Autonomous Vehicles

[dgaddis]

We've talked about self driving cars a bit here and there. Lets do it some more.

The recent review from the NTSB about the incident where a self driving Uber hit and killed a pedestrian/bicyclist (she was pushing a bike) has some interesting revelations (to me, anyhow).

NTSB: Uber Self-Driving Car Had Disabled Emergency Brake System Before Fatal Crash : The Two-Way : NPR

To review:
A Volvo fitted with Uber's self driving system hit and killed a woman. She was walking across the road pushing a bicycle. It was after dark. The "driver" wasn't looking at the road until just before impact.

I'm not trying to sound harsh or insensitive, but the woman who was hit and killed is primarily the one to blame here. She was walking across a four lane road in the dark and wasn't looking out for traffic. It's sad she's dead, but, had she been watching for traffic, she'd still be alive. It's not like the car randomly swerved and jumped the curb and hit her - she walked in front of it. That said, the car should have seen her and acted accordingly - I mean if I'm driving down the road and someone is walking across the road, and I see them, I try not to hit them, even if they shouldn't be walking out in front of me. Happens all the time actually, and so far I've yet to mow anyone down.

Now, lets discuss the car. What's surprising to me is how the Uber system works. It recognized something was in the road ~6secs before it hit her. It registered a bicycle a little over 1 second before it hit her.

The two big "WTFs" in my mind:

1. The car did not brake or use any evasive maneuvering. The Volvo has factory installed automatic emergency braking that will act on it's own, but the Uber system turns it off "to prevent erratic behavior". The system relies on the "driver" to intervene with braking and/or steering.
2. The Uber system does not give any alerts to the driver that it has seen something in the road.

In other words, the car 'thought' "oh look there's a person directly in my path of travel" and then did nothing.

It seems to me that #1 (auto braking is turned off) is done because the system isn't good enough to be trusted. And with that in mind, I can't for the life of me understand why #2 would be true. Why wouldn't it alert the driver?

With self driving cars, once people get used to them and start trusting them they will not be paying attention to the road. Therefore, you cannot count on them to do emergency braking or maneuvering. That's exactly what happened here, the "driver" wasn't looking at the road for quite some time. And if these systems do require 100% attention from the driver, then what's the point at all? Because they'll certainly pay even less attention than they do now with a self driving car vs one they have to drive themselves.

Further, I wonder what the process was like that approved the use of these things. Did the NTSB or Arizona DOT know that the system would turn off automatic emergency braking, and not alert the driver to potential dangers?

[becomingblue]

I read an article a few days ago similar to the one you referenced. I was surprised to read that the system(s) had been disabled on the car. When the actual accident occurred, I was stunned that it happened the way it did. These automated systems were supposed to save lives and in this case, it didn’t. Now we find out stuff was disabled. I wonder if the driver knew.

I’m not sure what to believe anymore related to the news reports on these cars: testing, miles driven with no accidents, etc. Are they working or not? I know there are teething pains with any new system of this magnitude, but I wonder what true reliability of this stuff is.

Personally, I think automated cars and the push for them is pure snake oil. It isn’t going to work. It can’t work just by putting gadgets in a car. There will have to be infrastructure added to the road system itself and right now in my area, merely filing craters/potholes seems to be a chore. If the system doesn’t work everywhere, it shouldn’t be used anywhere.

I could go on. I’m not normally a betting person, but I would wager the farm that in 10 years, 20 years, whatever, automated cars will not have saved any more lives than proper driver training. IF you could even quantify either.

[mmoore]

I would wager the farm that in 10 years, 20 years, whatever, automated cars will not have saved any more lives than proper driver training. IF you could even quantify either.

with the increasing amount of distracted driving I would take that bet.

[David Tollefson]

I read an article a few days ago similar to the one you referenced. I was surprised to read that the system(s) had been disabled on the car. When the actual accident occurred, I was stunned that it happened the way it did. These automated systems were supposed to save lives and in this case, it didn’t. Now we find out stuff was disabled. I wonder if the driver knew.

