Walking robot navigates bumpy ground (w/ Video)

(PhysOrg.com) -- A robot named MABEL with a human-like gait can walk over rough terrain in University of Michigan electrical engineering professor Jessy Grizzle's lab.

Grizzle and his students Hae-Won Park and Koushil Sreenath posted a YouTube video of MABEL's first steps over wooden planks, and within five days it was viewed more than 84,000 times.

(Spoiler alert!) It doesn't have a happy ending. MABEL breaks a leg and doubles over, but not before she clears planks stacked to 2.5 inches high.

"In the video, MABEL falls, but does not trip," Grizzle said. "Each shin has a built-in mechanical fuse that gives way when the load on the leg exceeds a threshold. This is done to avoid damage to the bearings in the hips. In the experiment, the fuse in MABEL's right shin gave way. The resulting fall is quite spectacular, but does not permanently damage the robot. It takes about an hour to reassemble the leg."

It was the researchers' intent to test MABEL's limits. Grizzle was surprised that she was able to perform as well as she did. What a can step over usually depends on what it can see, but MABEL is blind. She has no camera. The engineers had merely programmed her legs to swing higher to step over or onto obstacles. They hadn't programmed her specifically to navigate them.

That's what they'll do in future, more scientific experiments. The researchers are developing feedback control algorithms to give bipedal robots a strong sense of balance.
This early experiment was, for MABEL, the equivalent of walking confidently down a forest trail while talking to a friend—and not looking down to watch your steps.

"Humans rely on a keen sense of balance to pull this off," Grizzle said. "The challenge for engineers is to design this ability into robots."

Grizzle believes MABEL will eventually be able to gracefully step from a stair-height of about 7.5 inches.

So what's the point of all this? Grizzle says more than 70 percent of the Earth's land surface isn't navigable by wheeled or tracked vehicles. Bipedal robots could be capable of traveling in these places.

"If robots are going to perform rescue operations, they're not going to be walking on a nice linoleum floor," Grizzle said. "Even if they're in a building, it's going to be littered with things that have fallen. They have to be able to walk on an uneven surface without tripping."

Grizzle's ultimate goal for MABEL is to make her run. It's a feat he almost achieved with her predecessor, RABBIT, but RABBIT's heavy legs carried motors and no springs. MABEL's motors are in her torso, which leaves her legs light and agile. Her legs also have springs that act like tendons, storing and releasing energy.

"Running would really be one of the highest demonstrations of agile behavior in a machine. It takes such delicate feedback control," Grizzle said. "RABBIT failed to run in steady state. We got six really beautiful steps, but we never got seven or eight."

He hopes to see MABEL jog by winter.

Grizzle will present this research at MIT's Dynamic Walking 2010 conference on July 8 in Boston.

Explore further

Cockroaches offer inspiration for running robots

Citation: Walking robot navigates bumpy ground (w/ Video) (2010, May 28) retrieved 18 October 2019 from https://phys.org/news/2010-05-robot-bumpy-ground-video.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

Feedback to editors

User comments

May 28, 2010
Ok, better than nothing I guess, but this is really pathetic when you think about it.

It's got a stabilizer attached to a central pole. This can hardly even be considered "walking".

The other thing is that with the price of cameras being so cheap, I don't see why not make a walking robot that actually does use the cameras to navigate. You can give the thing eyes that look forward as well as downward at different angles to navigate perfectly. Cameras of sufficient quality are now so cheap they are made into disposable cell phones that are sold for $19.99.

Also, this foot design is terrible. There is no bipedal organism that has this foot design or anything similar. The best a team of university professors could come up with was peg legs? You gotta be joking...

I think they need to take a look at the hinds feet of cats and dogs, and especially the padding on their feet, or chickens. This allows a more gentle and smooth weight transfer, and a wider foot allows better balance.

May 28, 2010
Quantum_Conundrum, you are missing the point of the experiment. The team is isolating one aspect (balance through physical feedback) of robotic walking in order to analyze and improve it. Thus, this is an initial experiment in the design process of giving a robot human-like balance. Taking away the stabalizer, giving it a camera, or modifying its feet would defeat the purpose as those things have already been done and they add unnecessary complications to the project at this stage of development.

You have to break problems down into their component parts to effectively address specific issues. They have a specific issue of balance to address, and have designed an experiment that addresses that specific issue effectively for the current iteration of the project.

May 28, 2010
You have to break problems down into their component parts to effectively address specific issues. They have a specific issue of balance to address, and have designed an experiment that addresses that specific issue effectively for the current iteration of the project.

But I disagree. Even in BLIND humans, balance is not based simply on "touch". All humans have pressure air-pressure sensors in their inner ears, like virtually all macroscopic life on earth, which allows them to orient themselves even if they are blind, and being deaf doesn't necessarily effect this balance mechanism either.

So making a robot walk with no sensory input comparable to animals or humans is increasing the difficulty of the problem for no reason.

As humans, when we walk we are continually subconsciously evaluating our surroundings and planning ahead. We learn to do this so automatically that we don't give it much conscious thought. That is different than a blind robot stumbling around...

May 28, 2010
It's not for no reason. By making a robot balance without visual input they will be able to create a far better machine when all aspects (including visual input) are incorporated back into the design.

They are not trying to create a blind robot, as you seem to be assuming. They are merely isolating one factor in order to improve it.

May 28, 2010
After the leg breaks you can see that the bot is just being dragged along by the central pole. So the bot is not really walking. It is on a roundabout and just raising its legs.

May 28, 2010
Grizzle and his students should apologize for having wasted so much valuable research funding.
If this is the best they can do they should be restricted to Meccano or Lego sets.

May 28, 2010
They've got to be joking! Back in 2008, we already had THIS:


May 28, 2010
Why is it that videos never play for me?

May 28, 2010
Try YouTube; just search for "Boston Dynamics Dog".

Or generally speaking, get a browser with Flash support built-in, or download/install a Flash plug-in.

May 29, 2010
yeah that dog is tonnes more sophisticated and yet looks much cheaper as well.

May 29, 2010
They've got to be joking! Back in 2008, we already had THIS:


Yea that Dog is Freeking cool and imagine it at 200 feet tall with guns attached!

May 29, 2010
The Dog is absolutely amazing. They can even make it JUMP!

I think a six-legged design using similar technologies would be more stable, but that was absolutely amazing. It survived a very hard, random kick on partially wet pavement, walked up a snow-covered hill, and recovered from a nasty slip all while carrying a payload 50% heavier than itself.

I don't know of any human or animal that could have done that job, and not even a commercial ATV could have scaled the cliff, unless it had modified tires. And finally, a wheeled robot of similar scale definitely could not have climbed directly over the cinder block pile...

Oh yeah, "computations" of multi-threaded sensory input won't be an issue.

You can now buy a AMD 6 core, 2.8ghz processor with a total of 1.5MB L2 cache and 6MB L3 cache from New Egg for only $199. The intel 6 core processor is reportedly several times more powerful than a dual Xeon system for multi-threaded rendering.

Jun 08, 2010
First off, this robot design is bipedal and the dog thing is quadrupedal. Second goats, which navigate mountain passes with ease, have hooves which are like these feet. Its a great experiment to develop the complex algorithms needed to develop balance for a robot.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more