Hacker Read top | best | new | newcomments | leaders | about | bookmarklet login

Firstly this is a genetic algorithm designed walk. If that is your criticism, it's easy to fix. Simply make the head feel pain, as an extra constraint. This algorithm is trying to find ways of moving the body it was given in a way that would work (like a human or any animal would. This algorithm is trying to be smart about moving around, it's NOT trying to imitate human animators. You want that, we have algorithms for that as well.

Second, is this really a valid criticism ? In a pinch, I have in fact used my head as an appendage to move around, to support things when moving furniture, and a million other ways so ... I don't see the problem. I have in fact moved furniture with a partially disabled person that used his head as an appendage to keep furniture stable just like I did. I always found that disabled people prefer that you just ignore their disabilities, as much as possible (unless it makes something truly impossible for them, but hard is just fine). Is that a wrong attitude ?



sort by: page size:

What are you talking about? Evolutionary algos in walking animations have been used far and wide. These guys just took out the rules for using a upright stance.

And criticizing something like that, as unsubstantial as he did, is just fear of new tech. Those zombie movements are great.


Yeah, sadly I think it's the body not walking for me. Guess I should try a wheelchair simulator :)

I don't think it's meant to be a 100% replacement of walking. Personally I think of it as a toy.

The ability to walk without falling over takes a lot of computation, even if it seems effortless.

Huh, I once made a workshop where people used evolutionary algorithms to make a similar shape learn to walk (see gif on top here https://github.com/Matsemann/walkingea ).

I would call walking hard - most animals are just very good at it.

No programmer is good enough to build a walking algorithm that is 1% as good.


> You treat the walk as the default

Is it unreasonable to treat “moving around with your body” as the default vs “moving around in a $x,000-x00,000 device”?


I find it extremely disappointing, when a group is attempting to innovate on a problem that is very clearly difficult, and someone comes along to say "I'm sorry, but this can't work."

The researchers are obviously aware of the challenges and working to overcome them. No one is confused to think that an omnidirectional-treadmill can capture the true physics of walking. The researchers are working to get it as close as possible.

Naysaying, in cases like this, is not a virtuous skill.


it's pretty obvious that it can't actually walk.

I hate to be the one to break to this to you, but your robotics teacher was off his or her rocker or was trying to make a much larger point with a bad example.

If walking was simply graceful falling, it would imply that once we begin the process of "falling" onto the next foot, we couldn't significantly affect the outcome. However, because I can start walking and then stop with a single leg half way between it's highest point and the ground, this explanation of walking as falling becomes very problematic.


>>> We know this phenomenon from everyday life. Most of us are experts on walking. However, if we try to articulate how we walk, we certainly give a description that does not capture the skills involved in walking.

All the article is based on statements like the previous one. Scientifically is possible to nullify it with a counterargument.

[1] Boston dynamics Atlas. https://youtu.be/rVlhMGQgDkY

[2] Boston dynamics parkour. https://youtu.be/_sBBaNYex3E

QED.


the articles say 4 directions, but I'm going to assume that's because it's what they've been able to find clear enough patterns to map actions to. If they thought up a few more, ex "think about kicking your left leg" to move forwards, it could probably work. The leader talked about using wheelchairs and prosthetic limbs.

Since we have machine-learning based walking models for several years now, see e.g. [1] or [2], I'm wondering why walking isn't yet a solved problem by now (?)

[1] https://www.youtube.com/watch?v=pgaEE27nsQw

[2] http://mrl.snu.ac.kr/research/ProjectScalable/Page.htm


The notion that there's anything inherently wrong with walking is really bizarre.

> until you think about it as 50% of your body weight tacked on

That's not how vectors work. It's only 11% extra weight. If its reasonably smooth, that's easy walking.


So how did you ever learn to walk, then? Every time you tried to move your leg, the fact that you moved forward, would be anecdotal and the data should be thrown out.

Bandwidth needed to actually walk is not so small (keeping balance, etc)

It's probably easier to put this in hardware than sending it to the brain and back.


> Your legs are moving but your head / vestibular system isn’t experiencing acceleration that it expects with those leg movements

That's also true on a treadmill. I think the bigger problem would be that the world appears to be moving (unlike a treadmill) and you don't feel any acceleration.


Over 99% of humans can walk. I don't think that's very "me me me".
next

Legal | privacy