The automotive industry is beginning to come to terms with the realities of autonomous vehicles.
The technology is proving much harder to develop than many executives and industry leaders initially thought and even advanced semi-autonomous tech, like Cadillac Super Cruise, remains difficult and costly to develop and roll out en masse.
This change in approach is perhaps best exemplified by General Motors‘ and Cruise’s recent decision to postpone the rollout of its self-driving taxi service in San Francisco. The automaker had hoped to debut the service before year’s end and even wanted to launch with vehicles that featured no steering wheel or pedals, but eventually came to the realization that this timeline was too aggressive and delayed the rollout indefinitely.
Speaking at a recent industry conference, Farid Khairallah, a safety systems director at German automotive supplier ZF, provided more insight on the industry’s changing approach to self-driving vehicles. He said that many in the industry now recognize that Level 5 autonomous vehicles that feature no steering wheel or pedals are not the right approach, as they would require too much computing power.
According to Automotive News, Khairahalla elaborated on this, saying that a Level 5 vehicle would need 1 million times more processing power than today’s semi-autonomous vehicles in order to be safe 100% of the time. This would not only make the vehicle too expensive, but would present technical challenges with regard to cooling and other aspects.
“The perception in the industry was, a couple of years ago, that it is paved with gold. But now the industry is looking at it from a more sober, practical point of view,” Khairahalla said, as quoted by AN. “It’s not that easy to go there.”
“The whole industry is rethinking their strategies and what they want to do with this,” he added.
Regulation also presents its own challenges. Even if a manufacturer or supplier were able to solve the Level 5 self-driving problem, there’s no guarantee that regulators will allow them on public roads without a driver behind the wheel. If a safety driver were required, this would likely negate any benefit of the technology for ride-sharing companies, which currently lose a very large chunk of profit to paying drivers.
GM sent a petition to the National Highway Traffic Safety Administration earlier this year to operate cars on public roads with no steering wheel or pedals and despite delaying the rollout of Cruise, is still seeking approval. A decision with regard to the petition is expected to be made soon.
Subscribe to GM Authority for more Cruise Automation news and around-the-clock GM news coverage.
Comments
Yeah…not surprising. Why anyone would want a fully self driving machine with no steering wheel or pedals is beyond me. Who is liable if there is an accident? You cannot *guarantee* something will work 100% of the time especially over a longer range of life.
Apple makes good products and has easy to use software, but it’s still glitchy at some random points. Now think about how much more code and processing power would be needed for self driving. Self driving aids are pretty cool but I’d prefer to be in control of the car I’m driving, and I’m a millennial. Maybe Gen Z feels different but as they get older they’ll probably start to see the issues just like everyone else
I agree.
I can’t picture autonomous driving beyond level 3 being in the same class as other safety features that became standard. What automakers are asking for is no where near as reasonable as ABS or airbags.
On top of consumer sentiment and technological limitations, governments everywhere would have to be willing to drop untold sums of money; not only on new specialty infrastructure, but also repairing outdated infrastructure
Good luck with that.
Zero emissions is a much more attainable goal than zero crashes.
“governments everywhere would have to be willing to drop untold sums of money; not only on new specialty infrastructure, but also repairing outdated infrastructure”
Not if the AV’s can already use existing roads, as what Google is teaching their AV’s to do already. No need to waste government money making new roads and signs if you can teach an AV to use the roads we’ve already taught ourselves to use.
I sure hope Google does a better job of GPS/AV navigation than they seem to have done with WAZE on Android. I have found the WAZE application on my smartphone to be less than reliable on several occasions. It seems to crash, often in the middle of a trip, leaving you temporarily with no guidance. While sudden software crashing is just an inconvenience in a human guided automobile, it could have the potential to be disastrous in an autonomous vehicle. But realistically this will be the case in any vendor’s AV given today’s computer and communications technology. Truly self driving vehicle technology is not anywhere near ready yet.
