Hacker News new | ask | show | jobs
by tehf0x 1001 days ago
So a similar thing happened near some friends' in France. A military jet crashed into the forest near their house, but the air force couldn't figure out where it had gone. Eventually a farmer noticed that a new pond had appeared on his land. The jet made enough of a crater when crashing that drained the nearby swamp and created a new pond deep enough to conceal the full fuselage, thus completely hiding the airplane. Once the farmer alerted the air force, they were able to crane the remains out of the newly formed pond and recover the key parts of hardware onboard. Had the farmer not noticed the change in landscape they might have never found it. https://www.lefigaro.fr/flash-actu/2011/03/02/97001-20110302...
9 comments

I wonder if some of the "autopilot" functionality for military aircraft is to swan-dive into water/forest if everyone has already ejected.

If it's going to crash (and can't autoland, which it probably shouldn't attempt even if it could if something ejection-worthy happened), might as well obliterate the thing in a safe place.

The thing you’re thinking of is the pilot. If the plane is capable of control, the pilot will move it on a trajectory away from populated areas if possible. Protecting innocent people on the ground in the case of an emergency was always top of mind, you can see this in the crash reporting for multiple real world incidents (in airframes with, and without ejection seats, where the last actions of the aircrew were steering away from populated areas)

By the time a pilot ejects they’ve exhausted EVERY other option to control the aircraft, no AI is going to regain control at that point.

> By the time a pilot ejects they’ve exhausted EVERY other option to control the aircraft, no AI is going to regain control at that point.

It seems like they probably bail out when they've exhausted every option of being able to land it and survive. There are likely some scenarios where there's limited controls remaining, that wouldn't provide high enough odds of guaranteeing someone's survival, but that could optimize for something when it eventually makes contact with the land. In fact, it seems like there could be quite a bit of capabilities between where someone would want to bail out and where there are zero remaining controls.

I live in the real world, and there are incident reports that go into great detail about every aircraft loss. I’ve read every one since 1999, and I’ve never seen an incident where the pilot ejected over a populated area in an aircraft that was still controlled.

What you all are imagining does not happen, and the odds of it happening are vanishingly small.

https://www.thedrive.com/the-war-zone/russian-su-34-fullback...

“The statement went on to add that the crash was caused by an issue with one of the engines during takeoff, but the jet's two crewmembers are said to have ejected safely and survived. The extent of the injuries to the apartment building's tenants as well as civilians in the surrounding area is unknown at present.”

"To the last I struggled to lift the plane, [but] copilot Yuriy Yegorov hit the catapult [triggering ejection] and we two ejected with our seats."
Have you looked at the MiG-23 that crashed last month in Michigan?
https://data.ntsb.gov/carol-repgen/api/Aviation/ReportMain/G...

This one? Which reads as though the back-seater (not the pilot) was the one that ejected and took the pilot with him?

I think it's worth noting that that the MiG-23 that crashed in Michigan was a privately owned aircraft, flown not by an active service member. I highly doubt the military allows their pilots to eject without absolute certainty that the multimillion/billion dollar aircraft is totally lost.

Additionally, I highly doubt there are many privately owned military jets equipped with ejection seats that are allowed to fly, especially in residential airspace.

Also, as someone who works on FMS's the likelihood that a military program would spend the money required to code an AT/AP to have that capability is just too close to zero.

In flight school lore, during a training exercise, a plane righted itself from "uncontrollable flat spin" after ejection. Basically, pilot input can fight against the natural stability of the plane's design.

Also:

> During a training mission from Malmstrom Air Force Base, on Feb. 2, 1970, his F-106 entered an uncontrollable flat spin forcing him to eject. Unexpectedly, the aircraft recovered on its own and made a gentle belly landing and skidding for a few hundred yards on a field near Big Sandy, Montana, covered by some inches of snow.

https://theaviationist.com/2013/10/24/cornfield-bomber/

Worth noting that spin recovery is highly CG (center of gravity) dependent. Ejecting from an aircraft would significantly alter the CG. It's far more likely that the CG change broke the spin than the pilot was doing something unhelpful.

Additionally there would be a nose down moment from the seat firing (newtons third law) and that may well have broken the stall.

Visually the cockpit of a fighter plane is located forward, thus very likely ahead of the plane's CG. Ejecting the pilot would thus move the CG towards the back of the plane, which is a less stable configuration making it more difficult to recover from a stall.
https://en.wikipedia.org/wiki/1983_Negev_mid-air_collision

>Notably, the F-15, (with a crew of two), managed to land safely at a nearby airbase, despite having its right wing almost completely sheared off in the collision. The lifting body properties of the F-15, together with its overabundant engine thrust, allowed the pilot to achieve this unique feat.

