Self Driving Cars?

[kombat]

No, they did not accidentally initiate a go-around.

It's semantics. From your link:

The pilot flying (PF) selected an autopilot (A/P) mode (flight level change speed [FLCH SPD]) that instead resulted in the autoflight system initiating a climb because the airplane was below the selected altitude

The autopilot had previously been configured with the go-around altitude set, so when the pilot accidentally selected the incorrect mode ("flight level change speed"), the autopilot advanced the throttles in order to climb to the go-around altitude. It was configured this way so that in the event of a go-around, they'd simply select this mode and the plane would automatically climb to the appropriate altitude. So it was set up for the possibility of a go-around, and the pilot accidentally engaged it.

Upon immediately realizing his error, he manually pulled the throttles back to idle, and turned the autopilot off. This unusual sequence of commands is what disabled the autothrottles.

What really happened was that they were too high above the glide slope when they were 5 nm out from the runway. In trying to descend, they tried to slow down by pulling the throttle levers back. This action overrode the Autothrottle which had been controlling speed, and it went into HOLD mode, and stayed inactive. This was not noticed by the aircrew.

That's not exactly correct. Pulling the throttles back on its own would not have disconnected the autopilot. It was the odd combination of selecting the FLCH SPD mode, then turning it off, then manually overriding the throttles, that resulted in the unnoticed autothrottle disengagement.

At 100 ft, they gave up and did a go-around. The Autothrottle probably responded properly and advanced up

No, the autothrottles weren't engaged - that's the whole point. The pilots had complete control of the airplane. In its final configuration, the computer wasn't controlling anything. The pilots were in full manual control of the aircraft. They manually advanced the throttles to go-around power, but as you noted, big engines like that have nontrivial lag time, and it was too late for them to begin producing meaningful thrust to avoid a crash.

If the Autopilot was off, then of course the corresponding display would be off.

No, I don't mean his autopilot display was off - I mean his primary flight display (PFD) was off. The big, tablet-shaped panel front-and-center in the middle of the dashboard in front of each pilot. I don't know why any pilot would ever turn that off, as it conveys all sorts of useful information about heading, speed, attitude, altitude, weather, etc., but that's what the report said.

 

[kombat]

I beg to differ. In the above picture, several people were not recognized.

Are you referring to the ones indoors? Or the guy partially obscured by the street-sweeper thingy? I would hope that the car would avoid hitting those things for reasons other than correctly identifying them as pedestrians.

 

[kombat]

Earlier, I said "Pilots do not mess with "stuff"". Just now recall an incidence where one airline pilot did mess with stuff, trying to prove his knowledge of aerodynamics to his copilot.

I was going to call you on this earlier, I'm glad to see you've softened your point a little bit. You seem to have been advancing the narrative that everything in aviation is 100% fault-tolerant, fail-safe, and that pilots are highly trained professionals who would never take any shortcuts. This is all, of course, total fiction.

Airplanes routinely fly with one or more systems inoperable. All aircraft have a "snag sheet" that documents the problems with a particular aircraft. Most of these are fairly trivial and do not impede the safe operation of the aircraft, provided the pilots are aware of the limitation and have a safe workaround. But it happens all the time, and occasionally leads to crash.

Air Canada flight 143 (the "Gimli Glider") happened at least partially because the pilots didn't have a working fuel gauge, and instead relied on (miscalculated, as it turned out) fuel weight.

Turkish Air flight 1951 crashed short of Amsterdam because a faulty radio altimeter caused the autopilot to reduce the throttles under the mistaken belief that the plane was already on the ground.

Indonesia AirAsia flight QZ8501 had a faulty rudder trim limiter system that had been reported for maintenance dozens of times, but never actually properly fixed.

Pilots routinely take shortcuts or deliberately defeat safety mechanisms. In the crash I just referenced, Indonesia AirAsia flight QZ8501, the pilots got annoyed at the repeated warning messages, so they rebooted the entire flight augmentation computer (during flight!). This caused a slow, uncommanded roll, that went unnoticed until it was too late to recover.

