Adaptive Graduated Non-Convexity for Pose Graph Optimization

Abstract-- We present a novel approach to robust pose graph optimization based on Graduated Non-Convexity (GNC). Unlike traditional GNC-based methods, the proposed approach employs an adaptive shape function using B-spline to optimize the shape of the robust kernel. This aims to reduce GNC iterations, boosting computational speed without compromising accuracy. SLAM primarily consists of noise and challenging outlier loops. The top images two modules: the front-end, responsible for processing sensor illustrate the ground truth trajectory, marked by the gray line, data, and the back-end, tasked with estimating the sensor with the results of the riSAM algorithm on the left and the trajectory by solving a non-linear least squares problem.