I’m not sure what to believe anymore related to the news reports on these cars: testing, miles driven with no accidents, etc. Are they working or not? I know there are teething pains with any new system of this magnitude, but I wonder what true reliability of this stuff is.

Personally, I think automated cars and the push for them is pure snake oil. It isn’t going to work. It can’t work just by putting gadgets in a car. There will have to be infrastructure added to the road system itself and right now in my area, merely filing craters/potholes seems to be a chore. If the system doesn’t work everywhere, it shouldn’t be used anywhere.

I could go on. I’m not normally a betting person, but I would wager the farm that in 10 years, 20 years, whatever, automated cars will not have saved any more lives than proper driver training. IF you could even quantify either.

The only way it's going to save lives is what I see as the eventual end-game of autonomous vehicles -- all other users will be banned from the public roadways, including motorcyclists, pedestrians and cyclists.

[becomingblue]

Agreed. But, when they pry my cold dead hands from the handlebars, as they say.

[Too Tall]

Saw something similar to this concept being hauled on a flatbed in DC today >>

[j44ke]

I'm not trying to sound harsh or insensitive, but the woman who was hit and killed is primarily the one to blame here. She was walking across a four lane road in the dark and wasn't looking out for traffic. It's sad she's dead, but, had she been watching for traffic, she'd still be alive. It's not like the car randomly swerved and jumped the curb and hit her - she walked in front of it.

I don't agree that she is primarily to blame. The driver of the car wasn't paying attention to where the car was going. The car was an experimental vehicle. Wouldn't paying attention to where an experimental vehicle was going be the primary job of the driver? It is an experimental vehicle, not an actual finished done ready for prime time real world use car. I know that area - it isn't ever totally dark there. It is extremely confusing in terms of all the various sidewalks, some of them on the median, bike paths, bike lanes, turning lanes, impending traffic light, etc. Even though signs say don't cross over to the median from the sidewalk, what if you are riding in the bike lane southbound and need to get over to the northbound bike lane? They look perfect for that. And with all the ASU students in that area with apartments and nearby school buildings, you have to pay attention when driving around there. The lack of reflectors correctly positioned on the bike and the presence of drugs in the system of the woman killed are not conditions that make her primarily to blame, especially when there is no toxicology of the driver of the autonomous vehicle (evidently.) And wearing dark clothing also does not mean she is to blame for getting hit. At most, she showed bad judgment assuming that oncoming cars were either going to allow her to cross the street there or that she had enough time to make it across the street at that point. But still she isn't to blame for her death. Bad judgment has nothing to do with blame. Bad judgment could however be used to describe what is an unfortunate accident. Bad judgment which the driver of the autonomous car also showed by forgetting that even though she was sitting in an autonomous car, she was still the driver of that car, not a bystander in the driver's seat. Two people + two cases of bad judgment = one death.

The one to blame are the officials who allowed this car on the road to use unsuspecting people as guinea pigs with no warning that the car approaching them is being operated by a computer program. That car should have been covered with flashing lights and emitting a loud alarm sound to alert people to the possibility that this car might see them and not stop and that no one in the car would stop it.

Screen Shot 2018-05-25 at 8.40.42 PM.jpg

[j44ke]

The road where the accident occurred in Phoenix is a divided road with a planted central median and two to four lanes on either side, depending on how close to an intersection. In Europe it would be an autobahn with stoplight intersections. Crashes at intersections on these sorts of streets can be spectacular. I arrived at one just after a pickup truck broke a Honda Accord in half and then crashed into a quickee mart. The speed limits often seem irrelevant.

[caleb]

The road where the accident occurred in Phoenix is a divided road with a planted central median and two to four lanes on either side, depending on how close to an intersection. In Europe it would be an autobahn with stoplight intersections. Crashes at intersections on these sorts of streets can be spectacular.

