[johnbryanpeters]

This is one example; I don't believe that the problems are soluble:

Tesla driver blames Autopilot for crash, faces negligent driving charge

The driver reportedly told police he “must not have been paying attention.”

arstechnica.com

 

[Toshi]

Autopilot is not hands off. People are just idiots.

/me uses Autopilot responsibly

 

[Adventurous]

This is one example; I don't believe that the problems are soluble:

Tesla driver blames Autopilot for crash, faces negligent driving charge

The driver reportedly told police he “must not have been paying attention.”

arstechnica.com

No, those autopilot woes probably aren't soluble.

But I don't see any technological reason that can't be resolved. It'll just take time.

It would certainly be prudent for Tesla to rename their "autopilot" mode, as it is clearly not.

 

[Toshi]

A typical autopilot on an airplane requires oversight as well, no?

 

[Jm_]

A typical autopilot on an airplane requires oversight as well, no?

Yes, but, they are seeing the same issues as automation is becoming more and more comprehensive. The more automation, the more complacency, which causes a worse response from the humans when intervention is required, sometimes totally evading detection by the human. For human factors, it's more the simpler the automation, the better the human-results. The more comprehensive the automation, the less human interaction it requires and the human tends to have a much worse response to a crisis, but the problem becomes planning for "everything" via automation and when that unplanned event happens, the consequences are usually much more catastrophic, due to how much the human was cut out of the process.

Although not autopilot, this idea here is similar and this is a perfect example of not being able to take into account every situation...in this case, people died: https://www.avweb.com/aviation-news/fly-by-wire-helicopter-military-pilot-had-conflict-before-fatal-crash/

 

[iRider]

Yes, but, they are seeing the same issues as automation is becoming more and more comprehensive. The more automation, the more complacency, which causes a worse response from the humans when intervention is required, sometimes totally evading detection by the human. For human factors, it's more the simpler the automation, the better the human-results. The more comprehensive the automation, the less human interaction it requires and the human tends to have a much worse response to a crisis, but the problem becomes planning for "everything" via automation and when that unplanned event happens, the consequences are usually much more catastrophic, due to how much the human was cut out of the process.

Although not autopilot, this idea here is similar and this is a perfect example of not being able to take into account every situation...in this case, people died: https://www.avweb.com/aviation-news/fly-by-wire-helicopter-military-pilot-had-conflict-before-fatal-crash/

Isn't that the issue with the 737 Max?

 

[Jm_]

Isn't that the issue with the 737 Max?

That issue is complicated.

Due to wanting to keep costs down, both for materials/design, and certification process, they stayed with the classic 737, which required pushing the engines out further forward to fit the new larger engines. This created a nose up pitching moment at some high angles of attack, which is basically the exact opposite of what you want.

So to offset this, they put the MCAS system on it, which has been used successfully in other aircraft, except, they implemented it to react at a rate 4x of what they proposed and told pilots about. So rather than gently/smoothly pitching the nose forward, it reacted extreme. The reason for this is not clear, why it different from what they told everyone. Then, you can of course override the trim, but if the aircraft is pitching nose down and going very fast, as with the case of the two crashes, the only way to trim it back is counter intuitive. Due to the design of the system and aerodynamic loads, you have to push the nose FURTHER forward to unload the elevator, then you can put a bunch of nose-up trim in, but for human factors, that's the last thing you are going to be trying to do in a situation where you are screaming at the ground, even with a lot of training. Then of course there was the issue of not training the pilots on the new system (and the much faster rate of trimming) and not requiring flight training as part of the differences training for the new aircraft. Lastly, the system did not disconnect or post any warning when there were disagreeing angle-of-attack sensors. So when one sensor hits a bird or something, the system thinks it's at high angle of attack and automatically goes into this cycle.

It's a series of significant design errors, but those which seem that they could have been reasonably foreseen, as opposed to autopilot complacency, when you are monitoring the autopilot on approach or departure and it decides to turn the wrong way or fly into a mountain. The idea is that those are easily avoidable if the pilot is hand-flying, but in the case of MCAS, it intervenes no matter what. It is an example though of making the airplanes so complex that you can't control all of the situations that are possible to occur.

