Autonomous Systems

Would you ever own an autonomous vehicle?

  • Hell yeah Bring on our robotic overlords!

  • Fuck you! I'll keep my Indepenence


Results are only viewable after voting.

khorum

Murder Apologist
24,338
81,343
174d 4h 36m
New video for Agility's Digit V2 biped, it's the latest version of the van-to-door delivery system that Ford is trying to sell to Amazon. A lot of folks have been using it too but it gets nowhere near the publicity Boston Dynamics does.

In the video they play no music or any background sound effects to show off how quiet it is. Apparently BD's Atlas is loud as FUCK.

 
  • 2Like
  • 1Solidarity
Reactions: Ukerric, Tuco and Void

Captain Suave

Caesar si viveret, ad remum dareris.
2,002
1,768
64d 6h 40m
I doubt they'd use the drones for anything but the most remote destinations tho. If the recipient is in an urban area it's virtually guaranteed to be more efficient to use a van because of the densities involved.

I think the idea is human time savings, where density would be a good thing. Instead of a human making 20 stops in serial over half an hour, you could have the delivery van drive to a location and a battery of drones would detach deliver light packages to a dozen houses at once in the space of a couple minutes, and then they recharge off a bulk battery while the van drives to a new hot spot.
 

khorum

Murder Apologist
24,338
81,343
174d 4h 36m
I think the idea is human time savings. You could have the delivery van drive to a location and a battery of drones would deliver light packages to a dozen houses at once in the space of a couple minutes, and then they recharge off a bulk battery while the van drives to a new hot spot.
Well their first location is the middle of bumfuck Virginia so I dunno. There are practical challenges with urban deployment though, not the least of which being the FAA.
 

Captain Suave

Caesar si viveret, ad remum dareris.
2,002
1,768
64d 6h 40m
Well their first location is the middle of bumfuck Virginia so I dunno.

You're right, that's odd. Maybe it's a compromise for the purpose of authorization for testing and development? Last I'd read was one of their end-state ideas was along the lines of this, possibly with mobile drone hubs.

 

Tuco

I got Tuco'd!
<Gold Donor>
41,680
66,991
92d 6h 11m
Delivery is a tough problem because you're competing against low wage drivers who will jump out of the car, grab the box and run the box up stairs and around whatever obstacles are present without too many instances of failure to deliver due to complex geometries or whatever. Sure you get some people who will take a shit on someone's lawn, but that's the exception and it's not a systemic problem that the delivery company would be responsible for.

It's the same for virtually any automation problem: the less constrained the problem space is the harder it is to do better, cheaper and more reliably than a human.

It's still gon' happen though, it's just hard.
 
  • 2Like
Reactions: Flobee and khorum

khorum

Murder Apologist
24,338
81,343
174d 4h 36m
Delivery is a tough problem because you're competing against low wage drivers who will jump out of the car, grab the box and run the box up stairs and around whatever obstacles are present without too many instances of failure to deliver due to complex geometries or whatever. Sure you get some people who will take a shit on someone's lawn, but that's the exception and it's not a systemic problem that the delivery company would be responsible for.

It's the same for virtually any automation problem: the less constrained the problem space is the harder it is to do better, cheaper and more reliably than a human.

It's still gon' happen though, it's just hard.
Yep even when they get a self-driving delivery truck it'll be cheaper to get some shmuck to just make the door to door deliveries. The use case for robot loaders may exist for specific package classes eventually though.

As to drone deliveries, that's basically a net win now if the FAA can establish airspace regulations urban airspaces. They'd have to design distribution centers around the drone's ranges, and there's already a viable proof of concept for extending THAT:

 

Cad

scientia potentia est
<Bronze Donator>
21,048
28,788
94d 4h 37m

That's kind of an intractable problem, they drew the lane right into the wall. It "should" auto lane change over to avoid the wall (which it undoubtedly detects via radar and camera) but it is not programmed to do this yet. It relies on the human operator to intervene.

This is not meant to be a self-driving feature. Yet.
 
  • 1Like
Reactions: Seananigans

Tuco

I got Tuco'd!
<Gold Donor>
41,680
66,991
92d 6h 11m
That's kind of an intractable problem, they drew the lane right into the wall. It "should" auto lane change over to avoid the wall (which it undoubtedly detects via radar and camera) but it is not programmed to do this yet. It relies on the human operator to intervene.

This is not meant to be a self-driving feature. Yet.
1. I'm not sure what you mean by intractable, but this is a very tractable problem, as in, the tech to detect the wall is a solved problem and they just need to improve the software to account for it.
2. You see a lot of these kinds of vids from tesla because they have the most human operated systems, but also because their choice to not use lidar will both cause these problems and cause people to showcase these problems. Radar and camera are great of course, but this is a problem very well suited to lidar.
 
  • 1Like
Reactions: Captain Suave

Cad

scientia potentia est
<Bronze Donator>
21,048
28,788
94d 4h 37m
1. I'm not sure what you mean by intractable, but this is a very tractable problem, as in, the tech to detect the wall is a solved problem and they just need to improve the software to account for it.
2. You see a lot of these kinds of vids from tesla because they have the most human operated systems, but also because their choice to not use lidar will both cause these problems and cause people to showcase these problems. Radar and camera are great of course, but this is a problem very well suited to lidar.

It's intractable (with the current software) because it is not coded to change lanes or evade in that fashion. Intractable is probably the wrong word. If you draw the road directly into a wall, the current Tesla autopilot is going to follow it directly into the wall. I would guess the reason they don't have it coded to evade detected obstacles is because this would cause more problems than it solves (right now, anyway) with false positives and Teslas swerving all over the place.

If the camera + radar was detecting the wall, then it's not a lidar problem. it's just not coded to evade outside the lane. It's coded to follow the lanes.

Anybody who has actually driven a Tesla on autopilot knows this and takes over in construction zones or when the lanes start to look janky.
 
  • 2Like
Reactions: Seananigans and Tuco

Seananigans

Pistol-whipped
<Bronze Donator>
4,345
6,479
112d 4h 16m
It's intractable (with the current software) because it is not coded to change lanes or evade in that fashion. Intractable is probably the wrong word. If you draw the road directly into a wall, the current Tesla autopilot is going to follow it directly into the wall. I would guess the reason they don't have it coded to evade detected obstacles is because this would cause more problems than it solves (right now, anyway) with false positives and Teslas swerving all over the place.

If the camera + radar was detecting the wall, then it's not a lidar problem. it's just not coded to evade outside the lane. It's coded to follow the lanes.

Anybody who has actually driven a Tesla on autopilot knows this and takes over in construction zones or when the lanes start to look janky.

Yeah I’m not sure what the issue is. That video is yet another instance of an idiot driving a Tesla improperly.

Also, Tesla autopilot confirmed coded by ACME and tested by Wile E Coyote.