As highly trained as pilots are, they still make big mistakes, as in the Asiana Air 214 crash we've already discussed. Thankfully, there are usually enough redundancies in place to prevent these little mistakes and faults from resulting in a catastrophic crash, but when you combine enough of these flaws together into a single incident, well, that's really the only way airplane accidents ever happen. If everything worked perfectly all the time, and pilots never made mistakes or took shortcuts, there simply would never be any crashes. The very fact that there ARE accidents is evidence that even aviation is not the elite, spare-no-expense-for-safety world you seem to want to portray it as.

 

[ERD50]

I was going to call you on this earlier, I'm glad to see you've softened your point a little bit. You seem to have been advancing the narrative that everything in aviation is 100% fault-tolerant, fail-safe, and that pilots are highly trained professionals who would never take any shortcuts. This is all, of course, total fiction. ....

I never read his posts that literally (or are you reading something into them?).

I thought his point was clear - aviation systems have a much higher level of fault-tolerance, they do have highly redundant systems, and pilots are highly trained, and still sometimes they do take shortcuts or simply make mistakes.

That, while not perfect, is still a large contrast to the automobile systems.

-ERD50

 

[Midpack]

According to an 'expert' in an article from Consumer Reports, getting to the point where self driving cards can handle 90-95% of road conditions and driving problems is just about done, It's that last 5-10% of oddball things that happen while driving that are the big problems.

Also, apparently they are working on networking the cars so if a car on a road a mile ahead of you spots a big pothole, you car can be instantly updated and avoid the pothole.

It is astonishing how much progress several organizations have made. Here's a video (below) of a Tesla in level 5 mode, and Waymo has far more capable vehicles. It's important to note the conditions for the drive were almost perfect - roads, weather, other cars were reasonably predictable, no pedestrians, etc.

I AM NOT SAYING LEVEL 4 of 5 CARS ARE RIGHT AROUND THE CORNER. That last 5-10% is where the difficulty lies. And like others, I expect first gen level 4-5 cars will be quite expensive, out of reach for most. We may see them in a few years, but I think it may be 30 years before they're very common.

I'd be unnerved giving the car that much control for quite a while, trusting the car will take time. Even in the video, the car navigates some intersections too quickly for my comfort.

https://youtu.be/C3DbrYx-SN4

 

[kombat]

Self-driving cars that are expensive will be useful to people who need them and can afford them. But to reduce accidents, you have to deny the masses the steering wheel. Then, it has to be inexpensive so that it becomes mandated, else the poor have no transportation.

That's a good point, and it occurred to me that the answer is probably automatic public transit. This, to me, seems like low-hanging fruit for self-driving vehicles. Busses take the same routes, all the time, and don't need to worry about navigating poorly-delimited dirt roads, or deal with multi-lane, high speed highway traffic. They simply trundle along city streets, often in their own dedicated lanes, stopping at periodic intervals.

If you eliminated the driver, you'd eliminate a substantial part of the cost. That savings could be passed on to the users, giving the poor an accessible mode of transportation. Moreover, automating the drivers would feasibly reduce accidents, which would correspondingly reduce lawsuits, and again, costs for the municipality.

Reducing the cost is not easy to do. I do not really see how.

Economies of scale, eliminating all the controls and instruments that become unnecessary in a self-driving vehicle. Automating production, replacing complicated gas engines with simpler, more reliable electric ones.

I follow traffic laws, and do not understand why people oppose speed cameras, and red-light cameras.

Because the speed limits are set unrealistically low. I think the general public is far more accepting of red light cameras, because the vast majority of us don't run red lights. Running a red light is clearly dangerous, and thus we have no sympathy for anyone "caught" doing so by an automated ticketing camera. However, we all know there's nothing inherently unsafe about driving 35 mph in a 30 mph zone, and thus it comes across as more of a "tax grab" than a safety issue.

 

[kombat]

I thought his point was clear - aviation systems have a much higher level of fault-tolerance, they do have highly redundant systems, and pilots are highly trained, and still sometimes they do take shortcuts or simply make mistakes.