 

[OGRipper]

Personally I'd love to see roughly 80 - 90% of the drivers I encounter in self-driving cars. I'll take my chances with machine error as compared to their terrible, selfish, and distracted driving. Enforcement is a joke and far too many people are simply unwilling to put their phones down (or their cheeseburgers, vapes, etc.) and focus on the task at hand. The technology can't come fast enough as far as I'm concerned.

 

[maxyedor]

That issue is complicated.

Due to wanting to keep costs down, both for materials/design, and certification process, they stayed with the classic 737, which required pushing the engines out further forward to fit the new larger engines. This created a nose up pitching moment at some high angles of attack, which is basically the exact opposite of what you want.

So to offset this, they put the MCAS system on it, which has been used successfully in other aircraft, except, they implemented it to react at a rate 4x of what they proposed and told pilots about. So rather than gently/smoothly pitching the nose forward, it reacted extreme. The reason for this is not clear, why it different from what they told everyone. Then, you can of course override the trim, but if the aircraft is pitching nose down and going very fast, as with the case of the two crashes, the only way to trim it back is counter intuitive. Due to the design of the system and aerodynamic loads, you have to push the nose FURTHER forward to unload the elevator, then you can put a bunch of nose-up trim in, but for human factors, that's the last thing you are going to be trying to do in a situation where you are screaming at the ground, even with a lot of training. Then of course there was the issue of not training the pilots on the new system (and the much faster rate of trimming) and not requiring flight training as part of the differences training for the new aircraft. Lastly, the system did not disconnect or post any warning when there were disagreeing angle-of-attack sensors. So when one sensor hits a bird or something, the system thinks it's at high angle of attack and automatically goes into this cycle.

It's a series of significant design errors, but those which seem that they could have been reasonably foreseen, as opposed to autopilot complacency, when you are monitoring the autopilot on approach or departure and it decides to turn the wrong way or fly into a mountain. The idea is that those are easily avoidable if the pilot is hand-flying, but in the case of MCAS, it intervenes no matter what. It is an example though of making the airplanes so complex that you can't control all of the situations that are possible to occur.

Only thing I'd add to that explanation is that they screwed the pooch by programming the MCAS software to only rely on one AOA sensor. The plane can't tell a pilot about a disagreement in sensors if it doesn't know to look for a disagreement of sensors, all it knew was that according to it's data, the thing needed to climb, and climb hard. Had they used 3 sensors, it could have kicked out the bad data and flown just fine, and I believe that's the major software change they're getting recertified this week.

People also like to blame Boeing, and the FAA, but never the airlines who demanded the MAX and as a result MCAS be developed the way it was because they didn't want to retrain pilots. Now, Boeing maybe should have said no, but had they not programmed the system to rely on a single sensor, it's highly likely that the system would have worked flawlessly for the life of the 737 Max program. When the system gets good AOA data, it works just fine. This was all to avoid having to send pilots back into simulators to train on the MCAS system.


Self driving car tech will have the same pitfalls, and everybody should look t the MCAS issue and learn from it, because most systems currently in use are 1 bug strike away from utter failure. This exact issue can, and likely will crop up on self driving cars because people don't want $300k cars that have redundancy on every system. Relying on single sensor could be fine, as long as there's some reason check that the system can perform and fail safe when presented with a situation far outside of the normal parameters, MCAS software could have seen the elevator input from the pilot and compared it to what the computer thought should happen and thrown a red flag. That's where the biggest problem with how airlines and the FAA went about requiring and approving the Max comes in. They couldn't let pilots know the system existed, because then they'd have to retrain, and airlines didn't want to do that, Boeing thought they could pull it off, and the FAA allowed it all to happen. Your car doesn't have those regulatory issues.

You don't need to be certified to drive a Tesla or a Mercedes, so a fail mode can be created, and you can tell drivers about it without accidentally requiring they go drive simulators for several thousand bucks and hour, and it doesn't matter if every car does it similarly or not. You can just add a pop-up on the screen that says "shit's fucked, pull over", can't do that in a plane unless you let the pilot practice on it first. Hell, my truck has gone into limp mode on the freeway, no warning, just 75mph to 30mph, all because an O2 sensor had a hissy fit, no reason the same thing can't happen if your car gets confused by a bad repaving job.