Just posting here because it is awesome.

At first I thought you were referring to this case where a crew landed an A300 with a large part of a wing missing and no hydraulic controls, using only differential engine thrust to control pitch, roll, yaw and speed.

https://en.wikipedia.org/wiki/2003_Baghdad_DHL_attempted_sho...

Near the Weizmann Institute, there is a monument for a pilot who refused to eject over the populated area, and rather kept control over his plane for long enough to divert it away. He died in the crash, as he no doubt would have anticipated.

https://honorisraelsfallen.com/fallen/holtzman-chaim/

Not really, sometimes a plane will land itself even after the pilot ejects. https://en.m.wikipedia.org/wiki/Cornfield_Bomber
If you read that Wikipedia page, it’s pretty clear the plane would have not recovered without the change in nose attitude from the ejection.
Which contradicts your original claim that "no AI is going to regain control at that point."
Ok. I admit I was wrong then.

I doubt an autopilot would have regained control in a flat spin without an ejection to drastically change the dynamics of the aircraft. I also made the point later that a sufficiently capable AI that could would essentially be able to replace the pilot, rending them moot.

Look, the OP clarified that he meant he wanted an autopilot that tries to lawn dart the plane in an incident of ejection, the purpose of which was to save sensitive technology from the enemy. I still maintain that is silly and if we funded such a program it would be a waste of money.

It seems AI was ablento regaim control after the ejection stabilized the aircraft.
> By the time a pilot ejects they’ve exhausted EVERY other option to control the aircraft, no AI is going to regain control at that point.

Every other option with the pilot on board. The plane _may_ be easier to fly without it.

FTA: “The loss of the weight of the crew, seats, and canopy, as well as the shift in the center of gravity, have seen aircraft pull out of an imminent crash with nobody left onboard.”

What Scoundreller meant is that if most of the controls are lost, then the few controls remaining could still be used to intentionally destroy the plane as much as possible. For example, the elevator (to control the pitch) might be the last working control, and it could be used to intentionally nosedive the plane after ejection.
I got that and I was trying to answer in the least sarcastic way I could. I’ll try again.

Autopilot isn’t an all knowing AI that is better at dealing with emergency than a pilot. In the region of flight that involves a pilot ejection, especially in a plane like the F-35, the “autopilot” that would be created that could successfully scout a target area that was safe to crash in, eject the pilot, then somehow move the plane to that area with the canopy off the plane and the degradation of control that would involve, plus the million issues that could have caused the ejection in the first place… it’s not plausible to create this and still have a pilot. You’re talking about a system that is a better aviator, with more SA and more detail about aircraft systems than the human at the stick.

So, first, I don’t think we’re there with AI that is better than the pilot in this region of flight, second, to make that would be so expensive as to be ridiculous, for the incredibly rare event that a military plane needs to eject over a populated area yet also has a safe area like a body of water to crash into. Third, making that system would negate the need for a pilot in the first place so again, what’s the point?

I get the thought, but it’s silly.

I was not thinking of some highly advanced AI, just something ultra simple like: after ejection, attempt to pitch the plane to nosedive as vertically as possible. The intent being to to crash the airplane as badly as possible to prevent the enemy from recovering anything useful from the crash site.
Sounds like a good way to destroy your own runway, barracks near the runway, etc.
Modern ejection systems are zero-zero systems designed to allow ejection at zero speed and zero altitude which means that the aircraft may not have enough energy to successfuly destroy sensitive components across the full ejection envelope.

Why not rig it with explosives around the sensitive components and avoid the messy endeavor of trying to orient the plane for maximum destruction after ejection when that is likely to be unreliable at best?

to add

4 - if you had control with some badass AI why not land the plane safely for a recovery... the example in both the OP and the GP are planes lost in the country they are from not in enemy territory. The planes are not cheap... why would you purposefully wreck it in in a safe location.

Actually my example assumed being in enemy territory. Nosedive the airplane for maximum destruction, to prevent the enemy from recovering anything useful from the crash site.
EXACTLY
> Autopilot isn’t an all knowing AI that is better at dealing with emergency than a pilot.

b...bu... but all the twitter influencers told me... /s

I don't think that solves anything. With critical secret pieces of hardware you will still want confirmation that they were destroyed or recover them. So even if the plane can attempt to self-destruct as much as possible, the military is still going to want to confirm the result.
The skin and coatings of these planes is secret. How do you self destruct the skin?