You're probably right, it's possible I've inadvertently constructed a strawman argument in my head and am arguing against things NW-Bound hasn't actually explicitly asserted. I just wanted to make it clear that while certainly held to a higher standard, aviation is by no means perfect. And yet it's still extremely safe. Therefore a similar argument could be made for self-driving cars. They don't have to be perfect to be acceptably safe, and even relatively primitive systems would be indisputably safer than the current situation, with inescapably flawed humans hurtling down the road willy-nilly.

 

[Midpack]

That's a good point, and it occurred to me that the answer is probably automatic public transit. This, to me, seems like low-hanging fruit for self-driving vehicles. Busses take the same routes, all the time, and don't need to worry about navigating poorly-delimited dirt roads, or deal with multi-lane, high speed highway traffic. They simply trundle along city streets, often in their own dedicated lanes, stopping at periodic intervals.

If you eliminated the driver, you'd eliminate a substantial part of the cost. That savings could be passed on to the users, giving the poor an accessible mode of transportation. Moreover, automating the drivers would feasibly reduce accidents, which would correspondingly reduce lawsuits, and again, costs for the municipality.

This is but one of many interesting dilemmas with the transition to self-driving. I don't know what the answer is, your vision may be exactly right.

I can easily imagine an individual moving from today's cars, to a car with more advanced driving assistance features (developing trust/relinquishing control gradually - already underway), to a level 4-5 car that the driver can override as needed (further developing trust/relinquishing control gradually), and finally to a fully autonomous car with no driver-steering wheel-pedals. But economics will probably not allow the gradual transition for most of us.

It's harder to imagine a driver moving from a highly manual car, or even a level 2-3, directly to a driverless public or Uber. Though routes are very predictable, other cars, pedestrians, weather, construction, etc. are not. I guess a few brave souls may have to show the rest of us the way.

So far, Waymo seems to includes human drivers to monitor the car as a given, despite the Firefly.

 

[Texas Proud]

Speaking of TVs, I am amazed how cheap they are. And yes, in a way they are really "cheap" as in not lasting that long.

I bought my first HDTV in 2003. Visitors ooh and aah'ed when they saw it, but now you would not think it is really HD when comparing to the "real" ones. But the doggone thing still works. My brother is now on his 4th TV. Each time I visit him, he has a new one. They don't last. He buys a new one about every 3 years, and it bothers him because he is frugal, like me.

Maybe he should try a different brand... even todays TVs should last more than 3 years... and 4 bad ones in a row!!!

 

[ERD50]

Maybe he should try a different brand... even todays TVs should last more than 3 years... and 4 bad ones in a row!!!

We bought our first HDTV 10 years ago. Used daily, still working fine. Since then, we've bought several more, bought them as gifts for family, helped some family choose them. About a dozen TVs overall, and not a single failure that I'm aware of. Most of them were Vizio, through Costco, a 'value' brand, not high end.

-ERD50

 

[JoeWras]

I AM NOT SAYING LEVEL 4 of 5 CARS ARE RIGHT AROUND THE CORNER. That last 5-10% is where the difficulty lies. And like others, I expect first gen level 4-5 cars will be quite expensive, out of reach for most. We may see them in a few years, but I think it may be 30 years before they're very common.

That clarifies, and on this, I agree.

Tell you what... This morning, on my commute, I saw some "human driven cars" do some crazy things that is turning me into someone who would like to see level 4 sooner than later. It can be madness out there!

One of my concerns is the switch-over years. Maybe some of the dangerous expressways where people are passing on the shoulder at 25 mph over limit (what I saw this morning) can be limited to level 4 or above cars. All others disallowed. Or at least create exclusive SDC lanes.

But that's a societal issue that will be typically slow. It involves law making and funds for road improvements. Any road improvement around here takes 5 years minimum. Heck, they are finally adding some metering lanes on ramps here, and it has been over 5 years since conception to implementation. Just a simple metering ramp.

 

[NW-Bound]

Kombat, thanks for the corrections. As mentioned I do not know much about how the 777 is set up, and the NTSB report does not describe the mode sequencing in the same details as you did. When the report said the autothrottle was in HOLD mode, I assumed that it remained on, waiting for a GA button switch from the pilots in case it was needed. Some systems work that way.