It is a uniquely American species of stupidity to have high traffic four lane roads with cars often going 60mph+, at grade, with lights, and right turns on red.

[rwsaunders]

Look at the variety of sponsors on the University of Michigan’s Mcity’s website for example…Toyota, HONDA, Ford, State Farm, Verizon, Deloitte to name a few. CAV’s attract strong interest from researches and investors alike and are tied into the subject of “Smart Cities”. The relatively recent surge in research, fabrication and test deployment of urban vehicles, was a seed started many years ago and government support behind the subject matter has strong ties to the military as well. I’m not saying this in a negative manner, but the link is pretty clear when you start reading more about where the industry is headed.

https://mcity.umich.edu/

https://mcity.umich.edu/did-you-know...-mcity-opened/

https://mcity.umich.edu/our-vision/fast-facts/

[vertical_doug]

Look at the variety of sponsors on the University of Michigan’s Mcity’s website for example…Toyota, HONDA, Ford, State Farm, Verizon, Deloitte to name a few. CAV’s attract strong interest from researches and investors alike and are tied into the subject of “Smart Cities”. The relatively recent surge in research, fabrication and test deployment of urban vehicles, was a seed started many years ago and government support behind the subject matter has strong ties to the military as well. I’m not saying this in a negative manner, but the link is pretty clear when you start reading more about where the industry is headed.

https://mcity.umich.edu/

https://mcity.umich.edu/did-you-know...-mcity-opened/

https://mcity.umich.edu/our-vision/fast-facts/

Only in America is the real world condition have a stop sign shot up with buck shot.

[rwsaunders]

Only in America is the real world condition have a stop sign shot up with buck shot.

Wasn't that awesome? I liked the skater boy with the tossle cap pulled over his/her eyes and of course, the buck.

[72gmc]

I think the corollary is what a typical industrial robot is like versus 3-CPO in the movies or any of the dancing robots from Boston Dynamics.

That’s a good corollary. I’ll also guess that someone at some point didn’t want to accept that constraint, but constraints aid creativity, and there could be much more progress by now if the investments were focused on realistic use cases.

[72gmc]

“It feels like the widespread use of autonomous driving is seven years away, and it’s been seven years away for 10 years,” says U.S. Department of Transportation secretary Pete Buttigieg. “So the question is, will it be seven years away 10 years from now, or will we actually be getting somewhere?”

https://jalopnik.com/u-s-in-valley-o...hec-1849699276

[vertical_doug]

well, not to beat a dead horse but......

Tesla, Ford and VW Sound the Death Knell for Driverless Car Hype
2022-10-27 12:30:15.913 GMT


By David Welch and Craig Trudell
(Bloomberg) --

The autonomous-driving sector just endured a day that tech
and automotive giants may well look back on the way Wall Street
recalls March 16, 2008.

Whereas the day Bear Stearns collapsed was an epochal event
in the global financial crisis, when flawed assumptions about
the value of mortgages pushed banks to the brink — and some over
the precipice — Oct. 26, 2022, will go down as the date that
seismic consequences emerged from years of faulty presumptions
about driverless-vehicle technology.

First came the shock that Argo AI, the startup Ford and
Volkswagen had each seeded with multibillion-dollar investments,
was shutting down. Within hours, Reuters reported that Tesla’s
self-driving claims are under criminal investigation. A person
familiar with the matter told Bloomberg that Justice Department
prosecutors in Washington and San Francisco have been probing
statements by the electric-car company and its executives since
last year.

It’s difficult to come up with two more polar-opposite
approaches to a mission one Ford executive said Wednesday will
be harder than putting a man on the moon.
Elon Musk put a target on Tesla’s back in 2016 by starting
to charge thousands of dollars for what the company calls Full
Self-Driving (FSD) capability. Six years later, the CEO
acknowledges the system still isn’t “feature complete,” and
cautions customers to expect two steps forward and one step
back.

