hmm is that such a feat? landing on a carrier compared to landing on land? The variables involved should be pretty predictable and can be modeled accurately...
I know this very well, but the point is that this is irrelevant anyway, because the relative movement of the carrier matter much less than the difficulty of catching the hook. Whether you match the speed to the carrier deck or to the ground makes (almost) no difference. It's not as if the carrier was suddenly accelerating while a plane lands.
Catching the hook while the plane is in a controlled crash, that's the problem.
i would venture to say this is more of a controlled environment than autonomous vehicles. The variables you have to deal with in real-time for vehicles is much greater than landing on a carrier.
Having done both extensively, in all weather conditions, I respectfully disagree. Now, modeling that in an algorithm...you may be accurate.
The thing about a pitching deck is that it's somewhat, but not completely, predictable. Swells are measurable and generally periodic...but the generality breaks down enough to make it unpredictable. It's why boarding rates (i.e. the ratio of landings to attempt) drop by a factor of 2-3 (i.e. from around 90+% to 30-ish%) in pitching deck situations.
I can heartily recommend 'The Right Stuff' by Tom Wolfe, where he gives a great description of how challenging it is for new pilots to learn this skill.
There is more involved with landing on a carrier vs a land runway. The carrier deck is much shorter so the plane needs to grab the arresting cable and there is only a small window for this. If you miss it, you need to quickly speed up and take off again for another pass.
you need to quickly speed up and take off again for another pass.
In fact, the aircraft always go to max throttle at touchdown, just in case they need to go around for another try. This, as opposed to going to max throttle after realizing there's a problem.
It definitely is a feat. Carrier landings on US-supercarrier-style carriers especially are pretty notorious -- between the ocean swell, the short runway and tiny margin of error, plus a runway that's actually angled with regards to the carrier's direction of travel, it's pretty challenging.
> plus a runway that's actually angled with regards to the carrier's direction of travel, it's pretty challenging
I hadn't really considered that before. Say the carrier is doing 20 knots then even on a one minute final approach the carrier will have moved by getting on for half a mile. ( very much fag-packet estimated maths but enough to show you can't just point at the carrier and land in a straight line)
The US Navy has actually had automatic carrier landing systems since the 1960s. The advancement here appears to be the carrier "taking over" the plane in order to land it with no input from the pilot.
This is interesting. I’ve been out of the cockpit for about 5 years, but we had what we called “mode 1” approaches back then where the carrier would lock onto our f-18. We could land without touching the controls. At the end of my time, they actually incentivized us to fly these approaches more frequently to develop unmanned systems more quickly. (Incentive was giving a perfect grade for a landing...usually Mode Ones were a “no count” in the grading rubric.)
I guess the difference here is that an actual person (the LSO) is controlling the jet vice the computer?
I'm not really sure what the benefit of having a remote operator landing the plane vs having a computer do it. I guess the remote pilot can be the third tier in who can land the plane (in-seat pilot, computer, ship remote pilot) and for UAVs (UAV pilot, computer, ship remote pilot). Is the automated landing system that crappy that they want to add in a remote operator? If they can already land a plane automatically, the the next step they should focus on is an automated air traffic control, landing and launching any planes within the 5 mile radius (assuming that doesn't already exist). Imagine a smaller carrier like a LHD totally devoted to UAVs that launch and land without anyone on the tarmac.
I think the idea is that UAV operators/pilots might not even attempt landings. There's a future where assume control of UAVs as they approach station, and release as they return to base.
In that scheme, they were probably expecting the UAVs to land automatically on the carrier, with something like ATARI being the backup.
This way they can leave their operators focusing on mission tasks instead of recovery.
Seems like a small gain for a small chance of the loss of a $100 million asset. Wouldn't it be good to give pilots practice landing in adverse conditions just in case this "autoland" fails similarly to how Tesla and Uber's autopilot screws up fatally when the road lines are a bit faded?
On the other hand, landing probably is an easier task than driving given the small amount of ways it can go right versus wrong.
Is the actual hardware so expensive?
I've heard it's the systems, and increasingly, in them, the software.
So mostly the number bandied around might be just development and for total cost, upkeep costs, divided by the number of airframes.
So losing one isn't so hard, if it's unmanned.
That's why they really need to eliminate on board pilots.
There's always going to be a need for humans in the loop. In any sort of major wartime scenario countries will be shooting, disabling, and jamming any form of comms and guidance they know about.
What about autonomous drones that fallback to using stars and the moon positions for navigation if comms/guidance are jammed? I know this wouldn't work during sunny daytime hours but it's an interesting technique isn't it?
If communication or guidance is disabled, aircraft still need to function in a complex combat environment: they need to be able to intelligently identify targets, follow rules of engagement, engage correctly, and weigh the costs and benefits of different courses of action. An automated system will never be able to both practically do these things and also emotionally satisfy the humans in charge of militaries (you will never be able to convince a human officer that a drone will be able to autonomously follow rules of engagement and correctly discriminate friend from foe).
you'd have issues during instrument conditions (stars not visible regardless of daylight), but this could be interesting, especially combined with, say, SLAM fed by terrain-scanning radar.