Anyway, it’s moot. The people with the resources to actually make this stuff already hacked in and stole it. They compromised a whole CA just to get the F-35 design docs.

> If the plane is capable of control, the pilot will move it on a trajectory away from populated areas if possible. Protecting innocent people on the ground in the case of an emergency was always top of mind, you can see this in the crash reporting for multiple real world incidents

One of the things I remember from Chuck Yeager's autobiography (his first one - I think he eventually had several) is that he called bullshit on this. His view was that the pilot was putting all his attention and focus into saving the aircraft, and stories about how the pilot steered a failing aircraft away from something important on the ground were not true.

Of course, after the fact a pilot is going to say he tried to save people on the ground...

(his view might have been cynical, but I expect it often holds true)

There’s many voice recordings, instrument logs, and even ATC conversations of people aiming doomed aircraft away from populated areas. Military pilots may have different priorities as they can eject, but many civilian pilots have spent their final moments trying to minimize casualties.

Exceptions defiantly exist, but it’s a very common reaction.

Yeah these are by far the most harrowing ones to listen to. You can hear the resignation in their voices before they steel themselves to try anything left to avoid loss of life on the ground.

Real rough shit.

I do expect having an ejection seat might alter behavior, as well as the encouragement of an ATC. And Yeager was a test pilot who was part of a community of test pilots...

In my memory, my impression of Yeager's brief anecdote was that "the pilot heroically sacrificed himself to save the lives of innocents on the ground" was a common enough bullshit news story that he wanted to debunk it. A pilot would be focused on saving himself and his aircraft to the fullest extent possible.

Test pilots are also generally avoiding populated areas where civilian airports are often very close or even inside them. The trope of military jets crashing into a desert, ocean, or farmers field represents the most likely outcomes.
I've seen few episodes of aircrash investigation (mayday), and multiple times pilots try to crash or land where they will do less damage on ground.
My problem with the OP's comment is that a couple of different things are stated as universal truths, even though they definitely aren't.
Maybe Chuck Yeager was a selfish asshole.

There are many incidents that prove this wrong. The comment above yours is one example.

Chuck Yeager emphatically was a selfish asshole, but...

> There are many incidents that prove this wrong.

No, there aren't, and your saying so is just a failure in logic. Literally every instance of a crash would have to play out the way you insist it should, with a pilot taking into account the presence of population centers before ejecting, in order to "prove this wrong."

The tag team duo of MiG pilots who crashed in Michigan a couple of weeks ago didn't even agree on whether it was time to eject. They certainly hadn't placed the plane on some kind of safe trajectory (it landed literally right next to an apartment building). For that matter, maybe that incident is somehow an argument for selfishness: if they hadn't punched out right when they did, they would have been outside the envelope for a safe ejection, and that would be two dead people for sure.

All that said, I don't think you and I would disagree on how pilots should handle the situation. I just don't think it plays out as well as we'd like every time.

So you posit that the ejection seat adds some selfishness to the pilots? Because I’m not kidding there are a lot of incidents where we have inflight recordings of pilots without ejection seats steering away from populated areas in their last action, as well as the incident where an Israeli pilot with an ejection seat choose not to eject (and died) moving his plane off line.

Then, I’ll just tell you in my personal experience in the airplane, I’ve seen 2 pilots risk their lives moving away from a town and not die (luckily) but crash and destroy their airframe. The other pilots I flew with, I’d say 95 out of 100 would do the same.

Maybe Chuck was just a selfish asshole.

> By the time a pilot ejects they’ve exhausted EVERY other option to control the aircraft, no AI is going to regain control at that point.

See, you say that, but this one seems to have done a good job of flying off to who the fuck knows where after the pilot ejected.

lol. The first incident I’ve ever seen where this happens. Just perfect that I pop off with absolute statements and then am proven wrong by the incident I was responding to.

The F-35 is really a game changer. What a shitbox

Happens to the best of us.
And could potentially still be flying.
Will this always be true though? Not even thinking about advancements in AI, but from a human body g-force standpoint, surely those jets can already pull way more Gs than the pilots can handle.

I doubt the software is doing that today but why couldn't there be maneuvers the plane could do at 15 Gs that would help it survive?

I’m sure there is somewhere at the extreme edge of the bell curve that this might be true for, but in general, no.

For one, the jets are engineered around the limits of the pilots. These are high performance military aircraft where every ounce of weight matters, the airframe isn’t over engineered to support 15g maneuvers. Ripping the wings off wouldn’t help in an emergency.

