[nradov]

It's always hilarious to see ignorant developers on HN claiming that real world engineering problems are easy to solve based on zero actual knowledge or experience. This kind of comment is really peak HN.

An autopilot for airplanes is only "easy" until something goes wrong. For example, one failure mode for autopilots is that if the aircraft gets progressively more and more out of trim the autopilot will automatically compensate until it hits its design limit. Then it suddenly disengages, leaving the human pilots in manual control of a nearly uncontrollable aircraft. If you talk to an actual flight control engineer they can give you plenty more examples of why building a safe autopilot is quite hard.

[terminalshort]

And yet it was done decades ago. Air traffic control is just as solvable.

[nradov]

"Done" in what sense? Do you even understand how autopilots work and how limited they are?

[kayodelycaon]

That’s a nice strawman you’re creating there.

An aircraft has fewer and simpler variables to deal with than ground vehicle.

If a ground vehicle runs a red light, it’s potentially fatal error. There are more of these for a car than there are for an airplane.

You don’t have to write automation to avoid hitting trees in a plane. An airplane just needs terrain data and a few algorithms.

There are a few enough airplanes and airplane manufacturers that you could regulate a specific algorithm for traffic avoidance.

[Aurornis]

> There are more of these for a car than there are for an airplane.

Half of this comment section has strangely simplified ideas of how airplanes work and how a flight might get into trouble.

It's crazy that so many comments are convinced that completely automating airplane flight is some relatively trivial problem.

[nradov]

Those comments are coming from people whose aviation "knowledge" was learned by playing Ace Combat on Xbox and watching Snakes on a Plane. Totally disconnected from reality.

[nradov]

That's a nice strawman you're creating there. In some airspace classes and flight regimes an aircraft has more variables, especially when you account for possible failures. If an aircraft has a mechanical failure it can't just pull over and stop.

There are about 46000 aircraft registered in the USA, plus more that sometimes fly in from foreign countries. Many aircraft were manufactured decades ago by companies that no longer exist so major upgrades aren't economically practical.

[oasisaimlessly]

So why did we have airplane autopilots decades before car autopilots if it's not easier?

"Easier" != "easy"

[Aurornis]

This is such a strange comment section.

Airplane autopilot is more like the cruise control feature in your car, not a self-driving computer that does everything for the pilots while they sit back.

Car autopilot and airplane autopilot don't share much in common other than the word "autopilot"

[nandomrumber]

Modern auto pilot and flight management computer combos can fly way-points and perform full Cat III auto lands.

I’m not suggesting the pilots are sitting there doing fuck-all, or that they are not necessarily.

I think what the automate ATC advocates are suggesting is to bring ATC in to the 21st century.

[nradov]

Yes, and that's what the FAA NextGen program has been doing incrementally since 2003. There are probably ways to accelerate it but it seems like most of the "automate ATC advocates" are simply ignorant and haven't done their homework.

https://www.faa.gov/nextgen

[nandomrumber]

Thanks for that link too. I wasn’t aware of the extent to which all those is already underway.

My knowledge is mostly limited to a casual watching of the aviation YouTube boffins.

[tjohns]

Airplane autopilots are basically just cruise control.

You still have a human in the loop double checking everything constantly and stepping in as soon as something isn’t routine (which is actually quite frequently).