Visual Odometry for the Lunar Analogue Rover Artemis - Robotics Institute Carnegie Mellon University

Visual Odometry for the Lunar Analogue Rover Artemis

M. Wagner, D. Wettergreen, and P. Iles
Conference Paper, Proceedings of International Symposium on Artificial Intelligence, Robotics and Automation in Space (iSAIRAS '12), September, 2012

Abstract

In this paper we discuss the visual odometer development effort carried out by ProtoInnovations in support of the Artemis and Artemis Jr. rover development projects conducted by Neptec Design Group for the Canadian Space Agency. The Artemis Visual Odometer consists of a pair of downward-looking cameras that view terrain underneath the vehicle. Its close view of the terrain provides excellent visibility of down- and cross-track vehicle motion, thus allowing accurate measurement of vehicle motion even on rough terrain that induces vibration and wheel slippage. On Artemis, artificial illumination from a bank of LEDs allows it to take measurements at night as well as it can in daylight, thus fulfilling a key requirement.

Using this approach, the visual odometer must track features that pass quickly through the cameras’ fields of view. Motion prediction is used to improve the efficiency of feature tracking when traveling at high speeds. Shadow-edge masking is employed to improve the robustness of tracking the motion of features on the terrain. Sparse stereo-vision matching allows accurate ranging to features, which is important to accurately estimate motion when traversing rough terrain.

Here we present results collected from a testing program that began with evaluation on a bench-top apparatus capable of very accurate motion, and was completed by extensive testing on a number of vehicles, in a range of terrains, and at speeds up to 2 m/s. These results highlight the efficiency and accuracy of the approach.

BibTeX

@conference{Wagner-2012-120404,
author = {M. Wagner and D. Wettergreen and P. Iles},
title = {Visual Odometry for the Lunar Analogue Rover Artemis},
booktitle = {Proceedings of International Symposium on Artificial Intelligence, Robotics and Automation in Space (iSAIRAS '12)},
year = {2012},
month = {September},
}