Generalized Least Squares for Graph-Embeddable Problems
Speaker: Giorgio Grisetti
Title: Generalized Least Squares for Graph-Embeddable Problems
Date: October 29 2010, 10.00
Location: Aula Magna DIS
Several problems in robotics and computer vision can be reduced to least squares optimization of functions that are represented by a graph. Examples of these problems include simultaneous localization and mapping (SLAM) or bundle adjustment (BA). In this talk we present a general framework for graph optimization that can be easily extended to new classes of problems and supports arbitrary methods to solve the underlying linear system. We compared our method to existing approaches for SLAM and BA. While being general, our system achieves performances that compete with state-of-the-art ad-hoc approaches.