Jennifer L. Barry
MIT CSAIL Graduate Student
Email: JLBarry AT mit.edu
I graduated from Swarthmore
College with a physics major in 2007 and started as a
graduate student at MIT CSAIL in the same year. I work in the
Learning and Intelligent
Systems Group with Leslie Pack Kaelbling
Lozano-Pérez. When not doing research, I contra
dance, read about computers and robots much smarter than mine and play Ultimate frisbee.
My thesis work was on manipulation planning for multiple types of
manipulation primitives. With state-of-the-art planning
algorithms, robots can work in domains where only pick and
place are required. Given a flat table with reasonably
spaced objects, a robot can be trusted to pick an object up
and put it down in another location. However, this
situation is very rare; if we want robots that can function
in the real world they must be able to deal with clutter and
with objects that cannot be grasped. In my thesis work, I am
developing random sampling algorithms for planning with
diverse types of manipulation. Videos of the Willow Garage PR2 robot
carrying out plans generated by my planner are here, the code is available here, and a more
in-depth summary of my thesis work is here.
My previous work was on hierarchical Markov decision
processes (MDPs). In an MDP, the world consists of a set of
states that an agent can transition between using a set of
probabilistic actions. The agent always knows what state it is in,
but not necessarily to what state an action will take it. Although
this model fully specifies the world at the bottom level, sculpting it
into a useful hierarchy is a complex problem.
- J. Barry. Manipulation with Diverse Actions. Ph.D. thesis. Massachusetts Institute of Technology, 2013. [pdf, high-resolution, uncompressed pictures pdf (warning: this is about 250MB)]
- J. Barry, L. P. Kaelbling, T. Lozano-Pérez. A Hierarchical Approach to Diverse Action Manipulation. In International Conference on Robotics and Automation, May 2013. [pdf] (Note: This paper had a small implementation bug in DARRTH. For correct results, see chapter 5 of my thesis.)
- J. Barry, K. Hsiao, L. P. Kaelbling, T. Lozano-Pérez. Manipulation with Multiple Action Types. In International Symposium on Experimental Robotics, June 2012. [pdf]
- J.L. Barry, L.P. Kaelbling, T. Lozano-Pérez. DetH*:
Approximate Hierarchical Solution of Large Markov
Decision Processes. In International Joint
Conference on Artificial Intelligence, July
- M. Bollini, J. Barry, D. Rus. Bakebot: Baking Cookies with the
PR2. In IROS PR2 Workshop, September 2011.
- J. Barry, M. Bollini, A. Holladay, L. Kaelbling, T. Lozano-Pérez. Planning and Control Under Uncertainty for the PR2. In IROS PR2 Workshop, September 2011. Extended abstract. [pdf]
- J.L. Barry, L.P. Kaelbling, T. Lozano-Pérez. Hierarchical Solution of Large Markov Decision Processes. ICAPS 2010 Workshop on Planning and Scheduling Under Uncertainty, May 2010. [pdf]
- J.L. Barry. Fast Approximate Hierarchical Solution of MDPs. S.M. Thesis. MIT, June, 2009. [pdf]
- J.L. Barry, L.P. Kaelbling, and T. Lozano-Pérez. Fast Approximate Hierarchical Solutions of MDPs. Presented at EECS Masterworks. MIT, 2009. [poster]
- J. L. Barry. Monte Carlo Methods for Sampling Classical and Quantum Particles. Senior Thesis. Swarthmore College, June, 2007. [pdf]
- J.L. Barry and A. L. Bug, Path Integral Sampling Techniques. APS
March Meeting. Denver, CO. Abstract R1.00191, 2007.
[poster] (early version)
- J.L. Barry, M.D. Behn, and L.G.J. Montési, Melting and Mantle Flow
at Oblique Ultraslow Ridges. Eos. Trans. Amer. Geophys. Union, v86.
Abstract T41E-1351, 2005. [poster]
- L.G.J. Montési, M.D. Behn, and J.L. Barry. On the Geodynamics of
Oblique Spreading. Eos, Trans. Amer. Geophys. Union, v86. Abstract
- A. Holloway, J. Barry, H. Jones, and T. Newhall. Reliability for
Nswap. Journal of Computing Sciences in Colleges: Proceedings of
the Tenth Annual CCSC Northeastern Conference, v20(5), p. 123, 2005.
Other Cool Stuff: