Boston Dynamics : Atlas | Partners in Parkour

Boston Dynamics have released a new video of their Robocop prototype , Atlas ,as it goes though its its Parkour routine with its brother ... Atlas.

The back flips are mind-bending.

How Boston Dynamics put the video together.


You know you're in the future when rockets can land and robots make you feel uneasy.

a reply to: gortex

Creepy.

I can't even do a backflip.

Was this controlled remotely or did they do it all on their own?

a reply to: gortex

Yup, it's all over.

Just waiting for them to make it official and launch a T-101.

a reply to: chris_stibrany

I believe they were programmed to run the course , the second video shows a man running the course I guess that was part of the programming.
Give it 10 years they'll be doing it themselves.

Starting to resemble those droids from the movie Elysium but with awesome back-flipping action. So is that how its all going to end, robot ninjas?

But on the other hand if these robots can be utilized like ants there no end to the things we can accomplish.

originally posted by: gortex
a reply to: chris_stibrany

I believe they were programmed to run the course , the second video shows a man running the course I guess that was part of the programming.
Give it 10 years they'll be doing it themselves.

That's what I was wondering. If the entire course is preprogrammed. However, it is still amazing and scary at the same time.

I can only hope Isaac Asimov's three laws become a reality...

Anyone that doesn't know, famed science fiction writer Isaac Asimov proposed three laws of robotics in his stories from 50 years ago.

A robot may not injure a human being or, through inaction, allow a human being to come to harm.
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws.

"Welcome to Westworld where nothing can go wrong...wrong...wrooooonng..."

Incredible times. I remember when I thought the future would be here when we had Dick Tracy's watch.

In other news, I hear we're growing human ears on mice and don't even ask what the Chinese are doing. And that's not even talking about gain of function weaponized coronavirus.

Humanitarian science is a good thing. More of that and less weaponizing it.

Personally I think we will annihilate ourselves with science in service of the warmongers long before any of most of the other doom scenarios.

The second video explains how useless they are in the real world. If the course is altered in the slightest, they fail. It is all programmed based on repeatability.

If all else fails, use aluminum spray paint. They will be blind and lose networking. Or in this case, scoot the obstacles around a little and watch them crash.

a reply to: [post=26060130]chris_stibrany[champion on.


Per the segment shown some time ago on "60 Minutes", the robots are capable of calculating how to adjust and balance themselves depending upon the speed and surface they are traversing. Their path through the obstacle course is determined by the operator and programmed into the robot.

Similar to you being told to run up to those stairs, climb them, jump off the top, roll when you hit the ground and then run back to the starting point. You've been giving the instructions, the program, to run, but you have to use your learned skills and abilities to carry out the program and run the course.


It will be left for much more powerful and sophisticated AI's in years to come, to determine, for themselves, the optimized parkour course to follow.

Or whether such antics are worthy of their concern at all!

Talk about uncanny valley!

'Hairforceone' (agh), I don't think you know what 'uncanny valley' means.

It has to do with 'resemblance' not about movements or such. These robots don't resemble human beings enough for 'uncanny valley' effect.

To talk about the videos themselves; those movements look really weird, like they're not really happening, like they're just rendered animations not done quite right, or something.

I am not saying it's CGI, I am simply saying something about their movement is SO different from anything 'natural' (maybe same way as pixels are completely different method of producing pictures than what the eyes naturally do), that it just looks 'unrealistic', for the lack of better term. I have done some animations in my time, so my eye is keen in noticing 'small mistakes' in animation, like slightly too fast or slow leg movement compared to the movement of the whole object in relation to the ground and such (can look like 'slipping' or 'sliding').

I am not saying this is what I am seeing, but there's an eerie similarity - it's like something is happening between my eyes and my brains that alert me, as if this is not perfectly done animation or something. I am sure it's due to the way the robot legs are designed to move - all these robots move really differently from humans or even animals, they don't resemble something we've seen a lot before, so maybe that creates this odd effect when looking at their movements.

Ever since Asimo, I have always wondered why do the robots have to move like they're constipated or have a great urge that's about to explode.. why are their legs always somewhat bent, and why haven't we yet seen a robot that walks the same way as human beings do?

Is that some sort of 'failsafe' so they wouldn't fall so easily or what? If a human being walked like that, they'd look like they're trying to do Tai Chi but failing.. Are we ever going to see a robot walk the same way as people do? Or even animals?

The four-legged robots don't really move as fluidly and naturally as animals, either, not even some clumsy dog. Let alone some kind of graceful feline or fast-running horse.

I am suspecting the weird shape, heavy weight and also the gravity point causes some of why it looks so odd, but it looks almost unrealistic how they can just so weirdly and slowly jump around tilted surfaces and not fall, tilt or slide at all.

a reply to: Mantiss2021

..how to adjust and balance themselves depending upon the speed and surface they are traversing.

This needs a bit of clarification and explanation. Do you mean their own speed? How do they measure it?

Also, how do they know what surface they're 'traversing'? It's not like it's a continuous, one surface, either, it can vary with each step. How do they analyze and know what the surface under their leg is at any given moment?

To add, you can't analyze a surface fully or thoroughly by sight alone, do they use radar or some kind of heat sensor or what? Something can look like something else, it can reflect sensor signals, different lighting can make things look different, etc.

Heck, just SNOW can appear the same but be a dozen different textures, ranging from slippery to non-slippery and soft to rock hard. There can be soft snow on top of hard ice, so how does a robot analyze and know this multi-layered surface is exactly what it is, when even humans can't do that (and consequently slip, fall and hurt themselves)?

So your statement can't be factually true, if the robots can't know the surface completely and consistently (and I think this is impossible).

Maybe they're balancing themselves based on what they PROJECT or ANALYZE the surface to be, and then they can err and slip when they're wrong..?

But surely they can't just analyze any surface and balance themselves without fail - there's BOUND to be surface they misjudge or analyze wrong - would be neat to know _how_ they analyze a surface anyway, because that has got to be one of the trickiest things to do. They would have to know just how slippery or non-slippery, solid or soft, breakable or durable, etc. a surface is before stepping on it. What robot can do that, and how?

Think about glass surface that's painted black. It might not be slippery, but the robot would fall through if they stepped on it. The only way to really analyze it, would be to touch the surface, knock it, listen to the sound, and such, and still they couldn't really know how thick it is or if they can hold the robot's weight. Can these robots magically evaluate thickness and durability of a black (or any color) surface? There are also so many different glass types, from armoured to very weak, from super thick to super thin, how does the robot know all this?

I don't think your statement can be true, sorry.

a reply to: Shoujikina

Pressure sensor feedback. Accelerometers. Gyrostabilization. Plus the usual digitized visual suite.

Your example of the black painted sheet of glass really isn't relevant to the question of surface analysis, since the failure mode you suggest has nothing to do with a mis-analyis of the "surface" itself, but of the ability of the material underlying the surface to support the weight of the robot. In your given example, a human would be just as likely to fall through the glass plate as a robot, given a sheet of glass incapable of supporting the humans weight.

That being said, both humans and these robots must learn their moves by trial and error. That was made quite clear, for the robots, in the "60 Minutes" segment I referred to in my first post.