Just look at the rollout for ADS-B In/Out if you're curious how difficult that might be to implement.
I'm not sure what you mean with "is not predictable" when you also admit that it's repeatable.
But AI isn't different than existing software systems either. Both will take an input from reality and take actions upon it.
Repeatable means you get the same output a second time, given the same input. Predictable means you can say in advance what it will do, and can then check the output against your prediction.
If you can't predict the outcome, you can't validate the process, and can't guarantee its performance.
If your AI has linear / continuous output, testing it should be no different than any other software.
Almost all neural network implementations have continuous outputs (ie the nodes in the output layer produce a value between 0 and 1). That doesn't change the above issue at all.
This is much less of an issue with traditional methods
skip the buzzword, why couldn’t a computer program of any kind helped avoid this
It doesn’t alert you to planes on the ground because it would never stop going off. How is the system supposed to know that the pilot approaching the threshold as you land is going to stop? It can’t so it would issue an immediate correction alert and you have to go around.
Sure it may avoid this situation, but how many aircraft lifted off and landed at airports in the same day? In America I’d hazard a guess at aground the tens of thousands. Any new system has to reduce the complexity or risks of flying and that’s very hard to do.
I’m not an aviation expert either but it’s assume that some form of system exists for this also in a more manual form.
From a layman's perspective, it replaces one primary degree (pilot-control coordination) with another (a technological solution) and delegates the staffers to supervisory roles. That is a risk reduction due to the increase of confirmations and the independence between the staff decision and the software's decision.
ATC should be wearing VR goggles, visualizing approach and takeoff routing as it maps to flown with machines spotting the dangers similarly but differently from TCAS.
Move fast and break things has no place in aerospace nor aviation, just rolling whatever fancy new tech is there is not done for reasons. And this behavior made aviation as safe as it is today.
1) VR itself can lead to spatial disorientation and will introduce its own control issues.
2) A significant percentage of people (~1 in 4) cannot use VR without motion sickness. This is independent of #1. Modern VR (Oculus etc.) at first claimed to be better, but guess what, plenty of people still get sick. Sinus congestion can cause this even in those who are otherwise tolerant.
3) Position reporting of planes today is nowhere near accurate, reliable, or real-time enough to present a whole picture of runway ops. This is fixable with enough $$$...but who pays?
4) Even if these and other obstacles were cleared, I suspect the new "VR ops" procedures from the FAA would take years to be developed and approved, without some kind of urgent mandate.
My gut feeling is that we'll have automatic ground traffic control at major airports by the time the necessary systems are in place, and skip the goggled humans entirely.
You have been repeatedly dismissing people in this thread, but HN is about being curious. “Leave it to the professionals” is neither satisfying nor interesting.
Throwing ideas out to improve things, without properly thinking about the the root causes for incidents, not waiting for official investigations to be run, and all of that based on some audio recordings and headlines, or news coverage at best, is neither curious nor allows people to learn something. So yes, at a certain point, leave it to the professionals (whom else would one leave it to anyways?) is exactly the right thing to do. And maybe listen to people with more knowledge on a subject (probably not me in that case so).
Insisting on pre-formed "layman" opinions about something as peculiar as aviation, or aerospace, is the opposite of curiosity.
If you don’t agree with something someone posts, contribute to the discussion by explaining the issues. Why shouldn’t the ATC be sacked? Safety culture. Why shouldn’t they install ground radar? Complexity. Etc.
Appeals to authority are the worst kind of arguments because they don’t help us to understand what to expect from them.
Just think of it as CRM for the internet. :)
Humans in the loop are a SPoF if they're there:
1. Solely to read information over a lossy, slow medium like analog radio. Digital data between ground and air systems should be the primary means of comms with voice radio as a backup channel for clarification and stating intent.
2. To flawlessly plan and avoid collisions between dozens of objects moving at high speed in 4 dimensions. Never going to happen. These should be done and verified mechanically, continuously.
Humans should be guiding and assisting mechanical, reliable automation of decision-making rather than playing telephone or doing long division on paper when calculators exist.
You're sure about that?
That said, it makes the "moving in" redundant and is shy a few dimensions if you want to go full descriptive phase space diagraming given velocity and acceleration in each spatial dimension are missing.
