Partial Hierarchical Pose Graph Optimization for SLAM

TL;DR

This paper introduces a partial hierarchical pose graph optimization method for SLAM that significantly speeds up computation by a factor of 10 without losing accuracy, using a novel incremental hierarchy-building approach.

Contribution

It presents a fast incremental hierarchy construction for pose graphs and demonstrates the effectiveness of partial optimization in SLAM.

Findings

Partial HPGO achieves 10x speedup over traditional methods.

The proposed hierarchy-building procedure is fast and flexible.

Partial optimization maintains accuracy while reducing computation time.

Abstract

In this paper we consider a hierarchical pose graph optimization (HPGO) for Simultaneous Localization and Mapping (SLAM). We propose a fast incremental procedure for building hierarchy levels in pose graphs. We study the properties of this procedure and show that our solution delivers high execution speed, high reduction rate and good flexibility. We propose a way to do partial hierarchical optimization and compare it to other optimization modes. We show that given a comparatively large amount of poses, partial HPGO gives a 10x speed up comparing to the original optimization, not sacrificing the quality.