Autonomous Vehicles NOT Ready for Prime-Time

[KuroNekko]

For starters, people just seem to have a very hard time with seat belts. The police have to enforce laws to keep people from preventable risk of injury or death to themselves. While I'm sure some would advocate "Darwinism", the reality is that these injuries and deaths drive up health care costs and hospitals lose a lot of money treating stupidity. In a public hospital, that means a lot of tax payer money. If the government can take measures to reduce the costs, they will and they have.

Yes, put me firmly in the Darwinism camp. I also don't believe the medical costs should be passed along. Poof! There goes the phony justification for why the government has jurisdiction.

The cost of all of this nannying is staggering. Get into a minor fender bender and $3,000 worth of air bags go off (even though you were going 20 mph with your seatbelts on), the price of the repair has just been tripled for no reason. The never ending fiddling with tire pressure monitors which seem purposely designed to send you to the dealer (twice as expensive as your local shop and five times as expensive as doing it yourself) is exhausting.

I am fully capable of checking tire pressures, turning my head when backing up, modulating the brake pedal in a panic stop situation. If I choose to buy a rudimentary car without all of this BS on it, I should be able to as an adult who takes personal responsibility. If need be I'll sign some kind of disclaimer.

I agree with some of your points and personally, I don't want too much inter-connected technology in my car either. I like how things are in my Kizashi and it's one of the reasons I plan on hanging on to it for a long time. That being said, we can't attribute your level of attention and care to responsible ownership and driving to everyone out there. For most people, a vehicle is a transportation appliance and driving is a chore. Many don't care for either. Basically, a lot of safety has been largely up to automakers implementing technology and governments enforcing laws given many drivers can't do simple things like buckling in or not driving drunk for even saving themselves. Related to that, I predicted you'd take the Darwinism perspective but in a civil society, you just can't have a hands-off approach on a lot of things, especially related to public safety. It costs a lot of money when people get hurt or injured so it makes sense to mitigate that, especially when people are sharing public roads. After all, sharing public roads means one can easily get into an accident and get hurt at no fault of their own. This then drives up costs both in a personal and societal sense. For example, my brother is a medical doctor and during his med school and residency days, he worked at a large county hospital. There, they would treat all kinds of people ranging from drug addicts, gang members, the mentally ill, the elderly, the sickly, to people rushed there for serious trauma and ER needs. In essence, they are extremely busy with all kinds of things so if there was anything to reduce the number of people coming in from preventable injuries, it would be beneficial in ways ranging from cost to capacity. It's not only about cost-savings but also reducing the load on the healthcare system which is already overloaded and inefficient. Any doctor will tell you the best medicine is prevention. While it's annoying to have to deal with the cost of airbags being deployed, a fractured bone or ruptured organ costs a lot more and takes more resources to deal with.
Also consider that while cars have gotten safer, traffic accident injury and death rates have gone up. This is largely due to more people driving as gas prices are low. In essence, the driver is always the weakest link and the least reliable factor.

That being said, I still do like the idea of more basic vehicles that forgo more expensive components related to safety, technology, luxury, etc. for the reasons you state. There are merits in simplicity for the sake of reducing costs and improving reliability. I know these vehicles are available in many other countries. The problem, it seems in America, is that stripper models don't sell too well as most consumers want the features. Even if laws allowed such vehicles to be sold in the US, I wonder if automakers will deem them economically viable.

[Ronzuki]

Solid Darwinism proponent here as well...I know you're shocked Kuro. I'm curious as to why you feel I and everyone else should be forced to absorb costs laid upon me by 'laws', or any other method, promoting laziness and stupidity? Driving is a chore? Well boo-hoo for those individuals, either don't drive or get over it, be responsible and do the things that need to be done to operate your machine safely. There's many things in life that I don't care for such as the over-use of technology.

