For the younger guys
02-22-2020 | 09:40 AM
#171
Gets Weekends Off
Joined: Sep 2007
Posts: 1,506
Likes: 0
From: 777 Captain in Training
Yeah so that's a good question. That is how I started down this path, trying to "disprove" what I was researching was somehow pie in the sky aspirations. From what I have found, the vast majority of "it's not going to happen" articles are opinion and looking up the authors they have no real credibility to write what they do. If you find some worth reading, post here, and they can be discussed. There are lots of unknowns in how things will/could be overcome.
As far as avherald and screwing up goes, my point with that is that aviation is not as safe as it's made out to be in the media. Yes fatalities are low but you know as well as I do that the slop in the system keeps us out of the headlines. And automation continues to save our bacon.
There have most definitely been cases where human intervention prevented a mechanical failure from snowballing. They are few and far between. And it's unknown if future automation would have been able to handle those emergencies. But even if it couldn't, the benefits outweigh the risks, similar to how safe cars will become when computers can monitor and make safety decisions.
The secretary of the US Navy said in 2015 the F35 would be the last manned aircraft they bought. Then, a couple years later, they've backed off that as "pilotless doesn't mean unmanned, we still have operators on the ground" as the operating environment on a carrier deck was chaotic. So, maybe the logistics of running pilotless for an airline might be too tough to overcome, at least initially. Maybe there will still need to be someone directly responsible to coordinate all the nuances that happen before, during, and after a flight. And if you're going to have that person there, it would make sense that they have a traditional pilot background.
Soon enough we will take on more of an ATC role regarding aircraft separation. From cruise through landing, we will be assigned an aircraft to follow and make speed adjustments to do so. Approaches will be all RNAV like combining the arrival and approach (similar to how LAX does it now). The computer will make the speed adjustments, not the pilot. AA has scheduled trial runs of this starting next year. If you are interested in seeing what is coming, watch this short video:
https://youtu.be/1-MkAlwDtTg
What you're going to find is the automation is going to be doing a lot more of the flying and decision making.
As far as avherald and screwing up goes, my point with that is that aviation is not as safe as it's made out to be in the media. Yes fatalities are low but you know as well as I do that the slop in the system keeps us out of the headlines. And automation continues to save our bacon.
There have most definitely been cases where human intervention prevented a mechanical failure from snowballing. They are few and far between. And it's unknown if future automation would have been able to handle those emergencies. But even if it couldn't, the benefits outweigh the risks, similar to how safe cars will become when computers can monitor and make safety decisions.
The secretary of the US Navy said in 2015 the F35 would be the last manned aircraft they bought. Then, a couple years later, they've backed off that as "pilotless doesn't mean unmanned, we still have operators on the ground" as the operating environment on a carrier deck was chaotic. So, maybe the logistics of running pilotless for an airline might be too tough to overcome, at least initially. Maybe there will still need to be someone directly responsible to coordinate all the nuances that happen before, during, and after a flight. And if you're going to have that person there, it would make sense that they have a traditional pilot background.
Soon enough we will take on more of an ATC role regarding aircraft separation. From cruise through landing, we will be assigned an aircraft to follow and make speed adjustments to do so. Approaches will be all RNAV like combining the arrival and approach (similar to how LAX does it now). The computer will make the speed adjustments, not the pilot. AA has scheduled trial runs of this starting next year. If you are interested in seeing what is coming, watch this short video:
https://youtu.be/1-MkAlwDtTg
What you're going to find is the automation is going to be doing a lot more of the flying and decision making.
https://airlinesafety.blog/2016/01/3...be-eliminated/
02-23-2020 | 10:43 AM
#172
Thread Starter
Gets Weekends Off
Joined: Mar 2014
Posts: 4,153
Likes: 341
- (a) this guy is a professional pilot (biased about his job)
- (b) I *believe* flies 777's for FedEx (freight may be the first to go) and
- (c) it's four years old, an enternity in today's machine and deep learning environment.
