MIT EECS - Angle Undergraduate Research and Innovation Scholar
Object Placement as Inverse Motion Planning
Leslie P. Kaelbling
We present an approach to robust placing that uses movable surfaces in the environment to guide a badly grasped object into a goal pose. We model the physics of placing as a mixture model of simple object motions, and show that this easily computed model allows us to do useful planning for placing. We give an algorithm that searches over the possible configurations of the object and environment and chooses the configuration most likely to lead to a successful place according to the transition model. We show that this algorithm coupled with a very simple model allows the PR2 robot to execute places that fail with traditional placing implementations.
I have worked with Professors Kaelbling and Lozano-Perez at MIT on robust robotic manipulation on the PR2 robot. In my previous project for Professors Kaelbling and Lozano-Perez, I developed an algorithm for the PR2 robot to open doors. I spent a month in Mexico city in January of 2012 developing a physics curriculum for high school students.