More laws to protect...sure, how's that been working out?

http://6abc.com/california-town-bans-te ... g/3147005/
I find it interesting the places in the country that either have or are proposing this non-sense. You talk about expense, the creation of these types of ridiculous laws costs john-Q-taxpayer plenty. And, who is going to enforce this new law? The same police that I observe routinely breaking the very same laws they are supposed to enforce? Great.

When I saw the report above on the evening news last night they had video of some douche walking straight into the side (lucky for him) of a fire truck. LAMO, Without divine intervention (more laws) you can bet the farm Darwinism will fix that genetic defect right quick.

[Ronzuki]

Nope, still not ready....predictably, it's time to cue up the lawyers so the real expensive fun can begin.

[Ronzuki]

Hmm...how should I vote?

This week’s poll: Responses to Uber Crash

How do Engineer readers think the automotive industry should respond to yesterday’s
fatal Uber crash between a pedestrian and a driverless cab?
Uber was testing autonomous XC90s in the Phoenix, Arizona area.

The day that many observers of the development of autonomous vehicles have been dreading
arrived yesterday, when a pedestrian was struck by a driverless vehicle being tested by taxi
hailing app company Uber in Tempe, Arizona and later died. The US Consumer Watchdog
organisation has called for a national moratorium on autonomous vehicle testing, saying that
“there should be a national moratorium on all robot car testing on public roads until the
complete details of this tragedy are made public and are analysed by outside experts.”
Details of the accident are still unclear, but it appears that Elaine Hertzberg was struck by the
car, which was in autonomous mode with a human observer behind the wheel, at around 10
PM when she crossed the road away from a designated and illuminated pedestrian crossing.
According to Tempe’s chief of police, Sylvia Moir, the first that the observer knew of Mrs
Hertzberg’s presence was the sound of the collision, and in the San Francisco Chronicle she
states that the car was travelling at 38mph in a 35mph zone and did not attempt to brake, but
the driver states he did not have time to react.

The car in the incident was a Volvo XC90 SUV equipped with cameras. Uber chief executive
Dara Khosrowshahi said on Twitter that the company was working with local law enforcement
to understand what happened, while the US National Highway Traffic Safety Administration
and National Transportation Safety Board are both sending teams to Tempe. Uber is
suspending the autonomous vehicle trials in all North American cities while investigation takes
place; it has been testing in Pittsburgh since 2016 and is also carrying out trials in San
Francisco, Toronto and the Phoenix area, which includes Tempe. Last year, Uber took its
autonomous vehicles off the roads after an accident that overturned a Volvo SUV, also in
Arizona, but the programme was later reinstated.
Consumer Watchdog’s privacy and technology project director, John Simpson, blamed lax
regulation for the accident. “Arizona has been the wild West of robot car testing with virtually
no regulations in place,” he said. “When there’s no sheriff in town, people get killed.”

In a statement, the Institution of Mechanical Engineers said that Uber was right to suspend
autonomous vehicle trials while the accident is being investigated. “In 2016 the IMechE in our
case study on autonomous and driverless cars raised the needs to address societal questions
before highly and fully automated cars are both accepted and legally able to be positioned on
our roads,” said head of engineering Jenifer Baxter.

https://www.imeche.org/docs/default-sou ... f?sfvrsn=0

“Engineers will need to create an environment where connected autonomous vehicles can operate safely with or without an
operator during the transition period to a fully autonomous vehicle system. This transition
period could last for several decades.”

Volvo has issued the following brief statement. “We are aware of this incident and our thoughts
are with the family of the woman involved. We are aware that Uber is cooperating with local
authorities in their investigation.”

Although the investigation is in its early stages, we’d like to know how Engineer readers think
the industry should respond. Should it follow Consumer Watchdog’s call, and cease all testing
of autonomous vehicles until the investigation is complete? Should there be a focus more on
the technology involved, by which we mean the operating system of the car’s autonomy and
the sensors that provide it with information? Or should such tragedies, which are common on
the roads in any case, not discourage the development of autonomous vehicles and testing
should continue? In the comments, you might discuss whether this response should be
worldwide or confined to the US.

