Let's Make Robots!

DARPA's Pet-Proto walking robot navigates obstacles

In this new video released by DARPA, the Pet-Proto autonomously navigates a similated urban disaster environment. It can step up onto ledges, jump off landing on two feet, decide how to cross an open gap in the floor and fairly quickly scale stairs. Impressive.


Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Yeah but this is the prototype of their new model so that is why it is teathered. They eventually make them self sustained. Here's their previous model. It is teathered as well but clearly further along. 


What occurred to me when Lady Ada posted this is that so far the kinematics are impressive but this isn't going to be very useful unless the collapsed building you're trapped in comes with a humanoid tether and harness. Remember ASIMO and the stairs?

sorry that's a flawed comparison.

there's a lot more going on here than just better Kinematics because for all its hype Asimo isn't much different to those servo based humanoid toys just bigger. Fell off those steps because it didn't even know the step existed or the rest of the world for that matter it was just following a script. And On its best day if asimo tried to do the jump in the video it would break its gear train for its motors.

Atlas uses series elastic actuators (SEA), like big dog uses, they let the bot know the force on the joints which where a lot of that "impressive Kinematics" feel comes from. They lets atlas uses force control instead of only position control which is far closer to how a human does it. SEAs are also the reason it can do drops like that without destroying itself, the spring take the shock.

in the end asimo is little more then a toy they can never walk around in the real world.

Asimo is little more then a toy they can never walk around in the real world

Apparently neither can this. Take it off the tether or it's basically as impressive as the wired gundam walker my nephew has and hardly up to what a Wowwee toy can accomplish in autonomous mode. I'm not saying the technology isn't going somewhere, just that with the training wheels on that it's not demonstrating any real utility.

"Take it off the tether”

this is where I'm lost in this debate because your making the tether in this video out to be some kind drawback when it's just a safety measure much like the safely line on rock climbing walls. There’s a power line there but that's just so it doesn't have to run a noisy generator like big dog. Outside of that the tethers isn't holding the bot up in any way or helping it with the obstacles just there to catch it if it falls.

Just because it CAN take a fall doesn’t mean you don’t want to protect in while testing. 

I undesrstand protecting the investment, but isn't that another way of saying it's either not hearty enough to take the fall or that you're expecting its programming or hardware to fail? Alternately stated, if the point of this video is demonstration or proof of concept, the safety aspect negates both. What would you make of a parent who bragged that his child was a prodigy because he could walk before anyone else in his age category but who none the less never let the kid out of the walker to prove it? Also, it's my understanding that the processing that allows the complex kinematics and the problem-solving necessary to negotiate these complex situations isn't done in on-board processing, though whether the Telemetry is wired or not I can't tell and don't care. It's a technological achievement, yes, but it just doesn't represent an impressive level of utility to me yet.

What's interesting is this same video is on Boston Dynamics site (they actually built it). They call it the Atlas-Proto while Darpa calls it the Pet-Proto.