A Four-Dimensional Space-Time Automatic Obstacle Avoidance Trajectory Planning Method for Multi-UAV Cooperative Formation Flight
or any number of other similar papers.
Many prefer to think of an objects path as a trajectory in space-time (four dmensions) and for two ojects to "collide" their paths must coincide within that 4-D space within an Epsilon for some value of WTF.
Other common examples of working in N dimensional spaces for values of N bigger than someones mother include: https://en.wikipedia.org/wiki/Phase_space
Fun fact, now I can claim to plan in the 4th dimension by using lead times in supply chain planning!
Another plane (P2)'s travel through various {X,Y,Z} positions at various times forms another path in R^4.
If those paths come close to each other then P1 and P2 are close in both space and time - ie. they are very close to a collision.
Collision detection and avoidance is problem laid out and (hopefully) solved in an R^4 euclidean space.
(at the very least - throw in some more independant variables that parameterise motion and you've got a higher order puzzle
Eg: Collision avoidance for two robot arms with 6 or 7 degrees of freedom each is a maze solving puzze in 12 or 14 dimensions).
The path of the plane is a static curve in 4-dimensional space, yes.
But the plane is not located at any point in the 4-dimensional space, and the position in 4-dimensional space that it doesn't have is not changing over time. Both of those things are required before you can describe the plane as "moving" within the space.
There is no secret backup time that will allow you to track the plane's hypothetical motion along an explicit time dimension. That's not a thing.
I confess. I literally had no idea what you were intending to convey with those two sentences so I restated alternatively what I intended to convey in the hope it might make clear my position (if that was an issue for you) or that I might learn more from your response.
> Where do you think you're contradicting me?
That's not a thought that I thunk. Therefore I have no response.
> But the plane is not located at any point in the 4-dimensional space
Every point along the 4D path trajectory of the plane in {X,Y,Z,T} is a literal {X,Y,Z} location of that plane at time T.
> Both of those things are required before you can describe the plane as "moving" within the space.
I certainly did not describe the plane as "moving" within R^4.
> There is no secret backup time that will allow you to track the plane's hypothetical motion along an explicit time dimension.
I utterly fail to understand what you intend to convey here.
Although I note that the actual (not hypothetical) velocity of the plane projected onto the time axis is very likely to be on the order of approximately one second per second.
Regardless, I think it's too early to lay definitive blame until the NTSB report comes out.
They spent a minute extra before starting their takeoff roll; it wouldn't have hurt to wait an additional 5 minutes. Commercial flights build in a generous margin for ground ops these days. I doubt this was a factor in this specific case.
> maybe lost them their place in the queue, etc.
They were at the runway already, there's no way for them to have lost their place in the queue for that runway 18L [1].
[1] https://upload.wikimedia.org/wikipedia/commons/a/a9/AUSFAA.p...
Sure there is, enter the runway at bravo, right turn foxtrot, left on alpha and join the conga line. Happens all the time at places like JFK (admittedly Austin is not JFK).
> Sure there is, enter the runway at bravo, right turn foxtrot, left on alpha and join the conga line. Happens all the time at places like JFK (admittedly Austin is not JFK).
This routing appears to require the SouthWest aircraft to enter the active runway with FedEx Heavy on Final to that end, presumably the problem we're trying to avoid ...
In both cases, SW would be well within their rights to say "unable to comply".
More will surely be revealed in the follow up investigation and right now it's all armchair piloting. But I too am a little surprised the Southwest flight took its sweet time to do anything and was seemingly unaware of what was going despite being explicitly informed.
To your point though, we also don’t know what was going on in the cockpit of the 737. Even if they were aware of how close the FedEx jet was, it’s entirely possible something went wrong that caused them to delay the takeoff roll (assuming it was delayed).
There are plenty of things that could have gone wrong on the ground. Separating traffic is the ATCs job and clearly there was a loss of separation. Very hard for me to see how this could be Southwest’s problem particularly since they weren’t yet airborne.
Because you have unions protecting the manual, error-prone job of human operators.
As much as this frustrates me, it has always been like that, and always will be like that.
So the tower controllers had responsibility for separation. Here are the current FAA rules on separation between departing and arriving aircraft using the same runway. See section 3-1-3 of [3].
