Convergence: Public Policy and Autonomous Vehicles
At Eno’s March 24 Convergence, the first afternoon session brought together representatives of automakers, microchip manufacturers, and federal regulators together with an academic researcher in a discussion entitled “Beyond Speculation: The Pragmatic Policy Role in Autonomous Vehicles.”
Participants in the panel discussion were: Blair Anderson, the new Deputy Administrator of the National Highway Traffic Safety Administration; Harry Lightsey of General Motors; Ken Laberteaux of Toyota; Sam LaMagna of Intel, who until recently was Director of Advanced Driving Technologies for their Transportation Solutions Division (and who is a member of the Eno Center Board of Directors), and Dan Fagnant of the University of Utah Traffic Lab. The discussion was moderated by former NHTSA chief counsel Jackie Glassman.
“AVs” can mean a lot of things to a lot of people, so Glassman began the discussion by framing AVs in terms of the NHTSA framework levels:
- No-Automation (Level 0): The driver is in complete and sole control of the primary vehicle controls – brake, steering, throttle, and motive power – at all times.
- Function-specific Automation (Level 1): Automation at this level involves one or more specific control functions. Examples include electronic stability control or pre-charged brakes, where the vehicle automatically assists with braking to enable the driver to regain control of the vehicle or stop faster than possible by acting alone.
- Combined Function Automation (Level 2): This level involves automation of at least two primary control functions designed to work in unison to relieve the driver of control of those functions. An example of combined functions enabling a Level 2 system is adaptive cruise control in combination with lane centering.
- Limited Self-Driving Automation (Level 3): Vehicles at this level of automation enable the driver to cede full control of all safety-critical functions under certain traffic or environmental conditions and in those conditions to rely heavily on the vehicle to monitor for changes in those conditions requiring transition back to driver control. The driver is expected to be available for occasional control, but with sufficiently comfortable transition time. The Google car is an example of limited self-driving automation.
- Full Self-Driving Automation (Level 4): The vehicle is designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip. Such a design anticipates that the driver will provide destination or navigation input, but is not expected to be available for control at any time during the trip. This includes both occupied and unoccupied vehicles.
Glassman then noted that when it comes to safety regulation, there has always been a division of labor – the federal government regulates the safety of the vehicle, and state governments regulates the safety of the drivers. But when the vehicle becomes the driver, who is in charge of the regulations?
Anderson responded by discussing the AV that Google wants to test in Mountain View, California that is envisioned as a level 4 AV. In response to Google’s request that NHTSA remove regulatory barriers, NHTSA responded that while they could envision a way to consider the self-driving system as the vehicle operator, but that it would only apply to specific parts of the regulations, not to the parts of the regulations that currently refer to human-specific physical interactions (like the ability to press brake pedal). In terms of the broad picture, Anderson said that a lot of demonstrated testing must be done to prove that an AV is more safe than a driver.
Laberteaux said that amongst the AV community, there are two competing schools of thought. “Group A” believes that the AV market will begin with cars that are able to automate certain parts of driving, usually those with the least complexity (staying between lines, maintaining a minimum distance from the preceding car, moving through stop-and-go traffic) but that the scene of places where the car would take over would gradually grow from that. He said that day is essentially here, with Tesla introducing a “hands-off” feature and other automakers following.
“Group B” according to Laberteaux are people who think that Group A’s vision is a terrible idea because Group B fundamentally believes that the problem of handing back control from the car to the driver in a safe, but nearly instantaneous, way in the event of an emergency is fundamentally unsolvable. If the whole point of hands-off driving is to allow the driver’s attention to drift, and then something really unexpected happens, how does the driver take over quickly enough?
Lightsey said he also sees a “Group C” – deploying level 4-capable vehicles in controlled areas at first, with drivers ready to take over, but with a path to eventually go driver-free. He also saw ridesharing technology as an initial area of entry for AVs that would then get passengers more comfortable with the idea of owning an AV themselves.
LaMagna saw a gradual technological evolution – if you keep adding features, eventually the car will be able to drive itself. In the interim, LaMagna said that manufacturers need to think about the usage model. Currently, marketing focuses on individual consumers. – what will they want, and when will they be ready for a self-driving car? Beyond that, there are people who aren’t looking as car as asset to acquire for mobility service, who don’t look at car as extension of personality but thing as part of a fleet to make them money. He also noted that when consequences of a software failure are as high as they are in AV (as opposed to, say, an operating system bug in a new cell phone, which can simply be rebooted), regulators will place a high safety burden on manufacturers. But everyone will have to accept that has to undergo testing in the real world at some point.