As soon as the tech is more reliable than a human driver, it should be everywhere IMHO. The problem is people all think they're good drivers, in fact the worse they are, the better they think they are. Those people won't adopt the tech until it's perfect because they think they're already perfect. Nobody wants to be the corpse that results from a tech glitch that they otherwise could have avoided if they were at the helm, but ultimately more people will survive their commute if something similar in reliability to Tesla's autopilot becomes ubiquitous.

 

[Pesqueeb]

Personally I'd love to see roughly 80 - 90% of the drivers I encounter in self-driving cars. I'll take my chances with machine error as compared to their terrible, selfish, and distracted driving. Enforcement is a joke and far too many people are simply unwilling to put their phones down (or their cheeseburgers, vapes, etc.) and focus on the task at hand. The technology can't come fast enough as far as I'm concerned.

I got wiped out last thursday by a 76 year old woman who, if you believe her story, decided after waking up that morning suddenly blind in one eye, to spend the morning running errands behind the wheel of a 3500 pound car and only went to the doctor after she broadsided me in a cross walk, so she could be diagnosed as having a stroke.

I say, bring on the robot cars.

 

[Jm_]

People also like to blame Boeing, and the FAA, but never the airlines who demanded the MAX and as a result MCAS be developed the way it was because they didn't want to retrain pilots. Now, Boeing maybe should have said no, but had they not programmed the system to rely on a single sensor, it's highly likely that the system would have worked flawlessly for the life of the 737 Max program. When the system gets good AOA data, it works just fine. This was all to avoid having to send pilots back into simulators to train on the MCAS system.

For sure, but I still see Boeing as fully culpable there. It's wild that an airplane like the Max can have the same type rating as a 737-100 "classic", they are so different, like the Max 10 is roughly TWICE the weight of the original, it's just wild to think they are the "same". I don't think this is fully addressed anywhere, when a design "evolves" so far that it's not really the same as the beginning design, at least enough that it can't be called the same "type". More than a few companies have gotten this "handshake deal" to let their new version remain as the same "type" in the eyes of the FAA. For sure Boeing and the airlines were trying to avoid having to do flight training (in simulators), all resulting from trying to stretch out the 737 chassis.

 

[maxyedor]

For sure, but I still see Boeing as fully culpable there. It's wild that an airplane like the Max can have the same type rating as a 737-100 "classic", they are so different, like the Max 10 is roughly TWICE the weight of the original, it's just wild to think they are the "same". More than a few companies have gotten this "handshake deal" to let their new version remain as the same "type" in the eyes of the FAA. For sure Boeing and the airlines were trying to avoid having to do flight training (in simulators), all resulting from trying to stretch out the 737 chassis.

Is it possible for all parties to be fully culpable? 300% guilt!

Honestly, the 100 and 200 should barely have the same type rating, this is an issue that I think the FAA is guilty of that laid the groundwork for their other fuck-ups in regard to the Max, due to pressure from the airlines. Had somebody put their foot down 30-40 years ago, the Max wouldn't have been an issue because every major change should require a new type (sub-type?) rating beyond just the cursory glance at the manual. They're playing by the rules, but the rules are stupid, and since there were no adults in the room, nobody went to the FAA and told them to change the rules since it would really only effect Boeing in the near term as Airbus has newer airframe designs and hasn't run into this problem yet.


This is now an airplane nerd thread, GTFO Tesla nerds! Elon sucks.

 

[SkaredShtles]

I got wiped out last thursday by a 76 year old woman who, if you believe her story, decided after waking up that morning suddenly blind in one eye, to spend the morning running errands behind the wheel of a 3500 pound car and only went to the doctor after she broadsided me in a cross walk, so she could be diagnosed as having a stroke.

I say, bring on the robot cars.

I will 3rd this motion. And people will still scream when the occasional robot car gets into some situation where someone dies.

What those screamers will overlook is that instead of 38,000 people dying every year due to auto accidents, it'll be 1,500. 'Cause... people are dumb.

 

[Toshi]

Not sure if it was in the walls of text, but not only did MCAS dial in more degrees of control actuation than was originally proposed, it also would repeatedly actuate, its effects piling up until at max deflection.

 

[Jm_]

Not sure if it was in the walls of text, but not only did MCAS dial in more degrees of control actuation than was originally proposed, it also would repeatedly actuate, its effects piling up until at max deflection.