This is the rule: "Separate an arriving aircraft from another aircraft using the same runway by ensuring that the arriving aircraft does not cross the landing threshold until one of the following conditions exists .... The other aircraft has departed and crossed the runway end or turned to avert any conflict. If you can determine distances by reference to suitable landmarks, the other aircraft needs only be airborne if the following minimum distance exists between aircraft: ...when either is a Category III aircraft- 6,000 feet."
The trouble is, the controller can apparently say "Cleared to land" when they expect that departing aircraft will be airborne and at least 6,000 feet down the runway, and thus out of the way, before the incoming aircraft crosses the landing threshold. Not when it actually is in the air and out of the conflict zone. Or at least in this case, the controller did. This is apparently called "reduced runway separation". The intent is to increase traffic capacity.
But this was in fog. Tower probably could not see the departing aircraft, and they did not apparently have a sensor system to compensate for that.
[1] https://skybrary.aero/articles/autonomous-runway-incursion-w...
[2] https://skybrary.aero/articles/runway-status-lights-rwsl
[3] https://www.faa.gov/air_traffic/publications/atpubs/atc_html...
What scares me about this is that I have no idea how common or uncommon this is, and what other airports are in a similar situation.
And that professionalism seems to have failed in parts in this circumstance. (And has been well noted to have failed in many other circumstances with often tragic consequences.)
All that aside, I still want to understand why the Austin airport doesn't have ARIS? Cost/benefit? Not mandated by the FAA?
I'm afraid that "shush, and leave it to the professionals" isn't a very compelling argument.
For fatal accidents to happen in aviation and aerspace, more than one thing has to go wrong at the same time. Single, isolated failures are no longer sufficient to cause serious accidents.
That's just luck, not a system that's working as intended.
However, the plane component of the system failed, because the wings aren't supposed to fall off.
When you’re up to the last “hole” in the cheese, that’s a failure.
Then i also know horror stories from someone thrustworthy, about surgeons throwing extracted body parts after nurses in high stress situations. No, i will not name names, but life and death situations need a perfect team and you dont always got that perfect team.
And surgery quality being dependant on the actual execution of the surgery (chief surgeon is not automatic the best individual contributor to surgery).
You want the most hours on the knife guy for best outcome, not some administrator with white coat.
Also you want to be near univesity clinics, they are spearheading new stuff, the more you get away from them, the operation technique is depending on the age of the doctors education and the hospitals financials to keep up to date, send doctors to regular re-trainings and try new things. Then there are lawsuits, they keep the people responsible, but also risk averse.
So if you are in a high risk situation, in a lawsuit rich country you are fucked, cause nobody will try the dice throw when worst comes to worst.
Trust in systems working, is upkept by untrusting agents.
So like in capitalism, were the price aware negotiate the deals down for the crowd who does not want to haggle.
Finally: The infrastructure & hierarchy of hospitals (in europe), as we know them today, emerged largely from the two world wars were lots of nurses and doctors were integrated into army structures. Which means it suffers from the same inefficency and structural problems those armies back then suffered from. Now you have the MBA process micro optimization on top of that, so i wouldnt trust that machine blindly..
"What follows is an account of how Martha was allowed to die, but also what happens when you have blind faith in doctors – and learn too late what you should have known to save your child’s life. What I learned, I now want everyone to know. In a small way, I hope Martha’s story might change how some people think about healthcare; it might even save a life."
Allow me to chime in and respectfully disagree with this sentiment and the metaphor.
As the saying goes - the aviation (just as the automotive) industry saves lives (as in "advances security") one accident at the time. It would make sense that close calls like this Austin near mass casualty event should contribute to the future safety as well. In that context, it's absolutely legitimate to be suspicious and ask safety-related questions, including in a HN comment or in real life - instead of shutting someone down.
Speaking of the developer-OS metaphor - any non-rookie developer should be aware of not only the security and vulnerability of one's own software but also of the security and vulnerability of the underlying OS, infrastructure and even hardware. The number of building blocks gets larger by the day and nearly each building block is becoming increasingly complex. Yes, there are professionals working on each those blocks yet there are new CVEs and associated attacks all the time (incl. ransomware). If we add 0-days into consideration (a.k.a. "the unknown unknowns" in the software context) IMHO we should be able to conclude that the used developer-OS metaphor is not helpful.