For now, we use our trusty Garmin GPS when traveling in unknown areas. I would sure hate to be in a no wheel and pedal Google AV that lost its WAZE…
“this would likely negate any benefit of the technology for ride-sharing companies, which currently lose a very large chunk of profit to paying drivers” Ride sharing companies do pay the drivers but they don’t have to pay for the driver’s car, tax/tags, insure, fuel, maintenance costs or care about depreciation…Not sure if its even cheaper to buy robotaxis especially if that robotaxi has a high MSRP…
It’s about time somebody wrote something realistic about the problems associated with making autonomous vehicles practical. The original idea of guiding such vehicles by using RADAR and/or onboard cameras to “see” curbs—and even more sketchy—painted lane lines doesn’t hold up under even just casual inspection; nighttime, rain, snow and ice are just a few problems with such guidance systems.
Additionally, autonomous vehicle will be required to be able to differentiate between a pedestrian or a dog walking in front of it and, say, a plastic bag blowing across the road. And the default answer of just having the vehicle stop at any such occurrence could prove catastrophic under traffic the conditions such as seen in LA and New York City.
GPS is the obvious solution, but it currently doesn’t have anywhere near the resolution required to navigate a vehicle in traffic or even just in a parking lot.
“The original idea of guiding such vehicles by using RADAR and/or onboard cameras to “see” curbs—and even more sketchy—painted lane lines doesn’t hold up under even just casual inspection; nighttime, rain, snow and ice are just a few problems with such guidance systems.”
It does hold up when you teach the bot to remember where the lines are in fine weather. Then all it has to do is recall where the lines are in poor weather….just like how you already do.
You know where the lanes are when they are covered with hard-packed snow because you saw the same lanes in the summer months. Since we know the lanes are under the snow, it’s just as easy for an AV to remember where the lanes in the summer months are when it cannot see them.
The same means of inferred and recalled memorization of lanes and signs is what you and I do all the time behind the wheel. If you can recall and remember, an AV bot can too, even when the bot and the driver don’t have direct and clear line of sight.
“Additionally, autonomous vehicle will be required to be able to differentiate between a pedestrian or a dog walking in front of it and, say, a plastic bag blowing across the road. And the default answer of just having the vehicle stop at any such occurrence could prove catastrophic under traffic the conditions such as seen in LA and New York City.”
Did you fill out a Recaptcha to make that post? If yes, then you’re already on your way to helping an autonomous bot make the distinction between stuff you and I see while we drive. If the bots have been told millions and millions of times what a taxi or a fire hydrant looks like through learning algorithms, it’ll be able to easily recognize one when it sees it.
You might not have ever seen a fire hydrant painted yellow, but if you’ve seen thousands of them painted orange, and wouldn’t have much difficulty in knowing that you’re looking at a fire hydrant if you found one painted yellow.
Level 5 autonomous vehicles will not become available for years, perhaps many years. However, level 4 autonomous vehicles should become available soon, perhaps as early as next year. Of these, expect taxis to be the big success story. By about 2022 autonomous taxis should be good enough to be replacing other taxis everywhere.
So we’ve heard that AVs can’t even handle an elementary 4 way stop situation. Guess we can all sit and wait all day behind the AVs trying to decide who has the right of way. Ubers AV has already killed a person in the roadways at NIGHT when AV “vision” technology should be FAR better than the human eye. Now add in the millions of other scenarios we all deal with on a daily basis and these AVs will be crawling around town creating a nation of gridlock!! This tech isn’t even close to reality within a decade. I’m bettin the house on it. TAKE NOTE “rideshare” company imbeciles currently treating your driver’s like dog crap. You’re gonna need these driver “contractors” around for a LONG time!!
The Uber case was settled out of court when it was found that the pedestrians family couldn’t file a larger case for damages because their family member was in the wrong. The pedestrian was jaywalking and wasn’t wearing any reflective markings, nor was any reflectors on the bike they were carrying. Uber had the video evidence to their defense, not to their detriment in the case. The video wouldn’t haven’t been released otherwise if it exposed Uber to undue legal liability. The Uber AV wasn’t found to be at fault because the technology wasn’t looking for jaywalkers and pedestrians crossing the road illegally. The same accident happens to millions of human drivers the world over every year, so it’s unreasonable on your part to expect prototype AV software to be held to a higher standard this early in AV development.
Furthermore, the AV’s wouldn’t work like how humans do. Humans cause gridlock because the aren’t coordinating their actions with one another, and more often act greedy and poorly behind the wheel. A city of AV’s would coordinate their movements with each other, not just in line-of-sight, but from many blocks away. Humans will never have that level of capability while driving.