Fagnant was emphatic that full level 4 automation should be the goal. He said that the eventual level of handoff risk was unknown, as was the level of risk of level 4 AVs, and that two strong unknowns cancel each other out.
The moderator then pivoted the discussion to the impact of AVs on the environment. Laberteaux said that the environmental impact depends on marriage of automation and shared-use mobility. In order to get to that point, people have to get comfortable with idea of a full level 4 driverless car. He pointed out the fact that building laws used to require elevator operators and that now trains like those in the Detroit airport are fully automated, concluding that once you build enough public evidence and experience of safety, people come to accept it. While you could have limited no-driver zones for level 4, we will have level 2 and 3 before that. He also concluded that AVs could eventually cause total vehicle miles-traveled (VMT) – a key number in the aggregate environmental impact of driving – to go up, not down, saying that if driving becomes easier and more stress-free, all other things being equal, you get more driving.
LaMagna concurred with that VMT conclusion, asking the audience not to allow themselves to get trapped into thinking that the proliferation of self-driving cars will reduce the number of cars on the road, or to assume that due to self-driving cars there will be fewer cars manufactured per year. He said that if in-car mobility gets that easy, people will be traveling when they might otherwise travel, citing his elderly mother, who he said keeps bugging him for a self-driving car and threatening to move in with him unless she gets one.
Laberteaux asked a rhetorical question: if you are doing the same VMT with fewer cars, how is that an environmental win? He also pointed out that level 4 AVs would not necessarily take cars off the road but would take cars out of disuse and see them used more hours per day, which to him as an automaker is not necessarily a bad thing. He added that if a car has a 150,000-mile life, but the cars are level 4 AVs used in ridesharing, the automakers could sell the same number of cars but see them used up faster.
Fagnant said that level 4 AVs used in ride-sharing would be a prime market for electric vehicles, since “range anxiety” would not an issue because someone else is responsible for charging the car. A proliferation of shared on-demand fleets could result in total VMT reductions. He also pointed out a key environmental benefit from shared-use level 4 cars – the reduction of ground-level ozone. (Over half of carbon monoxide emissions, he said, are generated during the first two minutes of a car’s operation, as the catalytic converter is getting warmed up. Fewer cold starts equals less smog, even if the vehicles stay in operation much longer.)
He also envisioned huge efficiency savings from coordinating cars with signals and saw other environmental benefits from the decreased need for parking and lane construction (with more efficient platooning of trucks through AV technology).
The conversation then turned to cybersecurity, which all participants acknowledged was a key issue. Lightsey noted that both GM and Toyota belong to the new “Auto ISAC” (Information Sharing and Analysis Center) to work jointly on cybersecurity best practices. Anderson said that USDOT now has a dedicated office looking at AV cybersecurity and said that CS is incredibly important from the standpoints of both safety and consumer confidence. NHTSA held a roundtable discussion in January to talk about vulnerabilities.
At the end of the question and answer session, a questioner addressed the key issue of fleet turnover, which more than anything else will set the timeframe through which the AV transition will occur. Anderson said that full fleet turnover takes 20-30 years, so the expectation for automakers is that AVs need to be able to operate around human-driven cars for foreseeable future.
LaMagna closed by saying that we will live in a world with full and partial AV and human-driven. He said that today, cars are designed to be on road for up to 20 years and to have available spare parts to keep going for that time, so 2016, automakers are planning model year 2021 and intending to have parts for those cars available until 2041. He asked if perhaps the auto industry should move from this model (fixed technology to be repaired to the state of original manufacture throughout its useful life) to a more computer-like model of constant upgradeability (summarized as: do you really want to replace 20-year-old technology with more 20-year-old technology?).
In summary, this Convergence panel was a valuable look at the challenges that face the transition to autonomous vehicles. But with a group of experts unable to agree on questions so fundamental as “will it ever be safe to turn control of an AV back to the driver in an emergency?” or “will AVs result in much less VMT, or much more VMT?”, it is difficult for policymakers to set long-term plans for the future of AVs in the surface transportation system.