Why Semi-Autonomous Driving Will Never Be As Safe As Augmentation

22 Apr

Augmented Driving

The self-driving car industry is blowin’ it.

The definitions of self-driving—from ADAS to SAE automation levels to the inconsistent nomenclature used by the media—are a semantic disaster concealing a vast opportunity. There is no doubt increasing automation will make driving safer, but the safest possible implementation is one that maximizes human capabilities rather than treating them like a cancer.

Automakers are missing the biggest opportunity to profit from saving lives on what is likely to be a long, gentle ascent to Level 4. It requires tossing the insufficient logic behind L2/L3 semi-autonomy and probably even Advanced Driver Assistance Systems (ADAS), and deploying the same hardware and software being developed for L4 as a way to augment human driving.

Though augmented driving represents a clear break from the current crop of semi-autonomous systems, it’s not without precedent. Aircraft are being transformed by automation just as profoundly as cars, but because there is no impetus to move toward pilotless airliners, flight automation systems have been developed to enhance rather than replace human pilots. By following the example set by the commercial aviation sector, automakers can replace the risks inherent to semi-autonomy with the comprehensive assistance of augmented driving.

The Problem is the Transition Gap

Virtually all criticism of Semi-Autonomy focuses on transitions, meaning the length and nature of the control handoff from the system to a human operator.

Transitions are not the problem.

The flaw in Semi-Autonomous driving is inherent: it temporarily substitutes rather than comprehensively assists. The more it improves, the more human skills decline. Even as it improves, every “failure” is attributed to technology rather than human ignorance of it. Its perceived limitations discourage rather than encourage adoption of any form of automation, including future iterations decreasingly skilled drivers will need most, like L4.

Even if someone could “perfect” transitions the overall safety of partial automation will always remain hostage to the atrophying skills of humans in the loop. As Captain Chesley “Sully” Sullenberger stated in an interview about automation, driver’s education is “a national disgrace.” Human driving skills — especially in the United States — have never been great, and the recent spike in American road deaths suggests they are in decline well in advance of automation’s rise. If semi-autonomous systems continue to focus on replacing these skills rather than enhancing them, they will contribute to the very problem they are supposed to solve.

The “transition gap” between declining skills and rising automation will always exist, as untrained humans will always place more faith in technology (and their skills) than warranted. This gap is inherent to semi-autonomy because it is totally binary: it is on, or it is off. That such systems are safer than the average human driver when engaged makes commercializing them a moral imperative, but since they can never improve as quickly as human skill declines, and since the only solution offered by current thinking is L4, they will remain a conceptual dead end, a snake of safety technology eating its own tail until L4 magically becomes ubiquitous at some future date.

That’s nowhere near the best we can do using all the technologies developed along the way.

The Most Important Lesson of Aviation Has Been Ignored

What is Augmented Driving? It’s the synthesis of concepts pioneered in commercial aviation but so far ignored in automotive. What few examples automakers have tried to follow have been limited to infrastructure and protocols impossible to duplicate on the ground in the near or mid-term, like traffic control and ubiquitous communications/location broadcasting.

Augmented Driving ditches the pipe dreams of V2V and V2I by using technology already in place — like drive-by-wire and the increasingly commoditized radars and cameras already part of ADAS/L2 — and adds the higher-resolution GPS and LIDAR-based maps (almost) everyone will build or buy on the way to L4.

The core of Augmented Driving is a car-based implementation of aviation safety systems called Flight Envelope Protections. Airbus and Boeing have been debating and refining these systems for nearly 40 years. That everyone in self-driving research isn’t intimately familiar with them is a disgrace.

What are Flight Envelope Protections? Here’s the wiki:

“…A human machine interface extension of an aircraft’s control system that prevents the pilot of an aircraft from making control commands that would force the aircraft to exceed its structural and aerodynamic operating limits. It is used in some form in all modern commercial fly-by-wire aircraft. Its advantage is that it restricts pilots in emergency situations so they can react quickly without endangering the safety of their aircraft.”

Boeing and Airbus differ over the optimal implementation of Envelope Protections. An Airbus will not allow a pilot to exceed certain bank, roll and pitch angles, regardless of input. A Boeing will, but with deterrent haptic feedback that should shame automotive engineers into retirement. Either type of aircraft, flown properly by a trained pilot, is unlikely to bump up against the protections. An untrained, incompetent or drunk pilot, say, equivalent to most drivers on the road today? Envelope protections are their safety net.

Why don’t we have Driving Envelope Protections (DEP)? We do, in the form of ADAS, but they’re relatively primitive. They exist in the form of Anti-Lock Brakes, Traction Control, Stability Control, and Evasive Steering Assistance systems, but their intervention is inconsistent and largely invisible to — and misunderstood by — drivers. They are poorly or rarely integrated with peer technologies even within state-of-the-art ADAS suites like that in the 2017 Mercedes-Benz E-Class.

Why should drivers be able to panic steer into a wall clearly indicated by their car’s radar sensors? Or steer into a lane where their Blind Spot Monitoring system has identified a truck?

If we can’t guarantee an improvement in driver training outside of car, let’s move it inside by more closely integrating guardrail technologies with the human driving experience, improving both. Let’s gamify driving in a way that encourages and rewards safer driving, building trust between generations raised on analog driving and the technology that can save them from their own mistakes until L4 arrives.

Driving Envelope Protections Are The Future

No one wants to own a connected Porsche 911 capped at the speed limit for safety, but I would love to own an augmented, very-difficult-to-crash 911 that makes me a safer driver in all conditions, at any speed.

Sound crazy? Here’s what Sully had to say about automation vs augmentation:

“It would be much better — at least at a conceptual level — for humans to have more direct engagement with the operation, and technology to provide guardrails to prevent us from making egregious errors, and to monitor our performance. That would be, in terms of our inherent abilities and limitations, a much better way to go.”

All the pieces of real DEP are falling into place. Add high-resolution maps to ADAS, make driver monitoring systems and hands-on-wheel intervals mandatory, add windshield-mounted Augmented Reality/Head-Up-Displays to improve situational awareness, and you have the L2/3 we need, a system where people will remain sufficiently engaged to resolve edge cases binary L2/3 cannot. Drive well and — like aviation protections — DEP will remain completely invisible. Drive poorly and DEP will catch you.

How precisely would DEP work? What about the UI and UX? That’s for future pieces, but here’s my humorous take on how it might work in a 2036 Porsche 911.

Alex Roy is an occasional angel, Editor-at-Large for The Drive , co-host of Autonocast and /DRIVE on NBC Sports, contributor to 2025AD , author of The Driver, set the 2007 Transcontinental “Cannonball Run” Record in 31 hours & 4 minutes, and holds numerous endurance driving records in the United States and Europe. You may follow him on Facebook ,Twitter and Instagram.

Leave a Reply