We welcome discussion, although as always it will be moderated to ensure that it remains on
track and constructive. We will publish the results of this poll on 27th March.

[KuroNekko]

I guess some don't care that the victim was jaywalking at night. I read about 4 different articles on this incident and looked at the incident photographs, including an aerial map. There is no crosswalk in the area visible in the incident photographs, videos, or aerial map. The victim was walking her bike across a street at night without using a crosswalk which was likely illegal. Thousands of people are hit by human drivers in these sorts of conditions a year, but it's a big deal here because an autonomous vehicle did so (after several months of testing and millions of logged miles among them).
While I certainly do see the accident as tragic and something autonomous technology will need to work better at, it's foolish to negate all other factors here just because of the controversial technology involved.

As for, "it's time to cue up the lawyers so the real expensive fun can begin", don't worry, lawyers already make tons of money with human clients in auto accidents. Drive through any major city like Los Angeles or Las Vegas and "accident chaser" attorneys will have their billboards advertising their services. Some of them will even list huge sums of money like $150,000,000 to $500,000,000 (yes, that's half a billion dollars) to represent the amount of money their firm won in all their settlements and trials for all their clients. Hence, lawyers are already making tons of money from human drivers in cases involving accidents, DUIs, to speeding.
Also considering that an autonomous vehicle probably has a ton of data ranging from video footage, sensor log, GPS data, etc., it will likely be easier to factually determine what and how it happened.

Lastly, I recall some stating they were a "Solid Darwinism proponent", supporting the notion that some idiots in society should just die off rather than burdening society with their stupidity. Why shy away from it now? The cold brutality of survival evolution doesn't allow for exceptions depending on our biases. The victim here was very likely jaywalking at night and probably didn't even bother to look to see if a car was coming as it's evident she was hit immediately upon entering the road. It was likely just chance that an autonomous vehicle hit her because it happened to be there. It could have easily been a human driver from all I'm seeing and reading so far of the accident. She was hit immediately upon entering the road from the sidewalk which is evident from the damage to the vehicle, damage to the victim's bicycle, and location of everything. Reports also state she had just entered the road when she was struck. It really looks like she literally walked directly onto the street upon which she was hit immediately. I can't imagine a human driver would have done something much differently. But don't take my word for it. I've attached photos from a news video of the scene to show you the conditions.

Let's start with the likely point of impact. This is hinted by the location of the police evidence markers in the first photo. The Uber vehicle is located past it in the second photo (the impact location is behind the police truck with the door open in the second photo). See a crosswalk anywhere in any of these photos? Nope.

Location of impact.JPG (64.26 KiB) Viewed 6392 times

Location of impact and vehicles.JPG (51.58 KiB) Viewed 6392 times

Below is a picture of the Uber vehicle and the bicycle. Note the location of the damage to the vehicle. It is also being reported that the Uber vehicle was in the rightmost lane when the accident happened. This strongly suggests the victim and her bicycle were crossing the street from the sidewalk and were hit immediately upon entering the road.

bike and Volvo.JPG (61.96 KiB) Viewed 6392 times

The picture below shows a close-up of the damage to the Uber Volvo. Note that the impact damage is on the right side, closest to the sidewalk. Again, this supports the notion that the victim came from the side and was immediately hit as she attempted to cross the street. She probably did not even look for a vehicle given she was hit immediately.

Volvo.JPG (56.98 KiB) Viewed 6392 times

Note the damage to the bicycle. The front wheel is bent but the frame and rear wheel look undamaged. This indicates the impact occurred to the front of the bicycle. She was not rear-ended as she rode in a bike lane. She was either hit from the front or the side at the front of her bicycle. All reports state she was walking her bicycle as she attempted to walk across the road, away from a crosswalk. Considering the damage to the car, the bicycle, the location of impact, and other reported facts, it's quite evident the impact occurred immediately after the victim left the sidewalk to jaywalk her bicycle across the street.