I made the point later that by the time the AI is a more capable aviator, with the SA to do what OP was suggesting, the pilot is redundant. Take them out, engineer the plane to make those 15g turns with the extra weight you save not having life support, seats, canopy, etc.

Right -- I had the same thought coming back to this. We know planes aren't capable of this because as soon as they are you wouldn't have pilots. Maybe 'planes' is too strong since there are a bunch of people playing video game drones in the middle east from Alabama or whatever.
Depending on the situation, there might be not enough room or time to save lives in ground.

Sample dual ejection:

https://youtu.be/zN_Zl64OQEw?si=5Nk-LGBFxf4pmEDr

Actually, no. A pilot might punch from a perfectly flyable aircraft that ran out of fuel and can't be glided in due to terrain. Or a plane that has no propulsion for other reasons. They might even punch from a plane with some fuel but damage that precludes landing--if they're over civilization they might point it into nowhere and punch so they come down over civilization.
Ok, and here we go with the armchair aviators. Time in a Microsoft flight simulator doesn’t give you expertise.

I’ll just reiterate, by the time a pilot is ejecting from the aircraft, everything has been done. Read through the incident reports and find me an incident where this wasn’t done, cite it please. I can’t find one.

https://www.airandspaceforces.com/docs/type/accident/

I'm not saying they would punch if there was something to be done--the pilot could have found the problem was one that couldn't be dealt with. I'm saying that the control surfaces might be operational on a plane the pilot was punching from. Desert Storm an Iraqi pilot punched from an intact plane--he beat the Phoenix (long range shot) but ran out of fuel doing it.
I mean, are there detailed accident reports for the sort of military plane with “keep it away from the enemy” type equipment onboard? Seems like you wouldn’t want to give away information about crash behavior since that could help the enemy recover the equipment.

Not saying it is plausible or not in the first place, I have no idea, but I don’t see how a lack of published reports of this happening in the real world proves anything.

I think people think ejecting from one of these planes is like something you would do casually if things aren't going well.

My limited knowledge leads me to believe this would be an absolute last resort and has a high risk if injuring the pilot.

Also, installing these in the passenger seat of your car is expensive, difficult, and illegal just in case anyone is also thinking of this right now.

The key point is that the plane does not crash in a populated area (and secondarily that a military plane is destroyed if crashing in an area not controlled by the owner); in both of your examples, the pilot routinely does the first if not necessarily the second.
Well usually it is pilots responsibility to try to aim it in safe direction as much as possible and only then eject. Basically like captain on the ship you don't eject but you are last one to leave the ship after you made what is possible to save others from harm (unless they are of course hostile forces).

I think quite some pilots died this way because they were trying to the last second to save other lives.

If plane is that much out of control that you cannot do much, adding some code to try to do something might make things actually worse from my perspective.

Yup, plenty of stories of pilots going in with the plane because they were guiding it away from people on the ground. That's why modern ejection seats can save you even if you're at ground level (although you do have to be upright in that case.) It lets the pilot ride it as long as possible and yet escape.
Including Gagarin
I doubt it. You do not want an uncommanded activation of that system for any reason. You may actually _want_ your disabled plane to continue in a specific direction for tactical reasons. You generally do not allow auto pilot to make large deflections in control surfaces, as you always want a pilot on the stick to be able to overcome any uncommanded autopilot actions.
> I wonder if some of the "autopilot" functionality for military aircraft is to swan-dive into water/forest if everyone has already ejected.

Not quite the same you are thinking about but something similar is documented behaviour of the Global Hawk. Obviously since that is a remotely piloted aircraft “everyone has ejected safely” is not the trigger for it.

The way it works at the flight planning stage the operators define pre-determined points, and if the system detects certain faults it cannot recover from it tries to fly to these points and crash land into them. They call these point “termination point” on land and “ditching point” over water.

In this document[1] you can read more about the selection criteria of such points.

This document[2] details for air traffic control under which conditions flight termination points are used. In short (page 22) when the aircraft is uncontrolable for landing or landing at a suitable airfield cannot be achieved safely.

1: https://static.e-publishing.af.mil/production/1/af_a3/public...

2: https://www.eurocontrol.int/sites/default/files/2019-05/atm-...

Some Hollywood points in this direction. E.g. the Behind Enemy Lines ejection scene.

Seems like a pretty good idea to self-destruct the most sensitive hardware (computers/chips/storage modules/etc) if recovery of the aircraft is no longer feasible.

> self-destruct

Ooh, yeah. Let's add even more failure modes!