I think that was "normal" for the system, it would do it intermittently, as in in bursts, but it would keep doing it until the "high angle of attack" situation no longer presented itself. I don't think that was a significant issue. Stick shakers and stick pushers tend to operate like this. The real problem was the speed at which it did it got it to that full deflection real fast and that it was linked to one AOA sensor, even in the situation of an AOA disagree (one sensor going haywire). From a human factors engineering perspective, huge fails. Airbus has all kinds of protections designed into their flight envelope, so this is not really ground breaking...except they've been doing it to a much higher extent for much longer.

 

[Toshi]

The repeated actuation definitely was a problem in conjunction with the reading the one faulty AoA sensor, because it kept on acting on incorrect data, despite the pilots trying to override it (+/- being able to physically turn the trim wheel).

 

[johnbryanpeters]

I have come to like cruise control, and I like adaptive cruise control better.

The Sprinter has lane sensing (Active Lane Keeping Assist), beeps if you cross a line without signaling. It does not tell you which side you're off on. If you have reason to ignore the beep, it jams the brake, which is stupid as shit. I find that, when driving with the ALKA system, it requires a level of attention that tends to reduce overall situational awareness.

In heavy rain or snow, sensors tend to get occluded and all this stuff will shut down without warning. From the perspective of the driver, transition back to "unassisted" driving is jarring, especially in heavy traffic and/or shitty conditions.

Control settings are sufficiently complicated in all contemporary vehicles that I am familiar with that I doubt that less than 10% of drivers fully understand them, and I think that it's a problem.

There are reasons that we train pilots; I think that increasingly, if control systems continue in the direction that they're going, we're going to need to train drivers.

 

[Jm_]

I made adaptive cruise control the number 1 requirement for my current 2nd vehicle, that and remote start. I love it. Once you use it, it's damn hard to go back. I find there are a few different takes on it and it operates differently in those different vehicles. This one is a Hyundai, and it basically works exactly like the one did in my BMW, with the addition of optional lane-assist and inattentive alert. I don't like the lane-assist, so I keep it off, but the adaptive is a huge boon to me in any kind of traffic, like we regularly get on our highways due to summer rec. Secondary is the help in obscured conditions, like heavy rain and when it's cold and ice crystals are being throw up by all the vehicles. Sometimes it's enough to block vision totally, even if it's not really snowing. Both this and the BMW will stop and restart if the vehicle ahead moves in 3 seconds and then it's just pusthing a steering wheel button to re-engage from a stop. The only time I ever had it kick off with the BMW was when it was coated by heavy wet snow and the only really annoying thing if I remember is that in that situation it wouldn't even let me engage the normal cruise, but overall, I love the system. Brings you back to that complacency thing though, you start expecting other cars to slow down and stop if you engage the cruise. Did not care for the Subaru system, it was not stop-and-start and as you got real slow and real close to the vehicle ahead, it would "pop off" below a certain speed, at which point you'd start accelerating into the vehicle (at idle), which is the last thing you want when you are close to the vehicle ahead.

Airbus Completes Autonomous Flight Tests - Simple Flying

Airbus has today announced completion of its Autonomous Taxi, Take-Off, and Landing (ATTOL) project. The project has seen…

simpleflying.com

 

[Jm_]

The repeated actuation definitely was a problem in conjunction with the reading the one faulty AoA sensor, because it kept on acting on incorrect data, despite the pilots trying to override it (+/- being able to physically turn the trim wheel).

Yes, but if a pilot failed to correct, due to momentary incapacitation or some other issue (pilots do stupid shit), you would want the system to keep correcting until it was no longer at the high AOA. That's the whole point of the system, to get the nose down. The repeated actuation is not the issue IMO. Going off memory, I think it was supposed to be .5 units/sec and it was actually 2.5 units/sec. There are two big switches to override the system, but if the aerodynamic loads have increased too far, you have to use the "unload" technique. So in my opinion, it's that the speed at which it corrected was 4x faster than it was supposed to be and "advertised" as, not that it continues to correct. If the speed was slower, pilots would be able to cut it out before it reached the dangerous nose-down pitch. Transport category aircraft have a fairly narrow range of "trim" that they can fly/operate in compared to general aviation aircraft.

