Flight instructor here. The flying and driving environments are quite different, and what you need an “autodriver” to do is a bit different from an “autopilot.”
In a plane, you have to worry a lot more about your attitude, aka which way is up. This is the first thing we practice in flight school with 0-hour students, just flying straight ahead and keeping the airplane upright. This can be a challenge to do in low visibility environments such as in fog or clouds, or even at night in some circumstances, and your inner ears are compulsive liars the second you leave the ground, so you rely on your instruments when you can’t see, especially gyroscopic instruments such as an attitude indicator. This is largely what an autopilot takes over for from the human pilot, to relieve him of that constant low-level task to concentrate on other things.
Cars don’t have to worry about this so much; for normal highway driving any situation other than “all four wheels in contact with the road” is likely an unrecoverable emergency.
Navigation in a plane means keeping track of your position in 3D space relative to features on the Earth’s surface. What airspace are you in, what features on the ground are you flying over, where is the airport, where’s that really tall TV tower that’s around here? Important for finding your way back to the airport, preventing flight into terrain or obstacles, and keeping out of legal trouble. This can be accomplished with a variety of ways, many of which can integrate with an autopilot. Modern glass cockpit systems with fully integrated avionics can automate the navigation process as well, you can program in a course and the airplane can fly that course by itself, if appropriately equipped.
Navigation for cars is two separate problems; there’s the big picture question of “which road am I on? Do I take the next right? Where’s my exit?” which is a task that requires varying levels of precision from “you’re within this two mile stretch of road” to “you’re ten feet from the intersection.” And there’s the small picture question of “are we centered in the traffic lane?” which can have a required precision of inches. These are two different processes.
Anticollision, aka not crashing into other planes, is largely a procedural thing. We have certain best practices such as “eastbound traffic under IFR rules fly on the odd thousands, westbound traffic flies on the even thousands” so that oncoming traffic should be a thousand feet above or below you, that sort of thing, plus established traffic patterns and other standard or published routes of flight for high traffic areas. Under VFR conditions, pilots are expected to see and avoid each other. Under IFR conditions, that’s what air traffic control is for, who use a variety of techniques to sequence traffic to make sure no one is in the same place at the same altitude at the same time, anything from carefully keeping track of who is where to using radar systems, and increasingly a thing called ADS-B. There are also systems such as TCAS which are aircraft carried traffic detection electronics. Airplanes are kept fairly far apart via careful sequencing. There’s also not all that much else up there, not many pedestrians or cyclists thousands of feet in the air, wildlife and such can be a hazard but mostly during the departure and arrival phases of flight while relatively low. This is largely a human task; autopilots don’t respond to air traffic control and many don’t integrate with TCAS or ADS-B, this is the pilot’s job.
Cars are expected to whiz along mere inches apart via see and avoid. There is no equivalent to ATC on the roads, cars aren’t generally equipped with communication equipment beyond a couple blinking lights, and any kind of automated beacon for electronic detection absolutely is not the standard. Where roads cross at the same level some traffic control method such as traffic lights are used for some semblance of sequencing but in all conditions it requires visual see-and-avoid. Pedestrians, cyclists, wildlife and debris are constant collision threats during all phases of driving; deer bound across interstates all the time. This is very much a visual job, hell I’m not sure it could be done entirely with radar, it likely requires optical sensors/cameras. It’s also a lot more of the second-to-second workload of the driver. I honestly don’t see this task being fully automated with roads the way they are.
FiskFisk33@startrek.website 6 months ago
This is GPT.
After that intro I don’t trust a single word of what that site has to say.
tyler@programming.dev 6 months ago
Why in the world would you think that’s gpt? That’s not the normal style of gpt and it’s definitely the style of normal corporate sites.