For drones it makes sense. Drones could boot up with all config in RAM so as to be completely devoid of mission parameters once power is cut.
Sure. There's no way that in combat you could experience a temporary power loss. Nope, that could never happen.
If your drone has power loss during flight, I think you have other problems.
I think the failures leading up to ejection preclude any kind of automated flight control afterward. Seeing how modern aircraft are all fly by wire.
It's relevant to note that the aircraft disappeared from radar in the evening while doing "low altitude exercises" and was discovered the next morning (bad weather preventing the rescue teams from finding it earlier).

The linked article mentions that the pilot and navigator were considered missing. Other news reported also the presence of human remains near the crash site.

From the satellite images of the area, you have several urban areas with large swaths of farmland and some forest around them. One of the lakes is a dammed river. The flood plains downriver from the dam are quite lush. It’s in a field, under a tree, or in a pond it dug itself if the water table is high right now.
Just like Luke's X-Wing on Dagobah!
Surprised they didn't get Yoda to Force-levitate it out.
But why? Don’t these billion dollar machines have a GPS tracker on them?
They have transponders. But military airplanes can and sometimes do turn them off. When airplanes broadcast signals they can be detected. The military sometimes wishes to avoid this.
One of the article updates quotes from a Washington Post story:

> "The jet’s transponder, which usually helps locate the aircraft, was not working “for some reason that we haven’t yet determined,” said Jeremy Huggins, a spokesman at Joint Base Charleston. “So that’s why we put out the public request for help.”

It could activate after ejection, certain acceleration pattern (i.e. crash) or could start broadcasting only after receiving very specific signal (one-time code) on a specific frequency. There are probably plenty more options, it doesn't seem like a hard problem from engineering standpoint.
If you’ve gone down you probably don’t want the plane to be announcing it because that will tell the enemy where your pilot and hardware is.

There’s pros and cons to the enemy finding you but generally I’d imagine you want to give the pilot the chance to escape.

Civilian aircraft, even 2 seaters, have had ELTs for decades. They are being replaced with better systems, but they were designed to do just that: broadcast a signal on a specific frequency after a crash.

Doubtful you want to broadcast the position of the latest generation stealth fighter, however.

As mentioned, it can be silent until receiving specific signal. It can be single frequency it can be multiple, but if it's specific string of bytes or specific timing it's practically impossible to brute force when it's long enough, and when looking for the plane you can broadcast it with high intensity because it is only single use.
I don't think stealth aircrafts normally have trackers or anything broadcasting a signal. Kind of defeats the purpose...

I did read that stealth aircraft have transponders installed when operating in US airspace, so commercial radar can see them better. But it was not installed on this flight.

The article mentions that they have transponders that can be used but apparently in this case it was off or not working. They can also attach a radar reflectors when they want to be more visible to radar, and apparently the weapons configuration plays a role too:

> "Also, the jet's configuration and its avionics' operability are an issue. The F-35s wear radar reflectors when on transit flights, as well as on many training missions and some operational ones. The Marine jets often wear missile rails for AIM-9Xs, as well. But if the aircraft was in its full stealthy configuration and had avionics issues, tracking it may have been troublesome."

They're always installed, just selectively disabled as needed.
In civilian planes, sometimes, standard ADS-B tracker turned on automatically.

But in military planes, it must be turned on explicitly, because, depend on target of flight, it is possible, it must be OFF.

In general, this is very frequent case, when pilot just forget to turn on ADS-B.

They have transponders that transmit their identity and location on the most modern, but those depend on being active (for some reason the one in the F-35 was not) and on there being receivers nearby to pick them up. At lower altitudes in remote locations coverage is spotty.

So it's not a guarantee.

Why would an AirTag not work in this scenario? I’m guessing it would be a security risk?
First, yes it would be a security risk. Second, AirTags depend on there being a set number of iOS devices nearby to provide location data and an internet connection. You're not going to get that in a rural area.
Alright maybe not an AirTag but a Tile tag would surely work?
That would be a worse option, given they share location data.
It operates on the exact same principle. So, no.
Turn your stealth fighter into a regular one with just $30!
GPS is not reliable underwater, I'd expect it to have an ELT/EPIRB but those don't work with GPS as far as I know, and the crash could have been hard enough to render it inoperable.
Jesus, imagine the thrill, that you have a full blown airplane in your backyard. Going out during the nights and sleep next to it, sit in the cockpit, whatever.
Why don't these aircraft get 10Hz GPS updates and send them to StarLink as they are going down? It's 2023 already.
Because they're stealth military planes and broadcasting it's exact position to Musk owned network is literally the primary thing it should NOT do.
Well, for the plane in the parent comment's article, that reason is likely because the design for that plane was started in 1972.