A USC Viterbi School of Engineering research team won a top international robotics prize for a paper on the instruction of robots.
Peter Pastor and Mrinal Kalakrishnan, who work in professor Stefan Schaal’s Computational Learning and Motor Control Laboratory, co-authored a paper, along with Schaal and postdoctoral researcher Ludovic Righetti, about online movement adaptation based on previous sensor experience.
The paper was one of nearly 800 accepted – from a pool of more than 2,500 submissions – at the 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems held in San Francisco.
Pastor explained that the grasping actions programmed into their robots are vastly different from the grasps robots now routinely perform in, for example, the assembly of automobiles. In those circumstances, all the movements and the force used for these movements can be precisely programmed in advance.
For robots working in an everyday environment, Righetti said that the parameters are open and unknown.
“The robot looks [at locations] where objects could be,” Pastor said. “It scans the table and finds a cluster of data.” Then it uses 3-D visual processing to try to decode what it sees.
One cluster of data might be decoded as a cup, Pastor said.
“So assuming it is a cup, the robot has to decide how to grasp the cup – from the side, from the top, on the handles. And then [it must decide] how much force [to use]. And then we have motion-planning algorithms. We have to compute a collision-free trajectory so the hand doesn’t smash the table.”
Once robotic hands can negotiate the world, the possibilities are wide-ranging – from caring for the elderly to caring for pets. But first the robot’s hands must be safe and to be safe, according to the researchers, it needs to instantly adapt its behavior to unforeseen events – things that may look like a cup but are something quite different.
The robots create expectations about the objects they are picking up based on past experience, but they need to adjust their behavior if the expectations don’t match what their fingers feel.
In the future, the hope will be to build on the success achieved to expand the robotic flexibility of knowledge and adaptation.Its hand may not be able to gently pick up and stroke a kitten tomorrow or the next day, but give it time.
The National Science Foundation and the Defense Advanced Research Projects Agency supported the research.