And this, actually, is why those "hundred-million-dollar boondoggle" fighter jets are really so expensive: they're extremely fancy routers, connecting the (cheaper) planes around them into redundant mesh networks (well, redundant when you have more than one fancy plane) and connecting that mesh network to ground control with redundant-per-fancy-plane ground relays/satellite uplinks over several different physical media (broadcast RF, directed microwave, some kind of laser, etc.)
Since at least one of the fancy planes is pretty much guaranteed to stay connected to its flight, you can—if you have the room—stick your CIC in there, rather than on the ground. Then, rather than remote-controlling your fighters from the ground, you can remote-control your fighters from another fighter. Which means that now you only need one crew to pilot, say, 8 aircraft. (Well, for now, doctrine would still require humans in the cockpits of the subordinate aircraft; you just don't need to train them nearly as much on how to execute precise group maneuvers. Like driverless cars on highways!)
This is oddly easier in the air (because aerial maneuvers are something fighters need to do as a unit); the hard part is takeoff and landing, because there's not enough hands in an aerial CIC to pilot all the planes up and down. So you want these hypothetical subordinate drone fighters to be able to be passed off automatically between aerial CIC and ground control. (It's convenient if they mostly launch and park themselves, because you then need less ground-control staff on your aircraft carrier or whatever, but it's important that they can be passed off to humans on the ground when necessary.)
I mean maybe it's too much of a reach, and they need to use ATARI or something like ATARI on every landing. In any case, the idea is that maybe your mission operators don't have to be your recovery specialists.
Having options and fallbacks are good. I can't imagine the Navy not looking at all their data in 5 years and figuring out a reasonable plan.
We're all just armchairing it. Engineering data is really all that matters at the end of the day.
I suspect that landing on aircraft carriers is a skill that they do not want people to forget. For example, the military has recently started teaching officers celestial navigation because GPS can be jammed [1].
A few years ago, I saw Captain Cook-looking gear at a friend's house. I knew he like boats, so I thought these were collectibles.
Turns out that he was a sea scout, as a kid. Civilian GPS wasn't available in the early 2000s... at least not to amateur navigators. I think it only became standard for airlines in the 90s and militaries in the 80s.
That GPS is new is not surprising. That one tech-generation back is "celestial navigation" with a sextant is shocking (at least it was to me). It was like discovering that before smartphones the closest thing available was signal fires.^
Good reminder that tech can work in big leaps, and that a good technology can last a long time. Moore's law has been such a factor for digital technology, that we can forget it is not the norm.
BTW, the alternative to GPS (unless you have a terrestrial signal) for airlines is computerized dead reckoning, the equivalent of throwing a log tied to a rope overboard and counting knots, but a computer counts the knots.
Yes, my impression is the LSO can correct / guide it in ultimately.
However, I am wondering what happened to all of the technology that they put together for the X-47B - which I believe was fully qualified to land itself in most weather conditions:
> During testing, the ATARI system operators controlled an F/A-18 aircraft using a joystick, while a safety pilot sat in the cockpit as backup. The technology is capable of taking over an aircraft from up to five miles away.
It's an interesting choice of wording when they say the LSO can "take over" - I hope the encryption is strong enough that LulzSec doesn't also "take over" a military aircraft.
What could go wrong with remote control technology over our important military assets? Of course the answer is a lot, especially in wartime. For all the paranoid people thinking that the military or police or whatever US security service will take over the country in a secret coup, the people who make up those forces are an important bulwark against tyranny from within or without.
It's very dangerous to be a pilot and a soldier but as a software engineer I don't know how I could ever feel safe in a world without people making decisions about how to use weapons.
Iran has claimed a lot of things, but I have serious problems believing them when they got their nuclear facility hacked. My office network has more security than that.
It's not autonomous like a drone, it's controlled by human(s) on the aircraft carrier with a pilot in the cockpit. It's also being billed as a "backup system" for safety (to correct course) and for emergencies.
It still opens it to hacking, as much as people don't want to hear it. Essentially you have system for remote operation and if it can be hacked, then it can be used to direct the plane.
Cryptography works, but the implementations are frequently flawed.
Plus, things like this introduce systemic risk where an adversary can take over multiple aircraft simultaneously, or at least render them all inert midair.
If remote controlled aircraft are a problem then basically anything on a network is a problem. Pretty sure nuclear power stations, missile silos, satellites, dams, darn near everything has a remote controlling and monitoring element these days. You could just shut down a whole country.
We are going to have a very sorry series of tragedies if that doesn't change. Remote monitoring is fine, remote control is not. At least, not for things like nuclear power, weapons systems, vehicles that can drive very fast, etc
Secure remote control requires secure systems, which in turn requires secure humans, and we will never solve this last requirement.
The way around this is by preventing systemic attacks. Analogous is how paper voting—while vulnerable to things like vote stuffing—isn't susceptible to the systemic problems that electronic voting typically is.
