Simultaneous Localization and Layout Model Selection in Manhattan Worlds

TL;DR

This paper introduces a convex optimization approach for SLAM in Manhattan worlds that automatically handles data association and loop closure, producing the simplest consistent environment model.

Contribution

It reformulates SLAM as a model selection problem using convex optimization over layout structures, enabling automatic data association and loop closure.

Findings

Successfully applied to real-world datasets

Automatically performs data association and loop closure

Produces minimal consistent environment models

Abstract

In this paper, we will demonstrate how Manhattan structure can be exploited to transform the Simultaneous Localization and Mapping (SLAM) problem, which is typically solved by a nonlinear optimization over feature positions, into a model selection problem solved by a convex optimization over higher order layout structures, namely walls, floors, and ceilings. Furthermore, we show how our novel formulation leads to an optimization procedure that automatically performs data association and loop closure and which ultimately produces the simplest model of the environment that is consistent with the available measurements. We verify our method on real world data sets collected with various sensing modalities.