SNGR: Selective Non-Gaussian Refinement for Ambiguous SLAM Factor Graphs

TL;DR

SNGR introduces a targeted non-Gaussian refinement method for SLAM that improves accuracy and efficiency by focusing on problematic regions identified through covariance analysis.

Contribution

The paper proposes a novel selective refinement approach for SLAM that combines Gaussian and non-Gaussian inference based on covariance diagnostics.

Findings

SNGR detects failure regions effectively using covariance condition numbers.

SNGR improves local likelihood estimates while reducing computational costs.

Experiments demonstrate high-precision failure detection in range-only SLAM.

Abstract

We present Selective Non-Gaussian Refinement (SNGR), a SLAM framework that augments iSAM2 with targeted nested sampling on windows where Gaussian approximations are likely to fail. We detect such regions using the condition number of joint marginal covariances and selectively refine them using the full nonlinear factor graph likelihood, with a gating mechanism to avoid degradation in multimodal cases. Experiments on range-only SLAM with wrong data association show that SNGR achieves high-precision failure detection and consistent local likelihood improvements while reducing computational cost relative to exhaustive non-Gaussian inference. These results highlight both the promise and the limitations of selective refinement for approximate SLAM posteriors.