Bicycle.JPG (65.22 KiB) Viewed 6392 times

While the death of the victim is indeed a sad incident, I'd like to see a lot more evidence of a clear failure of the autonomous driving technology over a human driver in a similar condition. While there is clear reason for bias, it should be noted that the human back-up driver in the Uber vehicle stated that there was no time to react and take over to prevent the accident. Given what I can tell of the incident, I'm really not convinced a human driver would have done something much differently than the Uber vehicle driving itself. Technology cannot always prevent people from dangerously jaywalking at night in front of oncoming traffic and getting hit.

I'm going to wait for the details and the facts of the accident to come out to see if my theory on the accident is correct before I fault autonomous driving technology. At this point, I'm not convinced I, myself, could have prevented this accident if I was driving.

[KuroNekko]

Some more detailed articles are coming out now on the accident and as I suspected, the facts seem to show the victim literally walked out in front of the vehicle.
Some quotes from the article:

"The Volvo was in self-driving mode with a human backup driver at the wheel when it struck 49-year-old Elaine Herzberg as she was walking a bicycle outside the lines of a crosswalk in Tempe, police said."

"Tempe police Sgt. Ronald Elcock said local authorities haven't determined fault but urged people to use crosswalks. He told reporters at a news conference Monday the Uber vehicle was traveling around 40 mph when it hit Helzberg immediately as she stepped on to the street."

"The pedestrian was outside of the crosswalk, so it was midblock," Elcock said. "And as soon as she walked into the lane of traffic, she was struck by the vehicle."

"The public's image of the vehicles will be defined by stories like the crash in Tempe, said Bryant Walker Smith, a University of South Carolina law professor who studies self-driving vehicles. It may turn out that there was nothing either the vehicle or its human backup could have done to avoid the crash, he said."

https://www.yahoo.com/finance/news/cras ... nance.html

[Ronzuki]

I could care less that the woman was j-walking. Automation demands exception handling. Anyone that claims to be an automation engineer that is too lazy, in-experienced or flat-out too stupid to realize that exception handling is the largest component of an automation project has zero business being allowed to call themselves one. Unfortunately, this problem is getting worse, not better, as time goes on. When exception handling is ignored or the exceptions are not fully dealt with, these are the results. In this case, death. I can't say it any more clearly: there are just far too many exceptions to be accounted for in sending an autonomous vehicle down a road, period.

Automation functions best under predictable and repeatable conditions, the perfect world...roadways are about as far from that as you can get. Cameras in automation...a dicey, expensive, failure-prone proposition. We as a company no longer entertain vision projects because of those reasons. Unless all conditions are perfectly repeatable, lighting being the most critical and finicky, you can go broke very quickly trying to achieve 100% repeatable which is the delusion people expect when they commission us to solve a problem (usually removing people from their pay roles). Unless you're spending many, many 10s of thousands of dollars on the hardware (and ever-changing software), vision projects that are far less complicated, and more predictable, than navigating U.S. roadways rarely produce anywhere near that 100% index.

My opinions are defined by my extensive career experience in automation. None of what these companies are trying to accomplish in our not-so-perfect transportation system will ever work in the required repeatable fashion, reliably and predictably, thus not 'safely'.

The topic of the thread is "Autonomous Vehicles NOT Ready for Prime Time". And now the spotlight is on that fact. Well duh. The arrogance to think an autonomous machine can be set upon our abysmal roadways before the 'track' (the infrastructure) was made consistent and similar everywhere first is laughable. Well it always seems to take tragedy for the masses to get riled up about things that should not be allowed. I should have added "The over-use of Technology in Automobiles" as well based upon thread content. My apologies.