I think he also brushes over the logistical savings of having remote operators. Even if it was the same number of pilots employed, cost savings and reliability would still increase. No fatigue issues, no human factors issues like intoxication, or missing a commute, etc.
My guess is we'll see some combo form...ground based monitoring that has the ability to override what the system thinks it should do. AA has 15,000 pilots and 1000 aircraft. Let's say on average each aircraft operates at 12 hours per day, and you'd have a remote operator (pilot? Maybe or maybe not, depends on what the FAA says) monitoring it from push to block in.
1000 aircraft x 12 hours x 365 days a year = 4.4m man hours. 4.4m man hours / 2000 hours a year (ie 40 hour work week) is 2,200 "operators"/pilots. Boost it to 3,000 to account for sick calls, training, etc. So maybe there will still be some jobs...but man it would suck.
02-23-2020 | 11:05 AM
#173
Gets Weekends Off
Joined: Apr 2011
Posts: 3,517
Likes: 143
02-23-2020 | 11:13 AM
#174
Thread Starter
Gets Weekends Off
Joined: Mar 2014
Posts: 4,153
Likes: 341
I don’t think you need to look any further than the roll out of CPDLC. How long has it taken and we are only up to 3 centers now actively using it. The timeline is out to at least another year and a half before all centers are up and running with it.
Relatively speaking that’s basic technology yet NameUser is anticipating them to stand up a fully functional autonomous system guiding thousands of pilotless airliners around the country in 10 to 15 years.
The premise is laughable.
Relatively speaking that’s basic technology yet NameUser is anticipating them to stand up a fully functional autonomous system guiding thousands of pilotless airliners around the country in 10 to 15 years.
The premise is laughable.
So there are certainly different aspects of this.
- Pilot/operator monitoring of aircraft from the ground (I believe the tech exists today or will within the next few years to do so).
- Aircraft making decisions based on learned behavior such as when to divert, when to delay takeoff due to weather, when to go around, what to do if they experience an engine failure. The concept has been proven on a small scale with Garmin's autoland feature (basic level of executing scripts).
- Autonomous vehicles controlled by central computers and linked together. This will be quite a while off. Certainly the holy grail but getting all the partners together (government, airlines, manufacturers, etc) to decide on a standard and work together...yes a very long time away.
https://youtu.be/cF6QHTQqL7g
02-23-2020 | 03:27 PM
#175
Prime Minister/Moderator
Joined: Jan 2006
Posts: 45,167
Likes: 803
From: Engines Turn or People Swim
- Aircraft making decisions based on learned behavior such as when to divert, when to delay takeoff due to weather, when to go around, what to do if they experience an engine failure. The concept has been proven on a small scale with Garmin's autoland feature (basic level of executing scripts).
Active machine learning (non-deterministic AI) is not certifiable. As I've said before if they use machine-learning to develop an auto-pilot it will need to be frozen (ie learning disabled) and then tested exhaustively in the real world for years or decades before regulators might even consider certification. And if they find ANYTHING wrong, it's start all over from scratch.
- Autonomous vehicles controlled by central computers and linked together. This will be quite a while off. Certainly the holy grail but getting all the partners together (government, airlines, manufacturers, etc) to decide on a standard and work together...yes a very long time away.
This is one of my favorite videos about step 3 and how effective and efficient it can be. Imagine a human controller giving instructors to each of these drones to keep them positioned doing the figure 8's:
https://youtu.be/cF6QHTQqL7g
https://youtu.be/cF6QHTQqL7g
02-23-2020 | 06:56 PM
#176
Line Holder
Joined: Jun 2011
Posts: 527
Likes: 0
Using technology to control an aircraft is not a new phenomenon. The very first Autoland of an aircraft was accomplished in 1937!! Forms of autopilots have been used in commercial aviation since around this time as well. The Boeing Clipper flying boat fleets were using autopilot in cruise flight. And you could have pretty reliably flown a commercial jetliner from ATL-SEA without pilots on board 30-40 years ago.
But again, the two humans are up front for decision making and critical thinking ability. Heck, in the very recent past we've had two highly automated jetliners have to be dead-sticked into a river and a corn field. Both with no fatalities I might add.