The mode sequencing and reversion you described is quite a bit different from what I worked on. The operation of each aircraft is different. Hence pilots cannot just jump on an aircraft and fly it. They have to be trained.

And this brings up a point that I have tried to make, here and also in another thread, that semi-automated cars are getting to the point where one cannot just hop on and drive off. There is no standardized way that one can rate their capabilities. Just saying this car is a level 2 does not mean much. One car can see bicyclists, the other does not. One can detect crossing semi-trailers, the other does not. What can a driver expect to anticipate, so he can initiate a take-over before it is too late? These level-2 cars are not interchangeable. Yet, when I bring this up in another thread, a poster gets upset.

About pilots not messing with "stuff", I meant to say that they are formally trained to not do that. They are told that they would kill themselves and their passengers. And so, if we are releasing self-driving cars that are less than level-4 to the general public, we will have a bigger problem. Tesla owners are somewhat of a more selected bunch, and want to play with high tech. They would tend to be males, I guess. Many people, like my wife, my mother, would not care to explore the capabilities of a level-2 car to see what it can do. They would get frustrated, and in the worst case, get themselves or bystanders hurt when they do not know that they should have taken over. Yet, they are responsible if anything bad happens, because the car makers say so.

About me portraying the aircraft as having everything fault tolerant, no. I did describe somewhere how certain functions that are critical need to have more redundancy than the ones that are not. That has to be defined formally, and enforced. And people still cheat.

The above is about operating errors, but I also talked of design errors in a post somewhere, where engineers make a mistake despite our best effort. I mentioned that we show, by analysis and testing, that the possibility of a crash is 1 in a billion (not counting human operating errors), and yet we still have design errors. But we did do that hard homework.

I somehow convey a picture that things are perfect in the commercial aviation. That was not my intent. What I try to say is that if we did not take these measures, and make the aircraft so expensive, things would be A LOT WORSE.

But whenever I bring this up to show level-4 cars are tough and expensive, and level-2 and 3 are easier but have a different set of problems, people get upset. And whenever I brought up aviation as an example of how difficult it is, I was called a smart-ass!

 

[NW-Bound]

It is astonishing how much progress several organizations have made. Here's a video (below) of a Tesla in level 5 mode, and Waymo has far more capable vehicles...

What that video and Waymo have done demonstrate what is necessary for a level 4. It is not a proof of all that is required. They probably qualify for a level 3.

PS. Even Waymo says its target is a level 4.

 

[ERD50]

...

Tell you what... This morning, on my commute, I saw some "human driven cars" do some crazy things that is turning me into someone who would like to see level 4 sooner than later. It can be madness out there! ...

I definitely want to see something that helps protect us against poor drivers, and generally good drivers that have the occasional lapse.

I just don't think that SDC is the best path to get there. I want something as soon as possible, I don't want to wait many years for near-perfection. And I think driver assistance along with driver engagement is that path.

Although, even a high level of SDC probably won't protect us against the intentionally crazy driver. From what I've gathered, if the driver wants to take over, the system lets them. So if that idiot wants to pass on the shoulder, he gets to do it, right?

-ERD50

 

[NW-Bound]

From what I've gathered, if the driver wants to take over, the system lets them. So if that idiot wants to pass on the shoulder, he gets to do it, right?

-ERD50

Thank you. That's level 4. You need level 5, which Waymo is not targeting right now.

 

[Totoro]

I read that in Europe, there are cameras that capture your license plate at two different points, say 5 or 10 miles apart, and compare the times you pass them. If your computed average speed is higher than the speed limit, bingo!

They are everywhere in the Netherlands on highways. The cameras are at every onramp and offramp. Took a while and several thousands in fines to get used to . I like the fairness and clarity, but very much dislike the sometimes unreasonable limits.

A speed limit of below 50mph (80 kph) on an empty road with four lanes per direction, who came up with that? Especially when on a similar road a bit up ahead they increased speed limits up to 130kph (80 mph).