More simple explanation as to why I initiated this thread in the first place:

https://hotair.com/headlines/archives/2 ... -know-see/

On top of that, I also completely agree with the following since this ridiculous garbage was set loose upon public roads by people that do not fully comprehend what automating anything entails:

http://www.foxnews.com/us/2018/03/19/se ... arges.html

[KuroNekko]

Your first sentence shows an utterly unrealistic expectation. It's just not based in reality given someone jaywalking directly in front of a moving vehicle is not completely avoidable in all situations for anyone or anything. Even the best human driver in the world cannot avoid all collision scenarios. For one to expect this vehicle to have avoided hitting the woman, the vehicle would likely have had to defy the laws of physics given the evidence present at this time. I have high hopes for the future of automated driving technology but defying the laws of physics is not one of them.

Also, your experience with automation is certainly impressive and extensive but likely isn't directly comparable to automated driving and your comments reveal that. In your line of automation (was it in assembly equipment?) it may be realistic to have machinery with exception handling for nearly zero error. On a road with an infinite number of uncontrolled variables, that's an impossible standard to meet. I don't think anyone intelligently expects automated driving technology to be 100% perfect. It's just expected to be safer than a human driver, many of whom are flawed.

To expect an automated vehicle to avoid a jaywalker like in this case is just as unrealistic as expecting a bulletproof vest to stop any bullet at any range or a flu vaccine to prevent any strand of the virus at any time. It's neither how it works nor is realistic at all. These things are created to mitigate problems and offer alternatives to increased exposure and risk. As much as it's unrealistic for body armor to stop a .50 BMG round at 20 feet or for a particular flu vaccine to prevent all strains of the flu virus, it's unrealistic for automation to respond perfectly for matters that would cause failure for just about any other operator. Hence, the fact that this victim was jaywalking is not only important, it's essentially the most critical detail in this incident. It explains how and why this collision occurred. I'm now convinced she literally walked directly in front of this vehicle given what I can piece together of the actual evidence. I'm also convinced this accident is not something you or I could have likely avoided if we were behind the wheel.

With that in mind, if you were the operator of this vehicle, would you want someone saying that the woman jaywalking in front of your car is irrelevant in why you hit her? No. You would absolutely make it known that you didn't have the time to prevent the collision given she walked right in front of your vehicle, likely completely unaware of your presence. You would argue it was simply not avoidable. Why then would you expect a different standard for automation if a human driver couldn't avoid it under the best circumstances? I'm convinced successful "exception handling" in preventing this collision would have required Uber to defy the laws of physics. It's unrealistic, just like absolute perfection.

Referring back to the article I linked before, the quoted professor (Bryant Walker Smith, a University of South Carolina law professor who studies self-driving vehicles) said it best;
"We should be concerned about automated driving," Smith said. "We should be terrified about human driving."
The article highlights why:
"Autonomous vehicles with laser, radar and camera sensors and sophisticated computers have been billed as the way to reduce the more than 40,000 traffic deaths a year in the U.S. alone. Ninety-four percent of crashes are caused by human error, the government says.

Self-driving vehicles don't drive drunk, don't get sleepy and aren't easily distracted. But they do have faults.

In 2016, the latest year available, more than 6,000 U.S. pedestrians were killed by vehicles."

Given these facts of the status quo, we can't have unrealistic expectations of absolute, physics-defying perfection from automation. We simply need something better and safer which automated vehicles likely are, especially as technology progresses.

While this accident will likely cause setbacks from a suspension or a moratorium of automated vehicle testing/implementation, we should be careful. Moratoriums resulting from this will likely only be in the US and that will give other nations and their companies an opportunity to develop and advance the technology. The last thing a technology and innovation leader like the United States and its companies needs is to tie our own hands behind our backs as others take the opportunity to dominate the technology and therefore global market share. This would be a terrible mistake especially given these specific circumstances as the woman is very likely a victim of her own Darwinism, not automation.

[Woodie]

Where did her phone land? LOL

I don't think any vehicle with any driver could have avoided this one.

