Robots should autonomously use—and, when necessary, seek out—feedback from contact sensors to cope with uncertainty. We approach this problem by formulating contact manipulation as a stochastic process and use belief space planning to generate robust closed-loop manipulation policies.
Contact sensors are difficult to integrate into traditional Bayesian state estimation techniques because they are fundamentally discontinuous. We introduce novel state estimation techniques, based on the particle filter, that exploit the physics of interaction to address this challenge.
HERB is a bi-manual mobile manipulator designed and built by the Personal Robotics Lab at Carnegie Mellon University to complete household manipulation tasks. As a member of the Personal Robotics Lab, I developed many of the planning, perception, and control algorithms used by HERB to autonomously manipulate his environment.
The DARPA ARM-S program aimed to develop software to autonomously complete complex manipulation tasks. As a member of the CMU/NREC team for Phase II of the project, I developed manipulation and perception techniques necessary to complete tasks such as changing a tire or cutting a wire.