The German variable limits are the best solution: they get adapted based on traffic conditions. When the skies are clear and one is away from the city, no speed limit at all. While it can feel dangerous (if you drive on the left, you will have a Porsche flying past at 150mph pretty often) the road safety track record there is pretty great.

 

[NW-Bound]

They are everywhere in the Netherlands on highways. The cameras are at every onramp and offramp. Took a while and several thousands in fines to get used to . I like the fairness and clarity, but very much dislike the sometimes unreasonable limits.

A speed limit of below 50mph (80 kph) on an empty road with four lanes per direction, who came up with that? Especially when on a similar road a bit up ahead they increased speed limits up to 130kph (80 mph).

The German variable limits are the best solution: they get adapted based on traffic conditions. When the skies are clear and one is away from the city, no speed limit at all. While it can feel dangerous (if you drive on the left, you will have a Porsche flying past at 150mph pretty often) the road safety track record there is pretty great.

Hey, what other countries have these smart cameras? I am not planning to drive through the Netherlands this time, but will definitely go through France, Germany, Luxembourg, Italy, Switzerland, and Belgium.

PS. By the way, I am getting old, man. The lower the speed limit, the more I like it. But I do try to get reasonably close to the speed limit, else people behind me get upset on country roads.

 

[JoeWras]

I definitely want to see something that helps protect us against poor drivers, and generally good drivers that have the occasional lapse.

I just don't think that SDC is the best path to get there. I want something as soon as possible, I don't want to wait many years for near-perfection. And I think driver assistance along with driver engagement is that path.

Although, even a high level of SDC probably won't protect us against the intentionally crazy driver. From what I've gathered, if the driver wants to take over, the system lets them. So if that idiot wants to pass on the shoulder, he gets to do it, right?

-ERD50

I don't think I expressed it correctly. I'm trying to not write too much.

What I was hoping was that some expressways (or more likely, lanes) could be designated as "level 4 and above only." Any intentional disengagement a would require review to assure it was for safety reasons only with appropriate fines, etc. These lanes would be shutdown during adverse weather events.

Perhaps current express lanes on many highways can be the first to try this.

I'm trying to compromise here for the phase in. I don't like big brother either, but this stuff is all about big brother, so if you are in these lanes and you break protocol, you have explaining to do.

 

[Totoro]

I definitely want to see something that helps protect us against poor drivers, and generally good drivers that have the occasional lapse.

Don't know how it is in the US, but what boggles my mind is that today you generally only have to demonstrate capable driving once in your lifetime. After that, no-one will check whether you are a poor driver and major disaster waiting to happen. One is presumed capable. Insane.

Case in point: My grandmother has never driven a car at 86 years old, yet has a drivers license and is legally allowed to take a 4000+ lbs death machine on a public road.

Sadly, driving a car within the legally enforced rules has limited bearing on actual driver capability. There are people routinely treating the sidewalk as extended lane capacity that never get any citation.

If somehow we are willing to use current tech advances to open that discussion and do something about it, we'll be very far ahead already.

Imagine if your level 2 or 3 car signals a central evaluation center that you just hit a curb, with associated pictures. Or that you are dangerously tailgating, or slaloming (drunk driving). I hate big brother as much as the next person, it would save many lives I think.

 

[ERD50]

I don't think I expressed it correctly. I'm trying to not write too much.

What I was hoping was that some expressways (or more likely, lanes) could be designated as "level 4 and above only." Any intentional disengagement a would require review to assure it was for safety reasons only with appropriate fines, etc. These lanes would be shutdown during adverse weather events.

Perhaps current express lanes on many highways can be the first to try this.

I'm trying to compromise here for the phase in. I don't like big brother either, but this stuff is all about big brother, so if you are in these lanes and you break protocol, you have explaining to do.

I definitely agree that limiting the autonomy to specific sections of roadways is a very practical step in the development. A designated strip of highway, especially if it was designated "SDC only", could reduce the complexity by at least an order of magnitude, just far fewer scenarios to deal with. No bikes, pedestrians, no cross traffic, divided lanes, etc. And even that could be temporarily shut down if snow or some other condition would make the SDC less capable.

