Low Bandwidth Geometric Point Cloud Approximations for Subterranean Environments

Loading...
Thumbnail Image

Authors

Bugatto, Anthony
Mascarich, Frank
Khattak, Shehryar M. K.
Feil-Seifer, David J.

Issue Date

2019

Type

Poster

Language

en_US

Keywords

computer vision , communication , mapping , robotics

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

Aerial robots are becoming increasingly important for exploration and mapping of unknown, subterranean environments for scenarios where human labor is unsafe or costly such as mining and underground warfare. In order to explore and map subterranean environments LiDAR is needed, however, LiDAR maps can be extremely data intensive and are often unsuitable for wireless communication. In subterranean environments, robots are forced to use low bandwidth wireless communication channels. Because of the high bandwidth LiDAR data we currently have no way of communicating our maps over a low bandwidth wireless channel. In this paper, we present a solution for low bandwidth mapping using a geometric approximation of the LiDAR point cloud. We use RANSAC to fit cylinders to point cloud segments, adapting in real time for intersections and changes in tunnel curvature. The RANSAC models are able to condense the point cloud into a set of tuples containing shape, centroid coordinates, quaternions, and radius; thus enabling real time, low bandwidth map communication. The map can then be used to accelerate path planning in a constrained coordinate space using the cylindrical approximation and a dynamic, low bandwidth, connectivity graph. The algorithm is validated on a real time, subterranean robot in a commercial mine.

Description

Citation

Publisher

License

In Copyright (All Rights Reserved)

Journal

Volume

Issue

PubMed ID

DOI

ISSN

EISSN

Collections