Inference on networks of mixtures for robust robot mapping

Abstract

The central challenge in robotic mapping is obtaining reliable data associations (or "loop closures"): state-of-the-art inference algorithms can fail catastrophically if even one erroneous loop closure is incorporated into the map. Consequently, much work has been done to push error rates closer to zero. However, a long-lived or multi-robot system will still encounter errors, leading to system failure.We propose a fundamentally different approach: allow richer error models that allow the probability of a failure to be explicitly modeled. In other words, rather than characterizing loop closures as being "right" or "wrong", we propose characterizing the error of those loop closures in a more expressive manner that can account for their non-Gaussian behavior. Our approach leads to an fully-integrated Bayesian framework for dealing with error-prone data. Unlike earlier multiple-hypothesis approaches, our approach avoids exponential memory complexity and is fast enough for real-time performance.We show that the proposed method not only allows loop closing errors to be automatically identified, but also that in extreme cases, the "front-end" loop-validation systems can be unnecessary. We demonstrate our system both on standard benchmarks and on the real-world datasets that motivated this work.

Multi-modal and inertial sensor solutions for navigation-type factor graphs

“…The robustness of the iSAM algorithm is known to be a problem when "bad", inconsistent, or non-Gaussian measurements are added to the factor graph model [181,219]. This thesis is charged as a remake of the iSAM2 algorithm, by using the Bayes tree approach, but allowing non-Gaussian uncertainty models into the factor graph -called the Multi-modal iSAM algorithm.…”

“…A user specified penalty is used to ensure some attempt is made to keep the factor active during the optimization process. Switch variables are comparable to a null-hypothesis approach [181], and has the disadvantage of ignoring information and relying heavily selecting the correct penalty values.…”

Eliminating conditionally independent sets in factor graphs: A unifying perspective based on smart factors

“…Let us consider a setup in which a mobile robot has to perform SLAM in presence of outliers [13], [14], [15]. In this case, some of the measurements are distributed according to a given measurement model (say, a zero-mean Gaussian noise), while other measurements are outliers.…”

“…4 (Column: "Original"). In our approach, each smart factor independently performs elimination of a landmark via (14); then, we only optimize a smaller graph which involves vehicle poses and smart factors, see Fig. 4 (Column: "Smart factors").…”

“…In this case, the price of robustness is the increase in the computational cost of optimizing over a large number of latent variables (the binary variables). Olson and Agarwal [14] proposes a max-mixture model, to avoid introduction of binary variables. Agarwal et al [15] propose to fix the value of the relaxed binary variables, in order to circumvent the solution of a large optimization problem.…”