By Frank Dellaert, Georgia Institute of Technology, USA, dellaert@cc.gatech.edu | Michael Kaess, Carnegie Mellon University, USA, kaess@cmu.edu
We review the use of factor graphs for the modeling and solving of large-scale inference problems in robotics. Factor graphs are a family of probabilistic graphical models, other examples of which are Bayesian networks and Markov random fields, well known from the statistical modeling and machine learning literature. They provide a powerful abstraction that gives insight into particular inference problems, making it easier to think about and design solutions, and write modular software to perform the actual inference. We illustrate their use in the simultaneous localization and mapping problem and other important problems associated with deploying robots in the real world. We introduce factor graphs as an economical representation within which to formulate the different inference problems, setting the stage for the subsequent sections on practical methods to solve them. We explain the nonlinear optimization techniques for solving arbitrary nonlinear factor graphs, which requires repeatedly solving large sparse linear systems.
The sparse structure of the factor graph is the key to understanding this more general algorithm, and hence also understanding (and improving) sparse factorization methods. We provide insight into the graphs underlying robotics inference, and how their sparsity is affected by the implementation choices we make, crucial for achieving highly performant algorithms. As many inference problems in robotics are incremental, we also discuss the iSAM class of algorithms that can reuse previous computations, re-interpreting incremental matrix factorization methods as operations on graphical models, introducing the Bayes tree in the process. Because in most practical situations we will have to deal with 3D rotations and other nonlinear manifolds, we also introduce the more sophisticated machinery to perform optimization on nonlinear manifolds. Finally, we provide an overview of applications of factor graphs for robot perception, showing the broad impact factor graphs had in robot perception.
Factor Graphs for Robot Perception reviews the use of factor graphs for the modeling and solving of large-scale inference problems in robotics. Factor graphs are a family of probabilistic graphical models, other examples of which are Bayesian networks and Markov random fields, well known from the statistical modeling and machine learning literature. They provide a powerful abstraction that gives insight into particular inference problems, making it easier to think about and design solutions, and write modular software to perform the actual inference. This book illustrates their use in the simultaneous localization and mapping problem and other important problems associated with deploying robots in the real world. Factor graphs are introduced as an economical representation within which to formulate the different inference problems, setting the stage for the subsequent sections on practical methods to solve them. The book explains the nonlinear optimization techniques for solving arbitrary nonlinear factor graphs, which requires repeatedly solving large sparse linear systems.
Factor Graphs for Robot Perception will be of interest to students, researchers and practicing roboticists with an interest in the broad impact factor graphs have had, and continue to have, in robot perception.