The older and more experienced we all become, the more should we be cognizant of the potential risks (not only in our particular industry niche) and we should welcome and consider a normal, widely accepted practice to challenge the status quo and pose questions that should overall increase the number of brains and eyeballs on the problem and (unknown) unknowns - be it vulnerabilities or security risks, especially to our bare lives.
When the experts in the field have successfully prevented all accidents and near misses, maybe HN commenters will stop providing their input.
There is such a thing as tunnel vision in a field you have been in too long.
In other words, aviation regulation exists to serve the interests of the people.
If the airport or airline industry is trying to cut corners or save money at the expense of the safety of passengers, the interests of other air users, or the people who live under flight paths, items the job of congress to grant the FAA the authority needed to stop that - or to ensure that it is using its already-granted authority effectively to that end.
And ultimately the only check on whether or not that is happening is the electorate’s oversight of its congresspeople.
So yes, you can take an interest in how the airline industry is regulated, as a passenger or a random person who has planes fly over your head from time to time, because you have a say in making sure it’s not captured by industry and allowed to compromise your safety.
They designate the class of an airport not by the number of people living in the metro area around it, but based on the complexity of the airspace and the density of the traffic.
Austin probably isn’t a class b because while it may be an airport for a large city it doesn’t see a ton of traffic as it isn’t a huge hub. The regional hubs in that area are DFW and Houston and are way more traffic.
It’s also worth noting that making a this a class B airport wouldn’t have changed anything in terms of how the arrivals and departures are controlled at the runway level.
In spite of what happened with Boeing, the FAA is usually very good at airspace safety.
That’s WHY this incident is notable.
Lobbying can be good, of course! Airports and airlines likely have good knowledge about how regulations impact their operations!
But it’s the job of Congress in theory to represent their constituents’ interests - not merely their constituents who own airline stock or work for airport ground handling companies but all their constituents - to make sure that the framework the FAA correctly takes into account the competing interests of ensuring an efficient and high capacity aviation industry, as well as making sure that passenger safety is maintained.
As you say, the FAA has a good record of doing this.
But individuals saying ‘well I’m sure the industry knows best’ is precisely what leads to regulatory capture and that leads to the situation where airports are permitted to accept heavy automated landings under conditions where maybe they really ought not to without additional ground safety equipment in place.
Airlines are generally the ones asking for these safety upgrades before anyone else. They would love if every airport had Cat. III approaches on every runway, and there were designated parallel runways at every airport.
Airlines like higher classed airports since it frequently means the ability to handle more traffic safely, and better service from ATC.
Opposition to airspace changes are ALWAYS nimby's. ALWAYS. Go look at the public comments for any proposed change.
When I visited I remember thinking ‘how the hell does this place handle the influx for south-by?’
KSTL (St. Louis) and KMCI (Kansas City) are Class B, and at 10M and 7M carry fewer passengers than KAUS at 13M annually and have half the number of operations annually [1] [2].
There are many other airports on the List of Class B airports with smaller operations than Austin. [3]
[1] https://en.wikipedia.org/wiki/St._Louis_Lambert_Internationa... [2] https://en.wikipedia.org/wiki/Kansas_City_International_Airp... [3] [1] https://en.wikipedia.org/wiki/List_of_Class_B_airports_in_th...
KSTL and KMCI were formal airline hubs and are still Class B most likely due to legacy assignments. It's a lot harder for the FAA to "demote" a piece of airspace vs "promote" one.
I'm curious as to what you think upgrading KAUS to Class B would exactly accomplish, especially in this scenario? This was a runway localized mistake by the controller. Also, the word in the ATC community is this tower controller had been moved around the NAS multiple times as a problem child.
KSAT's airspace (not the airport, the airspace) farther south is FAR busier than KAUS and is considered some of the busiest airspace in the county due to multiple military fields in the area, extensive military training, etc. And it's also Class C airspace.
Under certain circumstances > 220k annual ops qualifies an airport for Class B status [1]. Austin's 2022 vs 2021 traffic is significantly higher [2]. When you include KEDC (Austin Exec Airport) and KGTU (Georgetown Municipal), the traffic certainly appears to exceeds the thresholds in question [3][4][5]. I didn't bother to look through the other airports within the area.
> I'm curious as to what you think upgrading KAUS to Class B would exactly accomplish, especially in this scenario? This was a runway localized mistake by the controller. Also, the word in the ATC community is this tower controller had been moved around the NAS multiple times as a problem child.
