Waymo’s small step/big leap from Level 4—to Level 4
The big news in the world of automated vehicles this week was Waymo’s announcement that they’ve been testing a few of their self-driving Chrysler Pacifica minivans on public roads around Phoenix, Arizona, without human “safety drivers”, since October 19th. This marks a major milestone in the development of the technology—before this, whenever an automated vehicle was being tested on public roads, a human was behind the wheel to act as a backup driver in case the automated system screwed up.
Actually, that’s not quite the case—the Google self-driving car project (Waymo’s predecessor) was already confident enough to run a demonstration sans backup driver two years earlier. On October 20, 2015, one of their automated prototypes took a legally blind individual, Steve Mahan, on a tour around the streets of Austin, Texas, with no backup driver.
Already in 2012, Steve Mahan had taken a brief demonstration ride through suburban San Francisco—with no human driver behind the wheel. Google/Waymo was able to pull off this early demonstration by fitting their car with training wheels, as it were. In the 2012 demo, their automated car had a police escort, and while there was no human backup driver behind the wheel, a test driver sat in the back with an emergency kill switch at hand. In contrast, for the 2015 demonstration, there was no police escort nor backup driver on board. While Waymo hasn’t made clear in precisely what respects the backup-driver-free tests they have been conducting this year represent an advance beyond the demonstration they pulled off in 2015, it appears to boil down to the fact that the 2015 demo was a one-off event, while the recent trials are part of a broader program of testing.
Earlier this year, Waymo’s self-driving test vehicles began offering free rides to select residents in the Phoenix suburb of Chandler. During this pilot project, a trained test driver has always been in the driver’s seat, ready to take the wheel if necessary. Waymo now plans to continue this service with an employee monitoring the vehicle’s operation from the back seat rather than the driver’s seat. The human safety monitor won’t be able to grab the wheel or hit the brakes, but they will be able to press a button to pull the minivan over and bring it to a stop at the roadside. In time, Waymo will dispense with the on-board human monitor, though passengers will still be able to hit an emergency pull-over button. Eventually, Waymo will open up the service to the general public, rather than just the pre-selected members of their Early Rider program.
To begin with, Waymo’s automated cars without human backup drivers will operate in a restricted zone within Chandler, where the less hectic suburban traffic makes driving more manageable for robot drivers. As the tech advances, the territory covered by the fully self-driving fleet will expand to cover the greater Phoenix region. And, crucially, Arizona’s dry, arid climate makes for friendlier conditions for robots that have yet to master driving in a wide range of weather conditions. This would minimize the number of situations where Waymo’s cars would have to pull over and wait out inclement weather, as they are currently programmed to.
As the vehicles become more reliable in rainy, snowy, icy, and otherwise difficult weather conditions, and in complex, chaotic driving situations that challenge automated vehicle sensors and algorithms, Waymo will expand their service to new cities. The question arises—when will their self-driving vehicles, operating without human backup drivers, be capable of driving anywhere, anytime, in the full range of situations humans currently drive in? In other words, when will Waymo—or one of its many challengers—reach Level 5, the summit of the hierarchy of automation defined by the Society of Automotive Engineers?
As ever, the answer is shrouded in uncertainty. Depending on who you listen to, it will happen in a few years or a few decades. But also as ever, there is no need to fixate on Level 5. Level 4 is very much here, as Waymo’s recent announcement reminds us. And Level 4 was around long before Waymo’s recent tests. A Level 4 vehicle is simply an automated vehicle that is capable of operating without any human assistance or backup, as long as it is within its so-called “operational design domain”—a range of situations and environments defined according to parameters such as road type, maximum speed, weather conditions, and time of the day. Waymo’s self-driving cars, with no human backup, are beginning with an operational design domain restricted to public roads in a selected area in Phoenix, in a particular set of weather conditions. But even when the domain of operation is far more restricted than this, automated vehicles can still be very useful. We can see this vividly in a particular variety of automated vehicle that uses technology far less sophisticated than that in Waymo’s test vehicles. These comparatively primitive automated vehicles are, of course, driverless trains.
Driverless trains are far from cutting-edge technology—the people mover at the Tampa International Airport was launched back in 1971, and larger-scale driverless urban transit systems have also been running for decades, with early examples like the Port Island Line in Kobe, Japan (1981) and Vancouver’s SkyTrain (1985). Stretching the concept of an operational design domain a bit to apply it to trains, you could say that these driverless trains have a very restricted domain—they of course don’t have to steer, and they operate in tightly controlled environments, running through their own tunnels, along exclusive elevated guideways, or down rights-of-way protected by fences. Despite their restricted operating domains, driverless trains excel in their job of providing cost-effective, efficient mobility to large numbers of city dwellers.
Waymo and others are pushing hard to expand the operational design domain of Level 4 vehicles, but experience with driverless trains shows that there’s no need to wait for automation technology to advance further, let alone to Level 5, before we can exploit it intensively. Waymo is clearly confident that their technology is already powerful and reliable enough to enable a pilot of fully self-driving vehicles in a selected suburb of Phoenix; what is more exciting is the fact that their technology, or something similar, would be more than capable of enabling automated bus rapid transit (BRT) in protected lanes in any city that wants it, today.