The ARMAR robots have been developing robust grasping strategies for awhile, thanks to software developed by at the University of Karlsruhe. Now researchers Daniel Kappler, Lillian Chang, Nancy Pollard, Tamim Asfour, and Rudiger Dillmann are developing pre-grasping strategies for objects that are tough to pick up straight away, like a cd lying flat on a table.
It’s easier for a humanoid robot to slide a flat object (even one with the height of a pizza box) so that it partially hangs over a table’s edge before picking it up, because their fingers can’t get a grip otherwise. And when an object is too big to lift with one hand, ARMAR-III is learning to slide it towards itself to make that easier, too. Check out the video to see how these strategies work on ARMAR-III as well as a Motoman robot with a Shadow Hand attachment.
These kinds of pre-grasping strategies are likely going to be just as important to a robot’s ability to grasp a diverse range of objects as the grasping strategy itself.