Designation as a Class B changes the care with which the airspace is managed and more importantly, staffed [6][7]. Perhaps the problem ATC is fine for a less busy area, perhaps the Tower staffing could use a little more oomph in the morning hours. There might also be additional RWLS and ARIWS requirements for Class B airports (needs regulatory research - maybe ChatGPT can give us pointers since this is all supposed to be public info!).
> KSAT's airspace (not the airport, the airspace) farther south is FAR busier than KAUS and is considered some of the busiest airspace in the county due to multiple military fields in the area, extensive military training, etc. And it's also Class C airspace.
The FAA Sectional for KSAT (San Antonio) shows that there are multiple Special Use Airspaces (Randolph, Laughlin, Crystal/North, Kingsville MOAs) to separate the civilian traffic from military airspace, and the military has their own controllers [8]. Caution: 111MB file.
[1] https://www.faa.gov/air_traffic/publications/atpubs/pham_htm... [2] https://www.austintexas.gov/news/november-2022-passenger-car... [3] https://www.gcr1.com/5010ReportRouter/EDC.pdf [4] https://www.gcr1.com/5010ReportRouter/GTU.pdf [5] https://en.wikipedia.org/wiki/Georgetown_Municipal_Airport [6] https://www.faa.gov/air_traffic/publications/atpubs/pham_htm... [7] https://www.faa.gov/air_traffic/publications/atpubs/pham_htm... [8] https://aeronav.faa.gov/visual/02-23-2023/PDFs/San_Antonio.p...
KAUS might be the most classic example of Class C airspace in existence. They only have 3 App/Dep positions with a spare 4th, and the top of the charlie is 4500 feet. I can overfly that in a light sport.
Austin airspace isn't nearly as busy as you make it out to be.
And in the context of this incident, arguing that KAUS needs to be upgraded to a Class B is kind of non-sequitur. A class B wouldn't solve anything except to make the area airspace wildly more restrictive.
> KAUS might be the most classic example of Class C airspace in existence. They only have 3 App/Dep positions with a spare 4th, and the top of the charlie is 4500 feet. I can overfly that in a light sport.
> Austin airspace isn't nearly as busy as you make it out to be.
Fair enough, local knowledge FTW. I made a suggestion - "it might be time to upgrade to Class B" as a systematic solution for an airspace that appears to be getting busier. It certainly appears to meet the FAA criteria based purely on stats.
> And in the context of this incident, arguing that KAUS needs to be upgraded to a Class B is kind of non-sequitur. A class B wouldn't solve anything except to make the area airspace wildly more restrictive.
You must have missed the part where I mentioned that Class B designation comes with staffing changes for ATCs which would presumably positively impact such occurrences :-) Maybe the alternative is better training, better operations in mornings, better rules around ATC IFR ground ops. 140+ dead people is 140+ dead people. Let's see what the NTSB says.
Also, with the plans to expand the airport with more gates, I'm guessing changes are coming to that airspace regardless [1][2]. Though the additional 17C-35C runway appears to be past 2037 [2] Pg 17, [3].
> > I fly in this airspace. (KAUS and KSAT and KDFW)
If you're flying in and out of KDFW you should have no trouble with a future Class B KAUS :-)
[1] https://www.austintexas.gov/department/aus-master-plan [2] https://www.austintexas.gov/sites/default/files/images/Airpo... [3] https://www.austintexas.gov/sites/default/files/images/Airpo...
Given my experience navigating the already PITA Bravo shelves in DFW (don't bust through those without an explicit clearance) and/or getting a Bravo clearance which adds to significant pilot workload, I'll take just flying over the charlie at 5500 and taking it easy while on flight following :) Also remember that Bravo airspace traditionally extends to 10,000 feet or more, so the considerations there for VFR traffic are significant.
Some bravo clearances are easy: "Cleared into the Bravo, direct to KXYZ, 4500 feet", others are "Expect vectors" and now you are flying around under direct control of a controller vectoring you around numerous aircraft for 45 minutes or more... yuck (or fun).
At this day and age it is bewildering to see all of this running on human communication essentially. Why cannot descending plane lay a claim on a runway in some computer system and a cabin in Southwest - blare a horn for pilots trying to steer into a claimed runway?