- Leslie Pack Kaelbling and Tomás
Lozano-Pérez, Integrated Task and Motion Planning in
Belief Space, International Journal of Robotics
Detailed Technical Reports (as of July, 2012)
- Leslie Pack Kaelbling and Tomás Lozano-Pérez, Unifying Perception, Estimation and Action for Mobile Manipulation via Belief Space Planning, IEEE Conference on Robotics and Automation (ICRA), May 2012. PDF
- Leslie Pack Kaelbling and Tomás Lozano-Pérez, Pre-image backchaining in belief space for mobile manipulation, International Symposium on Robotics Research (ISRR), August 2011. PDF
- Leslie Pack Kaelbling and Tomás Lozano-Pérez, Hierarchical Planning in the Now (Finalist, Best Manipulation Paper Award), IEEE Conference on Robotics and Automation (ICRA), May 2011. PDF
PR2 Videos (in MOV format)
These videos have been edited to cut out the "thinking" time. The planning time
is about the same duration as the motion time. The underlying motion planner is
a simple RRT with some smoothing, but not much, and so it generates motions that
are often "surprising".
Simulation Videos (in MOV format)
In the belief-space movies, the "shadows" around the objects and the robot
reflect the current uncertainty. In the planning sequences, the cyan robot
reflects paths and configurations generated during planning.
- A long simulated sequence of the PR2 in an "observable" kitchen/apartment. See description in MIT-CSAIL-TR-2012-018.
- Simulated PR2 manipulating two objects on a table; This requires placing
at the destination with high accuracy; the initial place fails and has to be
re-planned and executed. See description in MIT-CSAIL-TR-2012-019.
- Simulated PR2 moving an object to look at a second object. See description in MIT-CSAIL-TR-2012-019.
- Simulated PR2 manipulating across two tables. See description in MIT-CSAIL-TR-2012-019.
Some Planning and Execution Trees
Cupboard experiment - Cycle 1
Cupboard experiment - Cycle 5
We thank all the members of the LIS
for their help in this project, and in particular:
George Konidaris, Lawson Wong, Jennifer Barry and Jon Binney for
comments on drafts;
Jared Glover, Aaron Fryman, Will Grathwol, Sanja Popovic and
Dawit Zewdie for developing the 3D perception subsystem;
Lawson Wong for the UKF implementation;
Ashwin Deshpande and Dylan Hadfield-Menell for help with
Jennifer Barry for help in testing the planner and with PR2
Caelan Reed Garrett for help with Cython.
This work was supported in part by the NSF under Grant No. 1117325.
Any opinions, findings, and conclusions or recommendations expressed
in this material are those of the author(s) and do not necessarily
reflect the views of the National Science Foundation. We also
gratefully acknowledge support from ONR MURI grant N00014-09-1-1051,
from AFOSR grant AOARD-104135 and from the Singapore Ministry of
Education under a grant to the Singapore-MIT International Design
Center. We thank Willow Garage for the use of the PR2 robot as part
of the PR2 Beta Program.