Combinatorial Filters: Handling Sensing Uncertainty by Avoiding Big Models
Speaker: Steven M. LaValle , University of Illinois at Urbana-Champaign
Date: December 3 2009
Time: 4:00PM to 5:00PM
Location: D463
Host: Russ Tedrake, Robot Locomotion Group/CSAIL
Contact: Kathy Bates, 3-5817, kbates@csail.mit.edu
Relevant URL: http://msl.cs.uiuc.edu/~lavalle/iros09/Abstract:
Over the past several years, Bayesian filtering techniques have become
mainstream tools in robotics research that handle uncertainty.
Variations include the classical Kalman filter and recent particle
filters, all of which are routinely used for robot localization,
navigation, and map building. In this talk, I will introduce a new
class of filters, called combinatorial filters, that are distinctive
in several ways: 1) they simplify modeling burdens by avoiding
probabilities, 2) they are designed for processing information from
the weakest sensor abstractions possible, and 3) they avoid
unnecessary state estimation. In many ways, they are the direct
analog to Bayesian filters, but handle substantial amounts of
uncertainty by refusing to model it. The emphasis is on detecting and
maintaining tiny pieces of information that are critical to solving
robotic tasks, such as navigation, map building, target tracking, and
pursuit-evasion. Once a combinatorial filter has been designed, it
can then be extended to a reduced-complexity Bayesian filter that
exploits task-specific structure while achieve some robustness due to
statistical modeling. Our work is considered as a step in an ongoing
quest to find smaller information spaces on which to design filters
and planning algorithms.
This work is supported in part by MURI/ONR, DARPA SToMP, and the NSF
robotics area.
See other events that are part of
See other events happening in December 2009
