Last week I wrote about the fast pace of developments in autonomous vehicles, and looked at why the technology itself will be well ahead of other factors determining when it can be put into service. This week, we’ll see what steps there are on the way to total automation, and when substantial numbers of vehicles will be on each.
The U.S. National Highway Traffic Safety Administration recognizes four automation stages, but the first three are tied to specific innovations in cars which require active drivers. The difference between this organization’s level 3 and level 4 is a chasm, moving from technology dependent on human takeover to, as a May 8th Fox Business piece put it, “a fully autonomous car.” When we realize that there are a significant number of intermediate steps between these two, and when we want to put self-driving technology in groups characteristic of these phases, we can see a need for another scheme. Accordingly, here is one.
We are now primarily at Plane 1. Vehicles there, as I define it, require drivers to be at the controls essentially continuously. They can still have a variety of robotic or semi-robotic facilities, such as parallel parking capability, but such features do not remove the need for human attention. At this level, most automated tools must be explicitly activated by the driver, who must stay in his or her seat. Plane 1 does not require anything more from the road system or elsewhere in the driving environment than what is there now, and the current, visual-only “Check Engine” lights are perfectly appropriate ways of communicating possible mechanical problems.
At Plane 2, vehicles would only need partial human driving, with something like an aircraft autopilot taking over for long amounts of time. (Indeed, a similar feature already available on Tesla cars, which was announced yesterday to have been involved in what may be remembered as the first self-driving death, is named “Autopilot.”) Drivers here would handle the less routine trip parts, such as their beginning, their end, unusually heavy and complex city traffic, and turning and merging onto expressways. As this phase progresses it would require upgrades to the environment, particularly continuously available Wi-Fi, very detailed GPS, or both. Licensed drivers would still be needed in each vehicle, but they could leave their seats for as long as hours on end. Almost all automated systems could be manually overridden. Vehicles overall would have about as much autonomy as jetliners do today.
Plane 3 still calls for someone in the vehicle able to control it, along with a steering wheel and such, but that person would be more properly called a supervisor instead of a driver. They could still drive, but that would be unusual, mainly for emergency or extreme weather situations, for which the vehicle’s software would provide algorithmically-based notifications. This stage would require robust Wi-Fi or other satellite communication at the start, with further capabilities following as the phase progressed. As in Plane 2, a visual Check Engine light would need to be replaced by something auditory. The general autonomy level of Plane 3 vehicles would resemble that of washing machines or dryers today, with few problems once they are powered on and activated.
The next phase, Plane 4, would remove the need for anyone controlling from inside. The main characteristic here would be autonomous vehicles remotely supervised. One person could monitor, as the stage progressed, from one to dozens of them. As this phase became more established, the need for any kind of physical onboard controls, such as steering wheels and brake pedals, would go away. Passengers, who could now be in any seat, could set the destinations and some other factors, as could supervisors, who would be notified about mechanical failures electronically through systems that could also stop the vehicles or redirect them to a garage or other place where their problems could be solved. Supervisors would also initiate refueling in the same manner. Plane 4 vehicles would have the general autonomy level of today’s large computer system hardware.
The next phase, Plane 5, is where means of transportation would be totally autonomous, with no direct human monitoring. Passengers, police, or owners could set their directions, and compartments could look more like living rooms than the insides of today’s cars. Algorithms would determine when and where they would take on fuel, a process which would not need to involve humans. All trips would be automatically logged with their data uploaded, and precise vehicle locations would be available continuously. The amount of human intervention Plane 5 cars required would approximate that of a refrigerator.
Given these phases, the rapid progress of self-driving technology in general, and the resistance factors I named last week, when might we see which levels of progress around the world? Taking our best guesses that taxis will be one of the first applications at each stage, but that trucks will much more slowly reach them due to fear holding back their legal approval, how long will it take for jobs driving both to go away? Here is how all that might look:
To give some examples to explain this chart, I project that by 2018, 20% of the vehicles in the countries that prove to be the most advanced at driverless vehicle implementation, possibly the likes of Switzerland and Qatar, would be operating at Plane 2 as described above. The United States will have half of its vehicles fully autonomous by 2033. In 2021, there will be about as many Americans driving cabs as there are today, but by 2037 there will be 95% less.
Here is more about the future of self-driving vehicles as I see it. They will be produced not by software, automotive, ride-sharing, or electronic-technology companies alone, but by consortiums of at least three of these four. While that sounds like an easy opportunity for investors, in reality the industry will turn out to be a wild place to put money, with massive winners and total losers as happened with their driver-requiring versions a century before. Electric cars will probably not succeed even as taxicabs, as they have been pushed as viable alternatives or even replacements for hydrocarbon-powered ones since the 1960s and, despite heavy subsidies and even when grouped with electricity-gasoline hybrids, still have only 2% of United States passenger-car sales, with no new reasons for fundamental improvement. Mercedes-Benz has announced several Plane 2 features, including what they call “Drive Pilot” and “Active Brake Assist,” for some 2017 models. And, once again, invisible prosperity will improve while employment will worsen, continuing what looks like the main story of the 21st century.