-ERD50

 

[Totoro]

Hey, what other countries have these smart cameras? I am not planning to drive through the Netherlands this time, but will definitely go through France, Germany, Luxembourg, Italy, Switzerland, and Belgium.

France has them mostly on N-roads (national roads) and secondary routes. Germany just opened up their first section in 2016. Luxembourg none as far as I know (it's so small anyway ). Italy (Tutor) everywhere on the major highways. Switzerland in some locations like tunnels. Belgium is expanding quickly. Started with the tunnels around Brussels and highway around Gent.

If you want a more detailed overview you might want to try and toss this in Google Translate - also references therein. Happy to translate (from Dutch or French) any specifics.
https://www.wegenwiki.nl/Trajectcontrole

There is one aspect potentially relevant for you: breaking the speed limit is one thing, getting the fine is another ... information sharing is still far from optimal across the EU.

If you break the limit in Switzerland, you'll get chased towards the end of the earth for example. But if you have an Italian car in Belgium you are pretty much immune.

 

[kombat]

I definitely agree that limiting the autonomy to specific sections of roadways is a very practical step in the development. A designated strip of highway, especially if it was designated "SDC only", could reduce the complexity by at least an order of magnitude, just far fewer scenarios to deal with.

I think it could also dramatically increase capacity. I don't think it's a stretch to say that a lane that can accommodate "n" regular cars at 60 mph could probably safely handle "2n" self-driving cars. The cars could travel much closer together, as each one could react instantaneously to any braking action of the car ahead of it.

Of course, not all cars have the same stopping distance, so a safety factor would have to be figured in, but given two cars of comparable size, weight, and tires, on the same road surface, one probably isn't going to be able to stop more than 10 - 15 feet shorter than the next, so a gap of 20 feet should be ample room. At 60 mph, that will feel dangerously close with a human driver, but should present no problem at all for computers.

 

[Midpack]

Even Waymo says its target is a level 4.

Thank you. That's level 4. You need level 5, which Waymo is not targeting right now.

Though level 5 has been their ultimate goal almost from the beginning. The Firefly is a level 5 vehicle, though strictly for development only. And level 4 is fully autonomous.

The Google division completed in 2015 its first fully self-driving ride on public roads in Austin, putting a blind passenger in the front seat of a steering-wheel-less car. Krafcik reiterated that the company's goal is still creating a Level 4 or Level 5 autonomous vehicle in which passengers can sit back and completely relinquish control of the car.

"As we’ve demonstrated with Steve’s first ride is our goal to get there without those controls," Krafick said. "We dont think the human should be asked to monitor the self-driving system. As it turns out... it is a requirement to have those controls."

Google to spin out self-driving car project in new company, Waymo - Business Insider

 

[NW-Bound]

Krafcik himself in a recent video talked of the Pacifica as having enough for a level 4.

Regarding the little Firefly, it demonstrates some of the capabilities that will be needed, but not all. For example, it has a single lidar, so I do not see how it can drive in rain and snow.

Yes, level 5 is the ultimate goal for everybody, but again I caught Krafcik mentioning level 4 is the immediate production goal.

PS. I have not seen any Waymo demo in bad weather. I am looking forward to see something with the Pacifica. They are testing in my neck of the wood, actually driving a lot right in my neighborhood. No snow here, and very infrequent rain. I am willing to bet they are just starting to test the performance of the new sensor suite first, and do not want distracting publicity as they have in California.

I encountered them several times, while walking around. It's residential, nearly zero traffic during work hours, and the only pedestrians are me and my wife.

 

[Midpack]

Krafcik himself in a recent video talked of the Pacifica as having enough for a level 4.

Regarding the little Firefly, it demonstrates some of the capabilities that will be needed, but not all. For example, it has a single lidar, so I do not see how it can drive in rain and snow.

Yes, level 5 is the ultimate goal for everybody, but again I caught Krafcik mentioning level 4 is the immediate production goal.

Level 4 is the interim goal, primarily due to regulations as Krafcik has noted. Not sure why that's important if level 5 is the ultimate goal. I haven't seen anyone suggest the evolution won't follow,a stepwise progression, that's just common sense, especially for something this complex.