Stay on Top of Enterprise Technology Trends
Get updates impacting your industry from our GigaOm Research Community
They may not be the equivalent of the Jetson’s Rosie the Robot yet, but robots are coming along nicely as are the tools needed to make them smarter and more adaptive.
Brian Gerkey of Willow Garage, a Menlo Park, Calif.-based incubator dedicated to robotics, was at MIT today pitching ROS, a set of open-source libraries and tools to make software development for robots easier. Until now, most of those who did this work had to use pricey, proprietary software or program their code from scratch–a time-consuming proposition. ROS gives them a higher-level toolset, he said, and one that profits from hundreds of contributors around the world– a kind of Linux for robotics.
ROS was intially designed for the “PR2 human-scale, two-armed research robot” but it’s also used in more consumer-oriented robots on sale now, outdoor robots and robotic toys, Gerkey noted. His video showed ROS robots playing instruments, making breakfast, flying aerobatic machines and scratching the face of a quadriplegic man.
Gerkey was one of 35 entrepreneurs under 35 strutting their stuff at the Emtech 2011 conference at MIT this week.
Another robot-oriented presenter, Pieter Abbeel, assistant professor of electrical engineering and computer science at the University of California-Berkeley, thinks robots can learn from or mimic human actions given the right artificial intelligence foundation–and some time to learn.
“This is apprenticeship learning. You demonstrate the task, you present the task in way that’s amenable to computer execution and the robot can get to autonomous execution over time,” Abbeel said. The fundamental problem is that even a task that’s simple for humans–folding towels–can be difficult for robots because one item–a towel dumped on the counter–can appear in many different ways to a robot. But that can be ovecome, he said.
Likewise, sock sorting may be mundane to humans, but quite tricky for robots. The robot must be trained using a couple-hundred pair of socks, with a person showing the robot where the heel, toe, opening are, which socks match. “The robot can then internalize those concepts and sort the socks itself,” Abbeel said.
This may not be rocket science, but there are plenty of folks sick of sorting laundry that might relish these robots.