I'd say secure systems need to be based on the assumption of insecure humans. Nothing is absolute in security but we definitely should start the analysis expecting people to behave incorrectly and insecurely. This is not a new problem: https://en.wikipedia.org/wiki/Byzantine_fault_tolerance
At the scale of unique UAVs it is even possible to use one time pad loads before each mission from a codeplug. I can assure you that DoD "type 1" crypto does work.
73 comments
[ 4.4 ms ] story [ 147 ms ] threadBeing accurate enough in turbulent weather to catch the hook seems to be the main problem.
Catching the hook while the plane is in a controlled crash, that's the problem.
The thing about a pitching deck is that it's somewhat, but not completely, predictable. Swells are measurable and generally periodic...but the generality breaks down enough to make it unpredictable. It's why boarding rates (i.e. the ratio of landings to attempt) drop by a factor of 2-3 (i.e. from around 90+% to 30-ish%) in pitching deck situations.
In fact, the aircraft always go to max throttle at touchdown, just in case they need to go around for another try. This, as opposed to going to max throttle after realizing there's a problem.
I hadn't really considered that before. Say the carrier is doing 20 knots then even on a one minute final approach the carrier will have moved by getting on for half a mile. ( very much fag-packet estimated maths but enough to show you can't just point at the carrier and land in a straight line)
I guess the difference here is that an actual person (the LSO) is controlling the jet vice the computer?
In that scheme, they were probably expecting the UAVs to land automatically on the carrier, with something like ATARI being the backup.
This way they can leave their operators focusing on mission tasks instead of recovery.
On the other hand, landing probably is an easier task than driving given the small amount of ways it can go right versus wrong.
So mostly the number bandied around might be just development and for total cost, upkeep costs, divided by the number of airframes. So losing one isn't so hard, if it's unmanned.
How much of the hardware is devoted to keeping the pilot alive? Once you remove that, the hardware might be a whole lot simpler and cheaper.
Unmanned, with autonomous carrier take offs, landings, and aerial refueling.
There's always going to be a need for humans in the loop. In any sort of major wartime scenario countries will be shooting, disabling, and jamming any form of comms and guidance they know about.
It would; you just need to filter out the sun. Apparently this is already a thing: http://www.trexenterprises.com/Pages/Products%20and%20Servic....
Since at least one of the fancy planes is pretty much guaranteed to stay connected to its flight, you can—if you have the room—stick your CIC in there, rather than on the ground. Then, rather than remote-controlling your fighters from the ground, you can remote-control your fighters from another fighter. Which means that now you only need one crew to pilot, say, 8 aircraft. (Well, for now, doctrine would still require humans in the cockpits of the subordinate aircraft; you just don't need to train them nearly as much on how to execute precise group maneuvers. Like driverless cars on highways!)
This is oddly easier in the air (because aerial maneuvers are something fighters need to do as a unit); the hard part is takeoff and landing, because there's not enough hands in an aerial CIC to pilot all the planes up and down. So you want these hypothetical subordinate drone fighters to be able to be passed off automatically between aerial CIC and ground control. (It's convenient if they mostly launch and park themselves, because you then need less ground-control staff on your aircraft carrier or whatever, but it's important that they can be passed off to humans on the ground when necessary.)
Having options and fallbacks are good. I can't imagine the Navy not looking at all their data in 5 years and figuring out a reasonable plan.
We're all just armchairing it. Engineering data is really all that matters at the end of the day.
[1] https://www.npr.org/2016/02/22/467210492/u-s-navy-brings-bac...
Turns out that he was a sea scout, as a kid. Civilian GPS wasn't available in the early 2000s... at least not to amateur navigators. I think it only became standard for airlines in the 90s and militaries in the 80s.
That GPS is new is not surprising. That one tech-generation back is "celestial navigation" with a sextant is shocking (at least it was to me). It was like discovering that before smartphones the closest thing available was signal fires.^
Good reminder that tech can work in big leaps, and that a good technology can last a long time. Moore's law has been such a factor for digital technology, that we can forget it is not the norm.
BTW, the alternative to GPS (unless you have a terrestrial signal) for airlines is computerized dead reckoning, the equivalent of throwing a log tied to a rope overboard and counting knots, but a computer counts the knots.
However, I am wondering what happened to all of the technology that they put together for the X-47B - which I believe was fully qualified to land itself in most weather conditions:
https://www.naval-technology.com/projects/x-47b-unmanned-com...
btw: Thanks for your service.
> During testing, the ATARI system operators controlled an F/A-18 aircraft using a joystick, while a safety pilot sat in the cockpit as backup. The technology is capable of taking over an aircraft from up to five miles away.
It's very dangerous to be a pilot and a soldier but as a software engineer I don't know how I could ever feel safe in a world without people making decisions about how to use weapons.
Plus, things like this introduce systemic risk where an adversary can take over multiple aircraft simultaneously, or at least render them all inert midair.
The way around this is by preventing systemic attacks. Analogous is how paper voting—while vulnerable to things like vote stuffing—isn't susceptible to the systemic problems that electronic voting typically is.