02-24-2020 | 02:11 PM
#177
In a land of unicorns
Joined: Apr 2014
Posts: 7,072
Likes: 103
From: Whale FO
The tech exists right now. The problem is cost and reliability. To achieve reliability consistent with current 121 safety standards (which are quantified mathematically in regs) you'd need multiple redundant systems, almost assuredly both ground and space based. That would cost vast amounts, and only after you spend the money do you find out if it can deliver the needed reliability. And engineer and programmer proof is not enough... it has to be proof against offensive hacking and jamming as well. Because such a system would be an obvious attractive nuisance for malign actors.
Part 25 requires any catastrophic failure conditions (and the datalink/remote control system would very much fall into this) to be extremely improbable, which is 1 in a billion, or "are not anticipated to occur during the entire operational life of all airplanes of one type".
There simply isn't a system that can be reasonably thought of that could be quantified as such.
02-25-2020 | 08:25 PM
#178
Gets Weekends Off
Joined: Dec 2017
Posts: 2,014
Likes: 1
From: Retired NJA & AA
NTSB Report on Tesla Fatal Crash
This was a 2018 fatal accident where a Tesla driver was playing a video game on his phone and letting the car drive itself. Construction area, lane markings disappeared, autopilot decided to move to the lane it could see and hit a barrier in the process killing the driver.
https://www.theverge.com/2020/2/25/2...probable-cause
https://www.theverge.com/2020/2/25/2...probable-cause
02-26-2020 | 12:46 PM
#179
Thread Starter
Gets Weekends Off
Joined: Mar 2014
Posts: 4,153
Likes: 341
This.
Part 25 requires any catastrophic failure conditions (and the datalink/remote control system would very much fall into this) to be extremely improbable, which is 1 in a billion, or "are not anticipated to occur during the entire operational life of all airplanes of one type".
There simply isn't a system that can be reasonably thought of that could be quantified as such.
Part 25 requires any catastrophic failure conditions (and the datalink/remote control system would very much fall into this) to be extremely improbable, which is 1 in a billion, or "are not anticipated to occur during the entire operational life of all airplanes of one type".
There simply isn't a system that can be reasonably thought of that could be quantified as such.
I did watch a video where a Silicon Valley engineer was complying about the 10^9 and wanted 10^8 for operational reasons and felt the FAA was being "unreasonable".
02-26-2020 | 04:59 PM
#180
Prime Minister/Moderator
Joined: Jan 2006
Posts: 45,167
Likes: 803
From: Engines Turn or People Swim
Hmm, that is an interesting angle. I too cannot fathom a near term point in time where a datalink would not been required.
I did watch a video where a Silicon Valley engineer was complying about the 10^9 and wanted 10^8 for operational reasons and felt the FAA was being "unreasonable".
I did watch a video where a Silicon Valley engineer was complying about the 10^9 and wanted 10^8 for operational reasons and felt the FAA was being "unreasonable".
Also worth noting that 10^ -9 is the allowable chance of a serious accident per flight hour.
So it's system-cumulative, it does not mean every component needs that level of reliability, that can be mitigated with redundancy.
Multiple independent systems (ie parallel) can each have a lower reliability, with each backing the others up. This is why we have about six hydraulic pumps... hard to build a high temp, high rpm, high pressure pump with that kind of reliability.
But for remote piloting you'd need MULTIPLE comms system, and even then I'm not sure ANY known technology can achieve the needed reliability. How often do your HYD pumps "hiccup"? How often does your in-flight wifi hiccup?
Are you going to build six separate space and ground based systems? And you'd have to use different frequencies and protocols to mitigate hacking/jamming, can't just use six copies of the same system (like you can with hydraulic pumps). If it can be done technically, it most certainly cannot be done economically.
The way it will play out is anyone who seriously tries will soon be in the lobby of FAA HQ, hat in hand, looking for waivers and favors. Good luck with that: https://www.reuters.com/article/us-b...-idUSKBN20J2M3
Thread
Thread Starter
Forum
Replies
Last Post
