The latest version of the LittleDog robot | Hacker News Reader

The latest version of the LittleDog robot

youtube.com

173 pointspcof61 comments

I find amazing how difficult it is to successfully perform apparently trivial tasks such as walking.

There's more information about the 'LittleDog' on the Boston Dynamics website: http://www.bostondynamics.com/

I was really impressed 2 years ago by their video of the 'BigDog': http://www.youtube.com/watch?v=W1czBcnX1Ww (check out t=40s)

Yes, the way it recovered from that kick is incredible.

And, if you'll forgive me, here's one additional link for anyone who hadn't yet heard of Big Dog. (Unlikely, I know...) http://www.youtube.com/watch?v=VXJZVZFRFJc

If you saw the original Boston Dynamics BigDog Youtube video, should check out the version that researchers here in tokyo have come up with:

http://www.youtube.com/watch?v=X5BPrlOVX2o

I find it amazing HN readers believe walking is a trivial task.

The key word being "apparently".

I don't even think it's that apparent. It takes us 2 years to learn to walk, even with millions upon millions of years of evolution designing us perfectly to do it.

Yes, but many four-legged animals can walk mere minutes after having been born, and this is a more valid comparison for the task facing a four-legged robot.

>even with millions upon millions of years of evolution designing us perfectly to do it.

Birds can walk on two legs in very little time.

Anyone with some robotics experience able to explain more about the floating base inverse dynamics control system and what sort of accommodations make it so much more powerful than PD?

I don't know too many specifics, but it looks like their PD example was controlling the joint position. That means that if the ground isn't where the robot expects it to be, the leg will still assume the same position and cause the robot to fall over.

The dynamic force control system is, instead of controlling the leg position, controlling the amount of force that the leg is exerting on the ground. This means that the leg may end up in a different position than the walking controller intended, but the overall effect on the main body should be similar to what was expected.

Is it just me or does that thing look like a headcrab: http://upload.wikimedia.org/wikipedia/en/5/54/Headcrab.jpg

I thought it looked very insectile.

Reminded me of the houndeyes from Half-Life, with an extra leg

Its been said before, but needs to be repeated every time:

The implications of this line of R&D are horrific.

Removing the risk of say, US soldier casualties, combined with media control, and with "merely" the current level of inhumanity of political leadership, would be a very ugly mix indeed.

So yeah, its hard for me to get excited about the tech in this case.

I know I probably should be worried about this kind of technology, but I just can't help it - I see this and I'm excited like a little kid again. I want one of these :)

Until we devise a system where consumer products, or government grants, drive more research like this, the military war machine is necessary for this type of funding. Most public company's with deep pockets live quarter to quarter. Gotta beat those earnings estimates, by a lot.

They're only horrific if they're misused. Robots like this will also have plenty of non-military applications - especially in agriculture. Legged locomotion opens up the possibility of cultivating land which was previously inaccessible to conventional farming automation.

It should also be said that there is nothing inevitable about the close association between robotics and the military - it's just an artefact of American culture and the idiosyncratic way in which high tech research gets funded in America.

They're only horrific if they're misused.

Given that this whole line of research is funded by the US military, what do you mean by "misused" ? Wouldn't "used" be more accurate?

Under "misuse" I would include things like:

Use of robots for the purpose of prosecuting a war crime

Use of robots to carry out an act of terrorism

Indiscriminate use of deadly force by autonomous robots

Use of autonomous robots in a military context without proper supervision (for example, not being able to deactivate them once deployed).

Perhaps you will change your mind after viewing this horrific weaponized version of BigDog:

http://www.youtube.com/watch?v=ptyV1cpE14o

Unlike the A-bomb which is lovely.

Or the decidedly low-tech genocide in Sudan. Or the genocides and crimes against humanity in all the other places.

Bottom line is Homo homini lupus. It's not like the risk of US soldier casualties is that much of deterrent.

Bombs, A-bombs, robots, not that different, the more things change, the more they stay the same.

So existence of A-bomb or low-tech genocide somehow justifies creation of new means to kill people?

Sure. It'll be more transparent at least. And there might even be some due process eventually.

It doesn't justify anything. It just shows that we are already killing people in almost every way imaginable so another new one would not be anything all that new.

The same tech can also walk into a burning building to locate trapped people without endangering the lives of firemen, crawl through rubble to find survivors after an earthquake, power household robots to assist the disabled, and more. I know the source of funding here is distasteful, but this is a dual-use technology if ever I saw one.

This argument is moot.