By contrast, Argo CEO Bryan Salesky emphasized the need for
safe and limited deployments of test vehicles and close
partnerships with cities and stakeholders that its driverless
cars would share the road with.


Argo.ai seemed to take road safety more seriously than many
competitors, including a creative collaboration with A lot of people have seen Argo as one of the AV
"good guys."
Neither the scorched-earth nor the nice-guy method is
working.

Tesla is the subject of two defect investigations by the
National Highway Traffic Safety Administration and headed for
the first of several potential trials over crashes blamed on
Autopilot, its driver-assistance system. California accused the
company in August of misleading consumers, and a Golden State
resident who sued last month is proposing class-action status
for his claims that Musk has been stringing the public along
with perpetual promises that the company is on the cusp of
perfecting the technology.

Fans of the world’s richest man have gotten accustomed to
frequent posts from the soon-to-be Twitter owner about new
iterations of FSD beta software beaming to their vehicles. After
Musk tweeted recently about a next major release coming this
week, one follower replied with relief, writing that he’d been
hesitant to use the latest version of FSD after his Tesla veered
toward an oncoming car.

Elon Musk
@elonmusk
@teslaownersSV@28delayslater Update probably goes to wide
release next week. This is a big one.

Ford thought when it first invested in Argo five years ago
that it would be able to broadly market cars capable of going
driverless in certain conditions by 2021. Now, the automaker has
concluded it needs to invest in driver-assistance technology
that’s more achievable in the near term. Its decision to switch
gears led VW to walk away, too, according to people familiar
with the matter, and Argo was unable to attract new investors.
The $2.7 billion impairment recorded on its investment in
Argo dragged Ford to an $827 million net loss last quarter. VW,
which reports earnings on Friday, announced an almost identical
injection in the startup in 2019.

“The team we have at Argo has been working on what I
consider to be the hardest technical problem of our time," said
Doug Field, who Ford hired away from Apple’s car project last
year. “It's harder than putting a man on the moon.”

This is a world away from what car and tech leaders were
saying when driverless-car hype was at its peak. McKinsey
predicted just three years ago that global revenue generated by
autonomous vehicles could reach $1.6 trillion annually by the
end of this decade. The head of General Motors-owned startup
Cruise similarly talked in early 2020 of a trillion-dollar
addressable market. Chris Urmson, who said while at Google that
the goal was for his son to never need a driver’s license, sent
out a memo to staff at his cash-strapped startup Aurora
Innovation last month laying out options including cost cuts and
even a potential effort to sell to Apple or Microsoft.
Argo is arguably the most substantial casualty within the
self-driving space to date, though it isn’t the first. San
Francisco-based Zoox sold to Amazon in 2020, and Uber cut bait
with its self-driving unit months later, offloading it to
Aurora. Early last year, GM’s Cruise acquired Voyage, a startup
that had been trying the narrow use case approach to autonomy,
operating in Florida retirement communities.

When GM reported quarterly results this week, analysts
pressed Cruise CEO Kyle Vogt about the state of autonomy. He
argued companies that are deploying and expanding now have game,
and are distinguishing themselves from those that don’t.
“We're seeing increased separation between the companies
operating commercial driverless services, and those that are
still stuck in the trough of disillusionment,” Vogt said.
“What's happening here is that the companies with the best
product have pulled ahead and are accelerating.”
Time will tell just how far along GM and Cruise will get in
building a viable business. The unit aiming for $1 billion
revenue by 2025 lost $497 million in the most recent quarter,
bringing its total deficit this year to $1.4 billion.

Ford CEO Jim Farley is skeptical the industry is anywhere
close. “Profitable, fully autonomous vehicles at scale are a
long way off,” he said Wednesday. “And we won’t necessarily have
to create that technology ourselves.”
Like reading this newsletter? Subscribe to Bloomberg.com
for unlimited access to trusted, data-driven journalism and
subscriber-only insights.

--With assistance from Keith Naughton, Monica Raymunt, Edward
Ludlow and Tom Schoenberg.