Hacker News Reader

Comment by cdowns | Hacker News Reader

In General Aviation at least, there are varying levels of autopilot (depending on how nice the system is you can maintain heading and altitude or fly a pre-programmed course with navigational waypoints), but all of them will fly you right into the side of a mountain or into restricted airspace if you let them. You can't just set it and stop paying attention. It seems that is what people want when they think of "autopilot" for their cars.

View on HN

If that is the case, autopilot on a plane seems to be more like cruise control expanded to maintain pitch, yaw, and roll than autonomous flying.

There is tremendous variability, but this is basically correct. In addition, the sky is wide open and empty compared to even a not so busy street.

Airplane autopilots really are solving a much simpler problem than even a level 3 autonomous car has to solve.

The most advanced autopilots can automatically takeoff and land, but that requires substantial ground equipment at each airport to support.

+1, Instead of developing super smarts autopilots we could just instrument the highways so they can tell the car where to go. The instrumented roads could send out warnings and give control back to the driver when there are roadworks ahead.

That would require a massive infrastructure investment. And seeing as we can't seem to even get a good chunk of roads to even be in good repair, I don't have high expectations for the possibility of instrumented roads.

No more massive than it takes to maintain these roads to begin with.

So be it. We already have extreme amounts of signage for traditional drivers, it's time we stop treating autonomous cars like second rate vehicles and start providing detailed surveys of roadway boundaries and all of the signage that drivers rely on to autonomous vehicles in a manner suited to them instead of a human.

You say that it would take a massive investment but realistically existing signage, reflectors, and road markings probably cost more than the equivalent for an autonomous vehicle would.

GM is driving around with LIDAR to create very precise maps of specific highways and their cars use those detailed maps to navigate.

And what happens when the lanes are shifted for road construction? Killing a driver is bad enough, killing an innocent construction worker would be marginally worse.

Those maps will get outdated with construction, etc.

If we were willing to do that, we could have had self-driving cars decades ago. They were showing off systems with magnetic lane markers on the Discovery Channel when I was a kid.

The exciting thing about self-driving cars this time was that they worked on existing roads, making them financially feasible.

Key thing that unlike Tesla's self driving system. just about nothing can happen in a plane that will kill you if don't disable the auto pilot in 1-2 seconds.

Perhaps final stages of ILS approach? That'd be the exception, but pilots are trained and certified for this - and definitely not allowed to snooze off then.

(I'm not a pilot, but have spent a fair amount of time and money on various PC flight sims; edit: and used to drive a Mazda 3 with a very good radar cruise control and AEB, before ditching it for a bicycle because getting old and fat).

I'm a pilot and wouldn't consider them autonomous. In the most sophisticated systems, you can plug in a fairly complex route to destination but there's a disconnect in automaticity when transitioning from cruise to approach. There are standard terminal arrivals, but they all assume contact with ATC for specifics, and in the case of lost communication in instrument flight you're expected to use your best judgment for ambiguous sections of the route clearance. Autopilots don't have judgment, and they also can't do two way communications either, so they're not able to accept ATC clearances directly, a pilot has to do that. So yeah not autonomous.

Reply on news.ycombinator.com