1998 Winter Expedition
The 1998 expedition will demonstrate the robot Nomad's search and
classification of rocks and meteorites and its autonomous navigation of
polar terrain. The primary demonstration will evaluate Nomad's classifier
in distinguishing and characterizing rocks and meteorites using high-resolution
imagery and optical reflection spectroscopy. The secondary demonstration
will conduct field evaluation of newly developed autonomous capabilities,
including navigation with stereo vision and laser sensors, patterned coverage
planning, and "blind autonomy." Additionally, the field team will evaluate
a landmark-based navigation system, a millimeter wave radar, Nomad's locomotion
on ice and snow, and a communications architecture.
A field team from Carnegie Mellon University,
NASA Ames Research Center, and the University of Pittsburgh will reestablish
a base camp at approximately 80 deg 18' South - 81 deg 16' West, and will
operate there for six weeks. Field operations will take place between November
1 and December 15, 1998. The Chilean Air Force and the Chilean Antarctic
Institute will provide transportation and logistical support, and will
participate with scientists and field experts. A research scholar from
France's LAAS-CNRS (with funding from the French
Polar Institute, IFRTP) will evaluate stereo vision and laser perception
sensors in the context of a robotic program for Antarctic exploration.
Please click here for more information.
Robotic Rock/Meteorite Classification
The purpose of the December 1998 Antarctic field season is to test the
autonomous meteorite classifier under true field conditions. Nomad will
acquire data from new rocks and classify them with limited human input.
This will establish conclusively whether the current system (sensors and
classifier) is practical for field robotic deployment and will determine
its suitability for robotic meteorite search.
Nomad will explore the areas of Independence Moraine, the region where
data and rock samples were acquired last season. Nomad will collect images
and spectral data from new samples, and feed the information to its classifier
for assessment. Physical samples will be taken from each rock that Nomad
examines for analysis by a field geologist on the team. Nomad?s classifier
output will be compared to the geologist?s analysis for validation of Nomad?s
Nomad will be in its meteorite search mode for this expedition, so it
will either move autonomously or be directed to a waypoint by an operator.
Through an interface, the operator will monitor the panoramic images and,
if desired, direct the robot to an area of interest via teleoperation and/or
commands to the maneuver planner. Once closer to the area of interest,
the operator can then select a particular rock and enter it into the database.
The operator can generate a high-resolution image of the object and by
radio communicate with a sensor operator who will place the spectrometer
on the rock of interest. The operator will issue a command to the appropriate
sensor to take a sample. Following sample acquisition, Nomad can resume
autonomous navigation or receive further directions from the operator.
Real-time data processing and classification analysis will be performed
on board Nomad.
A Bayes network-based rock/meteorite classifier will perform assessments
from the visual imagery and reflectance spectra of rocks obtained in the
field. The classifier is capable of identifying and discriminating between
the various local rocks (mostly marble, limestone, granite, quartzite and
some igneous rocks) as well as meteorites and subclasses thereof (achondrite,
chondrite and iron).
The 1997 expedition demonstrated that, despite significant variations due
to cloud cover and other atmospheric conditions, the best results are attained
using spectral data from the visible to the infrared portions of the light
spectrum (350 - 2500nm). Even if confined to only the visible wavelengths
(400 - 900nm), approximately 90% of rock samples and 90% of meteorites
are correctly classified as either rocks or meteorites. When identifying
rock types, approximately 5% of rocks are misclassified using spectra data,
with limestone and marble being the most likely to be confused. This is
not surprising since both consist of the same minerals.
Because the Patriot and Independence Hills regions of Antarctica are
believed to lack meteorites, an area will be seeded with meteorites in
order to test the classifier utility. A geologist viewing high-resolution
images returned by Nomad will designate possible candidates for Nomad?s
classification. In addition, an expedition further inland from Patriot
Hills (to the polar plateau) will allow human investigators to search for
meteorites with a handheld version of Nomad?s spectrometer as well as a
digital camera. This expedition will be an opportunity to acquire
more data from different geographical areas, particular those in which
meteorites may be found.
