Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Extreme parkour with legged robots

extreme-parkour.github.io

Post reply on HN
Print view View original

81–90 of 195 posts

Re: Extreme parkour with legged robots

#81

Post by digbybk »

mytailorisrich wrote: post #17
stavros wrote: post #9

Earlier quoted context omitted.

What's the Black Mirror scenario?

Oh, right. There is a Black Mirror episode in which dog-like AI robots are hunting down humans [1] If you watch it then those research legged robots will look rather more sinister... or maybe it's just me. That episode really stuck with me. [1] https://en.wikipedia.org/wiki/Metalhead_(Black_Mirror)

Iirc the robots are the equivalent to land mines, continuing to cause harm after the war is over. That episode felt very plausible and horrifying.

Re: Extreme parkour with legged robots

#83

Post by blagie »

sbierwagen wrote: post #8
082349872349872 wrote: post #2

there ought to be plenty of money in this for delivering explosive payloads... (Edit: although come to think of it, in the animal world, jump height is much more constrained by 9,81 m/s2 than by body size)

Artillery exists. Delivering explosive payloads is a solved problem. Detecting humans is a solved problem. The hard part is reliably distinguishing enemy solders from friendly soldiers or civilians.

As does counter-artillery. Fighting against a technologically-savvy opponent, you get off one shot and you're dead.

Re: Extreme parkour with legged robots

#84

Post by gwern »

ShamelessC wrote: post #78
Animats wrote: post #42

" A single neural net policy operating directly from a camera image, trained in simulation with large scale RL, can overcome imprecise sensing and actuation to output highly precise control behavior end-to-end. " So that's what it takes. That's so much simpler than the way Boston Dynamics does it, working out all the dynamics in simulation first. It's amazing to see this done from vision to actuators in one net. It's…

Doesn’t the use of RL here imply some level of simulation?

Yes, but the key is the RL simulation can be simple & unrealistic since it's used to train offline, not to plan & control live: a control systems approach like the Boston Dynamics requires excellent physics modeling because that's the only way it can do planning or adjust for errors while controlling the actual robot in real time, while the NN is trained to cover many possible physics during development, and at runtime in an actual robot, it just shrugs and adapts to whatever crazy scenario it finds itself in.

Re: Extreme parkour with legged robots

#85

Post by gwern »

lukevp wrote: post #79

It’s pretty incredible how animalistic its behaviors are becoming (hesitating for a moment at the edge of a lip, coiling it’s back legs for increased actuation, and when it almost misses the high jump and moves its back leg really fast multiple times to give itself a tiny boost each time until it recovers). Is that because it’s trained on animal behaviors? Or is this emergent behavior and animals just also do it more…

It trains from scratch (no imitation-learning or hand-written policies to bootstrap off of), so it's all emergent: https://extreme-parkour.github.io/resources/parkour.pdf#page...

Re: Extreme parkour with legged robots

#86

Post by huytersd »

nightpool wrote: post #3

As a reminder: CMU and its robotics program is intimately associated with the US Department of Defense, and although there are certainly civilian applications for this technology, one of the most likely short-term applications is improving legged drones for battlefield deployment (often, it should be said, in humanitarian roles, including search and rescue, or support roles, such as ammo, food, and water resupply, bu…

Good. I want the best minds in our country working on getting our boys out of harms way and robots doing the hard, dangerous and dirty work.

Re: Extreme parkour with legged robots

#88

Post by sbierwagen »

PoignardAzur wrote: post #47
sbierwagen wrote: post #8

Earlier quoted context omitted.

Artillery exists. Delivering explosive payloads is a solved problem. Detecting humans is a solved problem. The hard part is reliably distinguishing enemy solders from friendly soldiers or civilians.

I don't know, loitering munitions are pretty much bringing a renaissance to the field of blowing people up.

Drones are a solution to the targeting problem: the operator watches through a live video link, identifies the target, then deploys weapons. (Drops grenade/launches Hellfire/navigates into the enemy)

This is why, despite being technically feasible, we haven't seen drone swarms yet. You can easily build 1000 drones, but you can't have 1000 channels of live video over radio back to 1000 drone operators.

Re: Extreme parkour with legged robots

#90

Post by modeless »

TaylorAlexander wrote: post #87

[dupe] https://news.ycombinator.com/item?id=37823440

Oh this is automatically reposted from the second chance pool, I originally posted it yesterday before that other story. It's a shame it isn't getting more attention. IMO it deserves to be at the top of HN for hours. This research represents how the methods in robotics are changing quickly. I think people underestimate how capable robots will be in the near future.

Also the comments on that other story are just sad. I understand war is in the news but I hope it's not all we can think about! Also I'm surprised by all the people who think this is not impressive because "a 10 year old could do this" or whatever, that's actually ridiculous.

I guess HN just isn't the right community to appreciate real hard technology like this anymore. This makes me sad, especially because I don't know of anywhere else that's better.

Post reply on HN