As long as humanity will want to kill eachother, everything will be used as weapons. Computers are used as weapons (hell would computerized bombs and missiles be possible otherwise?) boats are used to kill people. Airplanes used to kill people. So now robots, why not its the next step. Knowing the US government they will manufature them without any bit of encryption and with a radioshack remote control someone can hijack one of these :P

Don't worry, we're much farther from that than you might think from this video. This robot can't see where it's going. It requires external cameras watching it from known fixed locations to figure out where it is. Not only that, but the cameras cannot look at a piece of terrain and figure out the 3D structure; the terrain must be mapped out in advance as a 3D mesh, down to the millimeter.

Agreed. Computer vision is a steep challenge and will limit autonomy. The derivatives of these types of devices will likely be deployed as drones in the meantime.

We are rushing headlong into robotic automation. Civilian and military uses will abound. The day when hiding our head from technology was a solution is long past. Responsibility is the only option left that doesn't involve war. Our "leaders" will only be as insane as we allow them. People don't get the government they deserve; we get the one we'll tolerate.

I suspect that if you were rescued from a collapsed building by a descendant of this robot, you might possibly change your mind!

The most effective thing to get me off my ass and studying is to see something I should conceivably be able to create, but have no idea where I'd begin.

Can someone link to a good introduction to how the control systems for something like this work?

You should take in a couple of lectures of the "Introduction to Robotics" class on iTunes U. http://deimos3.apple.com/WebObjects/Core.woa/Feed/itunes.sta...

I want one of these to walk my dog at 6am on Sunday when it's raining.

until your dog falls in love with it ;-)

then maybe I'd have little cyborg dog puppies.. how awesome would that be??

Why can't it move 10x or 100x faster? Is speed constrained by computation or something else?

I'd think the physics would change dramatically. It's probably a brand new problem to have it work at higher speeds.

Pertinent facts from one of their papers: http://www-clmc.usc.edu/publications/K/kalakrishnan-IROS2009...

The robot drives its motors at 400 Hz based on control signals coming in at 100 Hz from an external computer. 100 Hz is probably the limit of the sensing system they're using, which is a motion capture system external to the robot. (Notice the reflective marker balls on top of the robot and also on the terrain.) My guess is that's the limiting factor here, though supporting a higher rate would probably also require faster software, and I have no idea what the limb speed/precision limits of the robot are.

I think the proverb "you have to learn to walk before you can learn to run" applies quite literally here.

That is only a matter of time. Then we will have the "rat thing" from Neil Stephenson's Snow Crash. It was scary to watch the video.

Very impressive. Sci-fi stuff.

I just found the big stair demo cheap. That "jumping" behavior looked hard-coded (and it was implied by the narrator) and seemed to possibly damage the robot. Still impressive, but a lot less so than the other demonstrations.

So inspiring. Reminds me of the Disney movie "The Brave Little Toaster":

http://www.youtube.com/watch?v=NbBEfZ8fK9Y

I like how they end the video with the perspective of what it will look like when you're being chased by one of these guys.

Indeed. Imagine an 8 foot tall version coming at you. Tanks of the future?

Immagine they are not controlled by humans. Even more scary.

Imagine if they could efficiently vacuum your floor and mow your lawn. We're still a long way from robot world domination.

The "funny" thing is that it's probably easier to make a robot-killer than a robot that will cook me a variety of healthy meals.

Yeah, after you have motion down, you would just need to be able to detect a heat signature and shoot at it. I doubt that's very difficult compared to what they've already done.

There are already robots for that. Not so fancy, but they somewhat work.

I wasn't speaking about "robot world domination". That implies intelligent machines and this is not. But it looks like this one is pretty efficient in recognizing obstacles, and shapes. Once they can recognize targets, humans just need to deploy them and they will be much more efficient, without the need of human pilots. And they can come in different sizes and shapes.

I find it quite scary, and the DARPA logo at the end of the movie does not help.

DARPA is the reason you are interacting with other people on this website. The logo doesn't automatically mean "to be used for evil".

Or even better, attack dogs of the future. When I got home last night the first thing my roommate did was show me that video and then remind me of the "rat things" in Snow Crash.

http://en.wikipedia.org/wiki/Snow_Crash#Rat_Things

Next advancements of this technology:

- running

- jumping

- seek and destroy

Definitely. I will add it to my list: http://creepyrobots.tumblr.com/

OMG, is the laundry bot for real? Want!

honestly this is just frightening.

I personally find unseen UAV's raining missiles down on our heads to be much more terrifying. I'm actually quite excited about this tech. The first thing I thought when I saw this video was one of these things walking on the surface of Mars.

Agreed re: the UAV comment. Imagine a world where big brother is watching everything you do, and can kill you from the other side of the world with the press of a button, with absolutely no warning, with no way for you to defend yourself, and no way to attempt to explain your actions to someone.

A walking robot is much less terrifying.

The first thing I thought was that countries such as Cambodia could one day exist without landmines.

Reply on news.ycombinator.com