 

[dan-o]

I will continue to purchase the lowest tech vehicles available.

 

[eric strt6]

then there was the Asiana Crash at SFO. reliance and complacency on the technology and pilot inattention

Crash of Asiana Flight 214 Accident Report Summary

www.ntsb.gov

 

[jdcamb]

The Cruise Control in the Toyota works pretty awesome. I seldom drive on highways to qualify.. I turn off all the other shit off. Get out of my way please....

 

[eric strt6]

The only time I use cruise control is on wide open fairly straight interstates Like I5 going from SF to LA or I80 from Reno to Nebraska across desert and grass lands. I would be bored shitless with a self driving vehicle. I like seeing something and going I wonder whats over there? then making that detour. The matrix need not know my flight plan

 

[Pesqueeb]

then there was the Asiana Crash at SFO. reliance and complacency on the technology and pilot inattention

Crash of Asiana Flight 214 Accident Report Summary

www.ntsb.gov

Don't forget these assholes. 110% these guys were both asleep. Not that they would admit it. "Distracted", my ass.

Plane crazy: NWA flight misses MSP by 150 miles

Military jets stood by as NWA pilots, apparently distracted, didn't respond to controllers for 75 minutes.

m.startribune.com

 

[eric strt6]

Don't forget these assholes. 110% these guys were both asleep. Not that they would admit it. "Distracted", my ass.

Plane crazy: NWA flight misses MSP by 150 miles

Military jets stood by as NWA pilots, apparently distracted, didn't respond to controllers for 75 minutes.

m.startribune.com

City Pages

www.citypages.com

 

[Pesqueeb]

City Pages

www.citypages.com

"Distracted", my ass.

 

[stevew]

I will continue to purchase the lowest tech vehicles available.

i want a 85 range rover with manual windows and locks....

 

[eric strt6]

i want a 85 range rover with manual windows and locks....

72 FJ40 LandCruiser for me

 

[stevew]

72 FJ40 LandCruiser for me
View attachment 147056

i prefer the ambulance style rear doors of the later ones and disc brakes but i wouldn't kick that out of the driveway...

 

[eric strt6]

i prefer the ambulance style rear doors of the later ones and disc brakes but i wouldn't kick that out of the driveway...

That was the first vehicle I ever bought. never should have sold it. Most fun adventures in that FJ 250,000 miles of them

 

[stevew]

That was the first vehicle I ever bought. never should have sold it. Most fun adventures in that FJ 250,000 miles of them

i had 35's on mine and no power steering and it was the easiest car i ever owned to parallel park...

 

[eric strt6]

I ripped out the seats in mine and put in high back wrap around racing poly bucket seats with 5 point safety harness and a roll cage. Had the stock 3 on the floor, 155hp inline 6 , added KC offroad lights and a Warn 8000 winch on the front

 

[Jm_]

"Distracted", my ass.

Sleeping is pretty distracting. The old way they used to do this was to dial in the frequency of a navaid at the destination that would start beeping loudly when it came into range. Works most of the time, except when the navaid is out of service.

 

[CBJ]

I have come to like cruise control, and I like adaptive cruise control better.

The Sprinter has lane sensing (Active Lane Keeping Assist), beeps if you cross a line without signaling. It does not tell you which side you're off on. If you have reason to ignore the beep, it jams the brake, which is stupid as shit. I find that, when driving with the ALKA system, it requires a level of attention that tends to reduce overall situational awareness.

In heavy rain or snow, sensors tend to get occluded and all this stuff will shut down without warning. From the perspective of the driver, transition back to "unassisted" driving is jarring, especially in heavy traffic and/or shitty conditions.

Control settings are sufficiently complicated in all contemporary vehicles that I am familiar with that I doubt that less than 10% of drivers fully understand them, and I think that it's a problem.

There are reasons that we train pilots; I think that increasingly, if control systems continue in the direction that they're going, we're going to need to train drivers.

I have some of the same experience in the Volvo. If I do not jerk the wheel and let it know I am holding on it will beep and warm me its turning off. However if self driving is engaged and for some reason it can not drive due to missing lines or some other reason it will just shut off without warning me. I have spend considerable time learning how the system works and as the info from Volvo is lacking. We are still in early days and I am sure very few people will take the time I have taken to understand their cars.