|
|
Foundations of Robotics
Seminar, December 11, 2007
Time
and Place | Seminar Abstract
Numerically
Stable Iterated Filters for Bearing-Only SLAM
Smith hall 100
Talk 4:00 pm
This talk discusses the
importance of iteration when performing the measurement update step for the
problem of bearing-only SLAM. We focus on an undelayed
approach that initializes a landmark after only one bearing measurement.
The conventional extended Kalman
filter (EKF) measurement update can often lead to a divergent state estimate
due to its inconsistency in linearization. We show that representing the
bearing-only update as a numerical optimization problem (solved with an
iterative approach such as Gauss-Newton minimization) prevents divergence of
the Kalman filter state and produces accurate SLAM
results for a bearing-only sensor. In addition, we show that even the well
established inverse-depth parametrization for
bearing-only SLAM can fall victim to the failures of the EKF. We propose the
use of an iterated Kalman filter to resolve these
issues as well. Also, for numerical stability, we demonstrate square root
filters for the Kalman update equation. Two outdoor
mobile robot experiments are discussed to compare algorithm performance. This
work is done with Stephen Tully, George Kantor, and Howie Choset.
The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.