence of a driver in the driver’s seat who is ready to control the vehicle,” and it “does not recommend at this time that states permit operation of self-driving vehicles for purposes other than testing.” These statements are understandable in the context of the kinds of fully functioning motor vehicles typically regulated by the agency–but not with respect to driverless shuttles.
The kinds of automated vehicles that could serve downtowns, neighborhoods, and campuses in the near future are unconventional in two senses: They have no on-board human driver, and they don’t look like normal cars or trucks. As government and industry move along the spectrum of vehicle automation, they should think broadly about the vehicles as well as the automation.