Abstract-State POMDPs for Grasping

Abstract: We provide a method for planning under uncertainty for robotic manipulation by partitioning the configuration space into a set of regions that are closed under compliant motions. These regions can be treated as abstract states in a partially observable Markov decision process (POMDP), which can be solved to yield optimal control policies under uncertainty. We demonstrate the approach on simple grasping problems both in simulation and on an actual Barrett Arm, showing that it can construct highly robust, efficiently executable solutions.

"Grasping POMDPs: Theory and Experiments," Ross Glashan and Kaijen Hsiao and Leslie Pack Kaelbling and Tomas Lozano-Perez, RSS Manipulation Workshop: Sensing and Adapting to the Real World, 2007.
"Grasping POMDPs," Kaijen Hsiao and Leslie Pack Kaelbling and Tomas Lozano-Perez. ICRA, 2007.

More details are provided in Chapter 2 of my thesis.

Videos (simulated robot):
Two fingered grasping of boxes
Two-fingered grasping of stepped blocks

Videos (real Barrett Arm):
Placing one finger relative to a stepped block
Two-fingered grasping of a box