This season, a human operator will control the prototype meteorite search
system by way of an interface created to allow control and testing of nearly
every component in the science system. Although it can track the
status of most science system components, this interface will deal most
directly with three systems:
Mission Planner. Through the interface, the user will control
the mission planner, specifying a robot task mode, i.e., go to a waypoint,
execute a coverage pattern, or take sensor measurements. The user also
will control the details of how the mission planner will coordinate the
science system, such as setting waypoint tolerance, robot field of view,
and coverage pattern type.
Database. The user interface will allow access to the database.
While the interface does not require updates to the database, the user
will be able to look up information in the database such as target images
or classifier probabilities.
Pan/tilt Camera. The ability to control the pan/tilt camera
to select new rock targets will be one of the most critical functions of
the interface. The user will be able to pan, tilt, and focus the camera
until an acceptable initial, or ?template,? image is taken. From
this image, the user will click on the target and the pan/tilt software
will insert this data into the database. Then the user will select
a high-resolution image, zooming into the target to get a close-up image.
This image, along with other information, will then be inserted into the
database for use by the classifier.
Autonomous Navigation of Polar Terrain
The autonomous navigation system will be vigorously tested and operated
under various polar conditions, with testing of the system?s perceptual
sensors as the primary focus. Additionally, Nomad?s ?blind autonomy,? which
is its ability to detect and circumvent a close object, and new advances
in Nomad?s patterned coverage search will be assessed.
Nomad?s performance will be evaluated on the four major terrain types
present at Patriot Hills: blue ice field, snow, moraine, and sastruggi.
The testing of the stereo vision and laser sensors on these terrain types
will be particularly important, as Nomad has never traversed such terrain.
The effects of and need for filters, particularly polarizing, on the stereo
cameras will be a major part of testing.
For the Antarctic expedition, Nomad?s navigation system is configured with
two types of perception sensors: stereo vision and laser sensors. Each
sensor independently acquires the appropriate data and creates a map. The
maps are grids in which each cell contains a goodness value and a certainty.
A goodness value is a rating of how beneficial it is for Nomad to be in
that particular position; a certainty is the sensor?s confidence level
in the data. These maps are then passed to Morphin, a program which merges
them and determines which direction will provide the maximal benefit.
The goodness value can be adapted to evaluate any number of measurements,
including science value and solar availability. This attribute, plus the
system?s ability to accommodate additional sensors, make this methodology
suitable for use in another context, potentially space exploration.
This new feature of the navigation system assists Nomad by evaluating the
robot?s ability to detect when it is too close to an object. While
Nomad monitors motor currents and angle sensors to determine if it is traversing
dangerous ground, blind autonomy allows both monitors to direct a backup
and turn maneuver in order to recover and turn away from the object.
This ability will be very important in Antarctica where the functionality
of stereo and laser on ice fields is not known.
Patterned Coverage Search
In addition to blind autonomy, autonomous navigation involving patterned
coverage will be evaluated this year. In patterned coverage search, the
operator will define a polygonal area to cover and the pattern to use.
Nomad will then follow that pattern in order to maximize coverage.
While Nomad travels, the operator will be able to select an object of
interest from Nomad?s panoramic view and enter it into the database. The
operator also will be able to direct Nomad in taking various sensor readings;
Nomad will execute these commands even if they require deviating from the
initial planner because the maneuver or path planner can create a route
to the region of interest. When the measurement is complete, Nomad will
resume the initial pattern, returning to the point from which it left and
continuing its autonomous traverse.
Initial activities will be held at Patriot
Hills, but in the course of the expedition, Nomad will traverse much
of the area between the Patriot Hills and the Independence Hills. The following
maps show the planned traversals and important distances.
||Equipment departs from Pittsburgh
||Expedition team departs from Pittsburgh
||Ice deployment of team and equipment
||FACH C-130 from Chile to Patriot Hills, weather permitting.
||Operations at Patriot Hills
||Initial tests and activities close to the Chilean camp.