[Ronzuki]

"Also, your experience with automation is certainly impressive and extensive but likely isn't directly comparable to automated driving and your comments reveal that. In your line of automation (was it in assembly equipment?) it may be realistic to have machinery with exception handling for nearly zero error. On a road with an infinite number of uncontrolled variables, that's an impossible standard to meet. I don't think anyone intelligently expects automated driving technology to be 100% perfect. It's just expected to be safer than a human driver, many of whom are flawed. "


This is exactly the kind of thinking that should never be permitted in automation. I've stated before I have worked with AGVs in the late 80s. AGV...Automated Guided Vehicles. What was that Volvo that killed the woman? Same damn thing. Has zero to do with industry. My experience is not impressive at all in the grand scheme of what it is we do for a living. Those AGVs I have experience with, loaded, weighed 8,000 lbs and hauled along quite fast. They were capable of damage to people, structures and other equipment. And when the tech failed, they caused big damage. People, were smart enough not to trust them and stayed the hell out of their way.

Let me be clear, I AM NOT INTERESTED IN THE UNIMAGINEABLE AND RIDICULOUS LONG-TERM COSTS ASSOCIATED WITH AUTONOMOUS VEHICLES that I will be forced to fund in multiple ways as this non-sense plows forward.

Woodie, I don't doubt that it was unavoidable by the average every-day distracted driver, or even a non-distracted driver. However, automation didn't solve the problem either. So we're right back where we started then aye? I'll drive my own vehicle at much less cost to me, without the added cost and aggravation of all the unnecessary tech.

Anyone here ever enroll in, and complete a full-blown motorcycle rider safety training course? I did in my early 20s. The very first thing they hammer in to your head, and over and over again throughout the course, is where's the threat and what are you going to do to avoid it or minimize the harm to yourself? I still use that training every day driving. They're training you to train your brain to look at everything and constantly scan for potential threats to you while riding.

Kuro, there is no positive argument for testing a 3000 pound automated pipe-dream on PUBLIC roads. The kind of automation testing that we would be required to perform on a system such as this (a fairly simple and basic test of an object jumping out in front of the AGV moving 40mph) was quite obviously not thought of, or performed, or worse, was, and didn't work out so well and they plowed forward anyway. That never happens right? "eh what the hell, we'll fix that in the next release". Again, the arrogance. We'll never know as this is business and the dirty laundry, the real facts, will not be made public for a host of reasons all detrimental to everyone involved in this venture. You're in the legal profession...why not dig into getting hold of the legal details surrounding the investigations of this "accident"? The juicy stuff...what controlled testing was done, by whom, what are their credentials, how was the testing done, what were the results? All that good documented stuff that these people won't share because its business and there's already been lawsuits regarding tech copyrights.

Your keep citing accident rates. Here's a low cost, high-impact solution to reduce accidents for you to chew on that would immediately and significantly improve your stats...ready for it?... render dumb phones completely useless in moving vehicles. Cause and effect, people would be looking up and around instead of focusing on their crotch, console, right seat cushion or steering wheel. Phones, like the one the butt-head in the Isuzu box truck was glued to when he almost plowed into my ass end 2 days ago as I had to quickly slow for and stop behind a horse and buggy that shot out from the right shoulder into the traveling lane to make a left turn. It's very simple you see...I was operating my vehicle, saw the buggy, realized what buggies are and what they're likely to do and reacted to the buggy. The moron behind me was distracted by starring down at his phone and didn't notice me slowing down and stopping for the buggy until glancing up from the steering wheel, where the phone was in his hand, at the last second and swerved on to the shoulder stopping by my right rear quarter panel. Human error (let's not sugar coat it, call it what it is, distraction) made possible by what? Correct, technology. Technology that has no business in an automobile. It's very simple logic don't you think? Technology induced problems that we must cure with....more and more, expensive, technology. The Cycle continues.