Google and Tesla are the indisputable leaders in autonomous car technology, even if they have radically different philosophies toward achieving their common goal. While Google is determined to skip straight to fully driverless vehicles that don’t need — or even allow — human intervention at all, Tesla is committed to a stepwise approach borrowed from aviation, what I often refer to as driver-managed Autopilot-enabled cars, where the driver is expected to always be attentive and at the ready.
While Google’s approach may well be the logical end point of the autonomous car revolution, it will certainly take longer to achieve than Tesla’s arguably more “interim” solution, and explains why Tesla’s first generation of Autopilot is already in drivers’ hands and avoiding accidents.
But now Ford’s gone ahead and (quietly) thrown its hat in the game too, seemingly out of nowhere. Despite little to show in the way of autonomous vehicle development so far, it now appears to be going all the way, closely mirroring Google’s approach rather than Tesla’s, and determined to have a fully driverless car on the road in just five years.
As reported by Wired, Ford thinks autonomous vehicles will actually pose greater risks than they will avoid unless and until they can be fully driverless rather than merely driver-managed as with Tesla’s Autopilot.
Spectrum of autonomous vehicle capabilities
To understand what Ford is on about, it’s worth noting that the good folks over at the Society of Automotive Engineers have established a scale describing the spectrum of autonomous vehicle capabilities, from 0 to 5. While 0 might as well be a your childhood tricycle, 5 represents full hands-off autonomous driving with no human inputs whatsoever.
While the Google Self-Driving Car (SDC) project does indeed aim for Cat 5 autonomy, eschewing all driver inputs altogether, Ford’s approach is essentially Cat 4, i.e., the car can indeed drive itself in all situations, but a human can still take over if desired, say, to carve up a mountain road on a Sunday morning drive.
Cat 4 autonomous vehicles allow — but do not need — a human to take over, while Cat 5 vehicles do not even allow human intervention.
The key difference between Cat 4 and Cat 5 autonomous vehicles then is that the former allows — but does not need — a human to take over, while Cat 5 vehicles do not even allow human intervention.
In contrast, Cat 1 systems include things like passive cruise control systems, while Cat 2 includes many of today’s automakers’ so-called “active” safety functions, like active radar-based cruise control, autonomous braking, pedestrian detection, and lane-keep assist.
Cat 3 is where Tesla’s Autopilot falls today: while it is certainly capable of driving autonomously on certain roads, it still requires a human manager: while it can flow with traffic, stay in its lane, avoid cars in your blind spot, and of course brake to avoid a collision, it can’t so much as change lanes without a human command. Nevertheless, it is by far the most advanced semi-autonomous car on the road today.
Tesla’s imperfect interim solution?
Ford’s decision to leapfrog Cat 3 and jump headfirst into Cat 4 autonomy is due to a legitimate concern about Tesla’s Autopilot solution. As Elon Musk has stated repeatedly, Autopilot is not meant to eliminate the human driver; on the contrary, the driver should remain very much alert and always monitoring and managing the drive. He draws the analogy to aviation explicitly:
“Autopilot is what we have in airplanes. For example we use the same term that is in airplanes where there is still an expectation that there will be a pilot. So the onus is on the pilot to make sure that the autopilot is doing the right thing.”
Thing is though, as we discussed earlier, “Google is applying a binary solution to the problem, while Tesla’s is more analog: the driver is neither fully driving nor fully asleep. Which is weird: if the issue were about driving and texting rather than driving and sleeping, we’d call it ‘distracted driving.'”
The issue surrounding this Cat 3 Autopilot-type of autonomy is that there will necessarily be a lag time between when the car signals for human intervention, and when the human actually reacts.
The issue surrounding this Cat 3 Autopilot-type of autonomy is that there will necessarily be a lag time between when the car signals for human intervention, and when the human actually reacts. Studies have shown it can take around three to seven seconds, and sometimes even as long as ten. Not good since, at freeway speeds, this is a distance of up to three football fields.
To be blunt then, this is arguably where the aviation-inspired argument breaks down: even at the immense speeds of commercial jetliners, a temporary autopilot hiccup is not likely to cause a disaster: there’s not much to hit up there, and, despite the narrow flight envelope in which commercial jets fly at cruising altitude, a competent pilot can still take over adequately quickly.
Should a Cat 3 autonomous car demand immediate assistance, however, there are far to many things that can go wrong in those precious few seconds, including and especially instances where the driver-turned-passenger is reading, or worse, dozing off, rather than paying attention. To wit, pilots are not allowed to read, nap, or otherwise be distracted from their jobs of monitoring the aircraft during all phases of autopilot flight.
Ford are echoing Tesla’s stance precisely because they think that this so-called “handoff” problem of how best to alert a driver to take over from the car in an emergency situation, is simply unsolvable.
This then is Ford’s position: they are echoing Tesla’s stance precisely because they think that this so-called “handoff” problem, i.e., the problem of how best to alert a driver to take over from the car in an emergency situation, is simply unsolvable: they don’t believe it is possible to reduce the delay between alert and human control to anything even approaching zero.
Arguably, Ford is taking the most carefully hedged approach by targeting a solution between Tesla’s Autopilot and Google’s SDC pods: by targeting Cat 4 autonomy, Ford aims to combine the best of both worlds, by providing full 100% driverless autonomy, while still enabling a human to take control if she desires.
It is far more likely that Tesla’s Autopilot five years from now will indeed also offer full driverless Cat 4 functionality, without eschewing driver input à la Google.
Truth be told, this is almost certainly the direction to which Tesla is barreling inexorably now: that they are “merely” in the Cat 3 region of the autonomous car continuum now speaks nothing to where they will be in the same five year period of time targeted by Ford. Indeed, it is far more likely that Tesla’s Autopilot five years from now will indeed also offer full driverless Cat 4 functionality, without eschewing driver input à la Google.
Still, the two need not be mutually exclusive: while surely the aforementioned risks inherent with Cat 3 autonomy are legitimate, technology has historically always improved with stepwise functions, cresting one plateau after another. Tesla’s more gradual path towards Cat 4 autonomy may simply offer a smoother transition for the market, if nothing else.
The market acceptance problem
This is a very important thing: driverless cars cannot simply be dumped onto the market with full Cat 5 capability. Put another way, if Google were to start selling its SDC pods to the public, frankly, they would sell only to the most hardened of early technology adopters.
Just because the car can drive fully autonomously, doesn’t mean consumers need to utilize such functionality.
That said, a Cat 4 car like Ford’s should pose no problem: just because the car can drive fully autonomously, doesn’t mean consumers need to utilize such functionality. Or, everybody who buys a 200mph-capable Ferrari must always do top-speed runs every time they enter a freeway. So that argument falls flat on its face then.
On the other hand, by first offering to the public a Cat 3 Autopilot-enabled car, Tesla has scored a monumental achievement, even if it is just an interim one: it will forever be known as the company that first got (semi-)autonomous cars into customers’ hands, so when they do inevitably upgrade Autopilot to support full Cat 4 status, they will have lengthened their lead leaps and bounds beyond even where they are today.
And that matters. That matters a lot. Not just for Tesla, but for consumers too.