Premium
Gamma‐SLAM: Visual SLAM in unstructured environments using variance grid maps
Author(s) -
Marks Tim K.,
Howard Andrew,
Bajracharya Max,
Cottrell Garrison W.,
Matthies Larry H.
Publication year - 2009
Publication title -
journal of field robotics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.152
H-Index - 96
eISSN - 1556-4967
pISSN - 1556-4959
DOI - 10.1002/rob.20273
Subject(s) - artificial intelligence , occupancy grid mapping , particle filter , simultaneous localization and mapping , computer vision , grid , computer science , posterior probability , filter (signal processing) , elevation (ballistics) , pixel , visual odometry , algorithm , geography , mathematics , robot , mobile robot , bayesian probability , geodesy , geometry
This paper describes an online stereo visual simultaneous localization and mapping (SLAM) algorithm developed for the Learning Applied to Ground Robotics (LAGR) program. The Gamma‐SLAM algorithm uses a Rao–Blackwellized particle filter to obtain a joint posterior over poses and maps: the pose distribution is estimated using a particle filter, and each particle has its own map that is obtained through exact filtering conditioned on the particle's pose. Visual odometry is used to provide good proposal distributions for the particle filter, and maps are represented using a Cartesian grid. Unlike previous grid‐based SLAM algorithms, however, the Gamma‐SLAM map maintains a posterior distribution over the elevation variance in each cell. This variance grid map can capture rocks, vegetation, and other objects that are typically found in unstructured environments but are not well modeled by traditional occupancy or elevation grid maps. The algorithm runs in real time on conventional processors and has been evaluated for both qualitative and quantitative accuracy in three outdoor environments over trajectories totaling 1,600 m in length. © 2008 Wiley Periodicals, Inc.