Hexapod Figures Out How to Walk After You Chop Its Leg Off

It's going to take more than a missing leg to slow this robot down

2 min read
Hexapod Figures Out How to Walk After You Chop Its Leg Off

If the movies have taught us anything, it's that chopping a futuristic death robot's leg off does not significantly diminish its capacity to hunt you down. Want to know where that capacity for being utterly unstoppable came from? It's this, right here.

After just 20 minutes worth of iterations, in the above video example the robot has come up with a five-legged gait that moves it along at 18 cm/s, as compared to the undamaged 26 cm/s gait. Not bad, considering that an unmodified five-legged gait had it limping along at just 8 cm/s.

The cool bit about this recovery model is that it doesn't require any specific information about what parts are malfunctioning or missing. Instead, it's just got a known model of how it's supposed to work, and if the actual performance that it measures is less efficient, it starts searching for new behaviors. To get all sciencey about it: "the robot will thus be able to sustain a functioning behavior when damage occurs by learning to avoid behaviors that it is unable to achieve in the real world."

Things get even more interesting when the authors of this paper (Sylvain Koos, Antoine Cully and Jean-Baptiste Mouret with the ISIR, Universite Pierre et Marie Curie-Paris) start to think about what this robot is doing in terms of what injured humans do:

Overall, learning to predict what behaviors should be avoided is a very general concept that can be applied to many situations in which a robot has to autonomously adapt its behavior. On a higher level, this concept could also share some similarities with what human do when they are injured: if a movement is painful, humans do not fully understand what cause the pain, but they identify the behaviors that cause the pain; once they know that some move are painful, they learn to instinctively avoid them. Humans seem reluctant to permanently change their self-model to reflect what behaviors are possible: people with an immobilized leg still know how to activate their muscles, amputated people frequently report pain in their missing members (Ramachandran and Hirstein, 1998) and dream about themselves in their intact body (Mulder et al., 2008). Humans may therefore learn by combining their self-model with a second model that predicts which behaviors should be avoided, even if they are possible. This model would be similar in essence to a transferability function.

You can read the entire paper (hooray!) at the link below.

[ Fast Damage Recovery in Robotics with the T-Resilience Algorithm ] via [ ISIR ]

The Conversation (0)

How the U.S. Army Is Turning Robots Into Team Players

Engineers battle the limits of deep learning for battlefield bots

11 min read
Robot with threads near a fallen branch

RoMan, the Army Research Laboratory's robotic manipulator, considers the best way to grasp and move a tree branch at the Adelphi Laboratory Center, in Maryland.

Evan Ackerman
LightGreen

This article is part of our special report on AI, “The Great AI Reckoning.

"I should probably not be standing this close," I think to myself, as the robot slowly approaches a large tree branch on the floor in front of me. It's not the size of the branch that makes me nervous—it's that the robot is operating autonomously, and that while I know what it's supposed to do, I'm not entirely sure what it will do. If everything works the way the roboticists at the U.S. Army Research Laboratory (ARL) in Adelphi, Md., expect, the robot will identify the branch, grasp it, and drag it out of the way. These folks know what they're doing, but I've spent enough time around robots that I take a small step backwards anyway.

The robot, named RoMan, for Robotic Manipulator, is about the size of a large lawn mower, with a tracked base that helps it handle most kinds of terrain. At the front, it has a squat torso equipped with cameras and depth sensors, as well as a pair of arms that were harvested from a prototype disaster-response robot originally developed at NASA's Jet Propulsion Laboratory for a DARPA robotics competition. RoMan's job today is roadway clearing, a multistep task that ARL wants the robot to complete as autonomously as possible. Instead of instructing the robot to grasp specific objects in specific ways and move them to specific places, the operators tell RoMan to "go clear a path." It's then up to the robot to make all the decisions necessary to achieve that objective.

Keep Reading ↓ Show less