Newsgroups: comp.robotics
Path: brunix!cat.cis.Brown.EDU!agate!ames!kronos.arc.nasa.gov!doctor
From: doctor@kronos.arc.nasa.gov (Terry Fong)
Subject: Dante II - Mt. Spurr Expedition Status (30-Jul-94)
Message-ID: <1994Aug1.070957.15023@ptolemy-ethernet.arc.nasa.gov>
Summary: status information on CMU/NASA volcano robot explorer
Keywords: Dante, teleoperations, volcano, Alaska
Sender: usenet@ptolemy-ethernet.arc.nasa.gov (usenet@ptolemy.arc.nasa.gov)
Nntp-Posting-Host: tardis.arc.nasa.gov
Organization: NASA Ames Intelligent Mechanisms Group
Date: Mon, 1 Aug 1994 07:09:57 GMT
Lines: 128

The following is a status update on the Dante II project. Dante II is
an eight-legged robot (developed by the CMU Robotics Institute with
sponsorship from NASA) which is currently descending into the active
crater of Mt. Spurr, an Alaskan volcano 80 miles west of Anchorage,
Alaska. The primary objective of the Dante II project is to develop and
evaluate new techniques and technologies which can be applied to space
and planetary exploration. 

Throughout the Mt. Spurr expedition, Dante II is being remotely
operated from a base station in Anchorage using control station
software developed at CMU and NASA Ames. Live video coverage is
periodically available via NASA Select. Full mission details including
current status, real-time images/video, and technical data is available
via a WorldWideWeb server provided by the NASA Ames Intelligent
Mechanisms Group. This site can be accessed on the Internet using
information browsers such as NCSA Mosaic and the URL:

	http://maas-neotek.arc.nasa.gov/dante

Questions regarding the Dante II project may be addressed to me via email 
(terry@ptolemy.arc.nasa.gov). Reponse time may be slow.

Terry Fong
NASA Ames Research Center

------------------------------------------------------------------------------

Daily Report (30-Jul-1994)

Dante has achieved fahrvergnugen!

Dante continued to walk into the crater of Mt. Spurr today, and is now
a total distance of 350 feet from the anchor point.  The majority of
the day's activities involved moving across a steep ridge at the top of
the second snow field, descending the snow field, then climbing a
second ridge which curves down into the crater.

In moving across the first ridge, Dante stepped from a nearly level
starting point down a four-foot tall embankment and onto a
thirty-degree slope.  This entire area was exposed volcanic mud, which
offered very soft footing and provided poor support for the legs of the
robot.  On more than one occasion, the snowshoes mounted on Dante's
legs provided support and stability in the soft mud.  Due to the
unusual footing and the severity of the transition this portion of the
traverse was watched very closely by the operators, and forced to
progress at a relatively slow pace.

Once beyond the transition, the next 80 meters of walking was across a
snow- covered slope at an angle of approximately 30 degrees.  After
characterizing the performance of the robot in deep snow during
yesterday's excursion, the operators  (Henning Pangels and Dave
Wettergreen) enabled the behavior-based control mode for Dante to
permit rapid progress  through the snow field.  By utilizing this
control mode the level of operator involvement with the navigation and
mobility of the robot is minimized, and the on-board autonomy
capabilities are exercised.  The operator gives the robot a course
heading and distance, and the robot proceeds to that goal without
operator intervention.  By combining the high-level goal to determine
the overall plan with reactive behaviors to control the motion of the
legs required to get the robot to the goal, a system is utilized which
minimizes any need for operator interaction while the robot is
walking.  The control system combines inputs from the force and torque
sensors mounted on the legs of the robot with the tether exit angles
sensors and the on-board inclinometers to get an understanding of it's
own orientation with respect to the surrounding environment.  It uses
this input to generate an understanding of where it is located with
respect to the tether and the terrain, and creates a plan to move the
legs accordingly.  Should the motion plan be insignificant or
incorrect, the control software will recognize sensor inputs which
exceed the expected normal values and replan the walking move
accordingly.  The control software will also recognize a breakaway
condition (where a support surface suddenly gives way or a rock
supporting a leg shifts under the weight of the robot) and correct it
automatically.

The behavior based walking was completed at the bottom of the second
snow field on the descent path.  At that point the machine was halted
for about 45 minutes while the operators discussed options for the
descend path and whether to go to the left or to the very far left of a
large boulder in the path of Dante.  After deciding to go a little more
than a little left of the boulder, the robot walked to the end of the
snow field and the start of the next ridge.  Dante was then manually
guided over the ridge of exposed volcanic soil and on to a snow-free
slope.  The robot descended another twenty feet and was then "parked"
for the evening next to a large boulder (which will provide some
shielding from rock fall) until the resumption of walking tomorrow
morning.

During the course of the traverse, more than twenty laser scans of the
terrain were taken and quilted together to provide and map of the
interior of the crater.  On several occasions this map was used by the
operators to analyze the true topography of the crater slope and gain a
better understanding of the mobility challenges posed by the crater
interior than could be gathered from the camera viewpoints provided by
the robot.   This was particularly useful on both of the transition
areas.

In addition to characterizing the performance of the robot and several
component technology subsystems in expected and unexpected surface
conditions (snow, mud, hard ash and rugged rocks), a considerable
amount of operator interface and sensor design information has been
gathered.  Much of this information has to do with learning to
understand terrain features and hazards from a single overhead
viewpoint combined with limited onboard camera views, as well as
learning which camera views provide the most useful information under
which circumstances.

For tomorrow, Dante will follow the ridge upon which it is now standing
down into the crater.  This large terrain feature, which starts about
half way down the crater, extends to the end of the large rock bulge
which defines one of the primary areas of interest for the science
portion of the mission.  The bulge is about 100 feet long and appears
to rise up to 40 feet above the bottom of the crater slope.   The plan
is to have Dante follow the ridge line for the next 200 feet to the
bottom of the bulge, then gather visual information on the region
hidden from direct viewing by the bulge.  Along the way, the robot will
provide close-up viewing of some unknown green material which covers
part of the ridge.  After that, an assessment will be made of the
overall robot performance and the quality and quantity of science data
gathered by Dante, and a determination will be made on progressing to
the secondary science interest areas.

------------------------------------------------------------------------------
-- 
_______________________________________________________________________________
 "Every once in a while declare        Terry Fong <terry@ptolemy.arc.nasa.gov>
  peace. It confuses the hell out      NASA Ames, M/S 269-3, Moffett Field, CA
  of your enemies" -- Rule of Acq. #76      (415) 604-6063, (415) 604-6081 lab