So I guess perhaps we need a new legal liability standard that requires victims of AV involved accidents to prove they are innocent of any wrong doing or even any human failing? Be sure you are wearing AV industry approved protective clothing before you venture into a public road? The claimed fact that the video wouldn’t have been released because it would expose Uber to “undue” legal liability speaks volumes. “The Uber AV wasn’t found to be at fault because “the technology wasn’t looking for jaywalkers and pedestrians crossing the road illegally”. Is the Uber programmed to only spare pedestrians that are crossing legally but allowed to hit “illegal” pedestrians? What a ridiculous argument.
Is there no end to the excuses the AV true believers will put forward to advance their “cause”?
You seem to be a know it all, with all the answers on AV’s in your unwavering support for these fantasy vehicles. What’s your story?? I drove part time for a$$hat Company Uber for 3 yrs, 6000+ till they threw me off the platform for 2 “Accidents” on my MV Report (1. Hitting a friggin DEER!! 2. Scratching a bumper pulling to the curb to safely pick up a rider). During MY Uber trips, IF I didn’t take EVASIVE action for literally 1000’s of pedestrians & bike riders ILEGALLY in the streets at night, sober & drunk, wearing black clothes head to toe, NO “reflectors”, no lights, etc etc the Carnage would’ve been staggering!! Using YOUR preposterous logic, Uber & Lyft will be the biggest mass killers in the country, running down people all over America because they were ILLEGALLY in the streets at night & the AVs tech can just run them down because they weren’t supposed to be there! Looking forward to seeing these wizard millennials program the AVs to handle the MASS CONFUSION at bar closing time with crowds of drunks all over, in the streets, seeking out their AV ride home. Good luck with that! ???
AV’s will never be perfect, when driving at highway speeds, a running animal, like a deer, with an randomly changing trajectory could cause the AV to crash…The goal is to make them safer than human drivers…AVs will not deploy nationwide in the US overnight, it has already started with fixed routes, campuses and office parks…
Since we are discussing AV-pedestrian accidents, your comment “AVs will not deploy nationwide in the US overnight, it has already started with fixed routes, campuses and office parks…” If the inclusion of campuses includes college and university campuses, my experience is that, when in session, there are large numbers of students that view jaywalking as a sacred right. In fact many college towns grant the legal right of students to jaywalk and place the responsibility of avoiding them on the motorist. So much for Uber’s claimed legal defense mentioned in the above posts. AVs better look out for not only wayward deer but wayward ,humans as well.
Read the wiki on the Uber AV fatality as its facsinating…Key points from the wiki:
“Evidence subsequently surfaced of a cozy relationship between Uber and Governor Doug Ducey that allowed and protected an immature technology on the streets”
“emergency braking maneuvers are not enabled while the vehicle is under computer control, to reduce the potential for erratic vehicle behavior, according to NTSB.”
Kind of shows how deep this AV conspiracy may go. Anything for “The Next Big Thing” stock appreciation? If this scam ever unravels, wonder what will happen to companies that invested heavily in it, like GM?
If it was a scam, why would so many invest in it? The only scam is that AVs will be absolutely perfect or “deathproof”; ulitemately they’ll only be better than human drivers…Ubers AVs were never rated very highly yet the whole idea is you have a backup human paying attention…Tesla would probably be profitable every quarter if they never invested in autopilot…AVs are already on the roads, the more AVs on the road the higher the odds of more pedestrian deaths…
Why? As PT Barnum may have said: “There’s a sucker born every minute”. Even quicker on Wall Street what with the business media, internet trolls, and computerized trading…
The fact that you immediately smell conspiracy in AV development tells me you’re unstable. AV development is far from covert, and the products of such research are’t driven by scam-hardy individuals.
If this was a conspiracy, it wouldn’t be public knowledge, nor would it have private investors. Back to future fear-mongering FB with you.
Grawdaddy: I guess I now must assume you are a troll, since when you have no real response you resort to ad hominem attacks. I don’t believe I attempted an amateur psychoanalysis of your motives, I stuck to the generally accepted common sense regarding our current societal and technical problems. As to being covert, telling stock analysts that your company would start producing cars with no brake pedal or steering wheel in 2019 was either incredible stupid or intentionally fraudulent, perhaps even a stock “scam”; I don’t know, which do you prefer?