Side flight to southern locations.
||Operations between Patriot and Independence Hills
||Field demonstrations and technology experiments.
|Nov 30 - Dec 6
||Final leg of expedition. Nomad returns to Camp via East
||Packing and camp break, return to Chile
||End of research activities. Public display of Nomad in
Punta Arenas, Chile.
||Equipment departs, team breaks
Data and voice transmission will be conducted using the INMARSAT satellite
constellation. Powered by the Sun, an INMARSAT terminal will sit at a hilltop
location on the Patriot Hills to transmit data and voice to the US. The
hilltop box will also house a router, ARLAN, power inverter, power regulator,
and a battery. The ARLAN allows wireless ethernet communications from the
operations tent to the hilltop router. To turn on the communications link,
the team will telnet to the router from a computer on the local network
and issue a simple command that initiates a phone call from the INMARSAT
terminal. About a minute later, all the computers in Antarctica can talk
to all the computers on the SCS network at CMU. This satellite link will
be used periodically every hour or as needed when there are new hi-res
images of rocks, spectrometer data, classification results, and telemetry
to send to the web server at CMU.
The map below describes the communications scenario during the 98 Antarctic
One of the unique technologies being field tested on Nomad during this
expedition is the science autonomy system (SAS). The aim of the SAS is
to be able to spot interesting rock/meteorite targets in the robot's vicinity,
move near them, deploy sensors, and classify the mineral composition of
the target to determine whether it is a meteorite. Ultimately, the SAS
should be capable of doing all of these things autonomously. This season
a prototype version is being tested to field test the individual components
of the SAS under human guidance. Therefore, a science
interface was created that allowed the users to coordinate and evaluate
the entire system. An SAS white paper is available here
that describes each of its components.
Additionally, the expedition is using an operator interface designed
at NASA Ames Research Center to drive
Nomad and monitor weather and robot telemetry information. The original
version of this operator interface is described here.
The operator interface used this season is designed to run more reliably,
and on a low-resolution laptop in intense sunlight.
Throughout 1998, the improved perception, planning, and autonomous navigation
capabilities of Nomad were tested at a barren slag site in Pittsburgh.
This allowed the team to evaluate aspects of Nomad's modified locomotion,
autonomous safeguarding, patterned coverage planning, and servo-pointing
of the high-resolution science camera. To assess the readiness and functionality
of Nomad's major subsystems for polar weather conditions, the project team
tested Nomad in a sub-zero storage facility, where the robot was subjected
to temperatures as low as -25oC.
After sitting in the chamber overnight, the robot successfully started
up. Its heaters successfully increased operating temperatures to levels
acceptable for the electronics systems. On-site test of the science sensor
heating systems were also performed.
Additionally, the expedition team was exposed to the very low temperatures
they would experience as well as the logistical problems posed by operating
the science autonomy system interfaces while wearing polar gear.
These are images of Nomad taken during its cold chamber testing:
Here are new images sent back from Antarctica on November 15th. The
first are of Nomad in transit to Antarctica in Chile:
without the outer shell, showing its internal electronics box
and the Chilean hanger
These are images taken near the CMU main camp:
These next images show Nomad's traversals on Antarctic terrain.
These images display the various sensors and systems that were field
Nomad's sensor mast
LAAS laser from Riegl
Nomad's science camera
The Patriot Hills viewed by Nomad's science camera
Another image from the science camera
Ice mobility tests
Robotic Search for Antarctic Meteorites 1998
All material on this page is property of NASA and Carnegie Mellon University.
Any image or text
taken from this site and incorporated into another document without
consent violates the Copyright
Law of the United States and the Berne International Copyright Agreement.
Send comments, questions, or suggestions to Dimitrios
This document prepared by Michael