It can now be trained like a baby, which opens up its applications to those of us with no robot knowhow.
In the future, it could be trained to pick apples or replace lightbulbs.
Jochen Steil and Matthias Rolf used a process called "goal babbling". This mimics the way a baby learns to grab things by continually reaching. It involves plenty of trial and error, but allows them to figure out which muscles they need to move.
In the robot's case, it remembers what happened to the trunk's position when tiny changes were made to the pressure in the pneumatic tubes that feed the artificial muscles. This makes a map that shows the trunk's position relative to the pressures in each tube.
Hence the trunk can be manually put in a series of positions, and can adopt these on command.
This means average Joes like us could soon have our own home robot trunks to train to carry out such arduous tasks like changing the channel on the TV, or fetching another beer from the fridge. Essential, we're sure you'll agree.
No comments:
Post a Comment