Giorgio Grisetti, Cyrill Stachniss, Slawomir Grzonka, Wolfram Burgard
A Tree Parameterization for Efficiently Computing Maximum Likelihood Maps
using Gradient Descent
Abstract:
Building maps is one of the fundamental tasks of mobile robots since
maps are needed for several high-level applications. In 2006, Olson
et al. presented a novel approach to solve the graph-based SLAM
problem by applying stochastic gradient descent to minimize the error
introduced by constraints. Together with multi-level relaxation, this
is one of the most robust and efficient techniques published so far.
In this paper, we present an extension of Olson's algorithm. It
applies a novel parameterization of the nodes in the graph that
significantly improves the performance and enables us to cope with
arbitrary network topologies. The latter allows us to bound the
complexity of the algorithm to the size of the mapped area and not to
the length of the trajectory as it is the case with both previous
approaches. We implemented our technique and compared it to
multi-level relaxation and Olson's algorithm. As we demonstrate in
simulated and in real world experiments, our approach converges faster
than the other approaches and yields accurate maps of the environment.
Bibtex:
@InProceedings{grisetti07rss,
author = {Grisetti, G. and Stachniss, C. and Grzonka, S. and Burgard},
title = {A Tree Parameterization for Efficiently Computing Maximum Likelihood Maps using Gradient Descent},
booktitle = {Proc.~of Robotics: Science and Systems (RSS)},
year = {2007},
address = {Atlanta, GA, USA},
}
Document (pdf file):
full paper