Abstract
In this letter, we present an algorithm for iterative nonlinear least-squares which increases the adaptive nature of previous methods in the literature. Our method uses two parameters to learn the best fitting distribution of the measurement residuals and performs Iterative Re-weighted Least Squares (IRLS) based on these two parameters. This adaptive nature of the weights is shown to be helpful in situations where the noise level varies in the measurements and is shown to increase robustness to outliers. We test our algorithm first on the point cloud registration problem with synthetic data sets, where the true transformation is known. Next, we also evaluate the approach with an open-source LiDAR-inertial SLAM package to demonstrate that the proposed approach is more effective than constant parameters for the application of incremental LiDAR-inertial odometry. This increased adaptivity can help in a wide range of estimation problems in robotics by better modeling the measurement errors.
Abstract (translated)
URL
https://arxiv.org/abs/2206.10305