clock menu more-arrow no yes

Filed under:

The super advanced Atlas robot can’t vacuum better than a Roomba

New, 9 comments

A humanoid robot that takes out the trash and cleans our floors is a far more pleasant view of the future than an AI army of automatons bent on enslaving mankind. Unfortunately, we still haven't figured out how to make robots move around the real world with the precision of a human, let alone perform household chores. Florida's Institute for Human & Machine Cognition (IHMC) is getting closer, however, with its in-house Atlas robot.

The team released a video this week of its Atlas, a bipedal humanoid bot originally developed by Google-owned Boston Mechanics with government funding back in 2013, performing a set of activities that require "whole-body coordinated motions." Those activities include vacuuming a carpet and placing trash in a receptacle. The experiments, conducted as part of a routine code test, are designed to help expand the robot's capabilities, albeit with a human operator helping it along the way.

Watching the Atlas perform these tasks isn't quite as terrifying as the animal-like movements of the robotic quadruped BigDog, but the test is a reminder that we humans are far off from creating the humanoid robots we so often see in Hollywood. After all, bipedal bots today tend to fall over a lot, and they're unable to perform tasks with any degree of complexity without close human supervision and control. Also, this most recent IHMC video was sped up 20 times its normal speed, so this type of precision still requires the robot to function at a painstaking pace.

ATLAS can barely walk without falling over, let alone clean your house

IHMC was one of several teams using Atlas at last year's Defense Advanced Research Projects Agency (DARPA) Robotics Challenge, for which DARPA contracted Boston Mechanics for a price of $10.8 million to provide Atlas units to participants and tech support for the contest. IHMC came second place in the challenge, and now the group is hard at work trying to improve its Atlas' control algorithms while trying to increase the number of activities and movements the robot can perform autonomously.