THE NASA AUTOMATION AND ROBOTICS TECHNOLOGY PROGRAM

By
Lee B. Holcomb
Director9 Information Sciences and Human Factors Division
Melvin D. Montemerlo
Manager, Automation and Robotics

NASA HEADQUARTERS
WASHINGTON~ DC 20546

ABSTRACT

In 1977~ the NASA Study Group on Machine

Intelligence and Robotics was commissioned to
study
the
influence
of
artificial
intelligence9
computer
science,
and
autonomous systems on the full range of NASA
activities I.
Chaired by Carl Sagan~ the
group
included
leading
researchers
from
government, industry and academe.
The group
found that NASA had fallen behind the leading
edge in computer science and that machine
intelligence
and
robotics
technology
was
essential to render future space missions
economical and feasible.
As a result of this
study NASA took aggressive steps to initiate
a
computer
science
research
program ,
initiate
research
and
application
of
artificial
intelligence
to
aerospace
prohlems~
initiate a space human factors
program
with
a
focus
on
teleoperation,
establish
the Center for Aeronautics and
Space
Information
Sciences
(CASlS)
at
Stanford
University 3,
and
initiate
the
Research
Institute
for Advanced
Computer
Sciences (RIACS) 4.

This paper presents an overview of NASA's

Automation
and
Robotics
(A&R)
technology
development program,
covering its history,
objectives~ organization and content.
This
program is being carried out by the Office of
Aeronautics and Space Technology (OAST) which
has the responsibility to provide long range,
high risk aerospace technology.

INTRODUCTION
NASA has been one of the pioneers in the
field of automation and robotic research.
NASA's initial motivation was to enable the
exploration of space in environments hostile
to man.
NASA has conducted and funded
research
in
teleoperators 9
slave
manipulators,
vision
systems,
planning
systems
and
extraterrestrial
roving
vehicles.
Major
operational
robotic
achievements by NASA include the Surveyor
lander on the moon 9 and the Viking/Mars
lander.
NASA's
current
capability
for
robotic operations in space is realized by
the Canadian-built Remote Manipulator System
(RMS).
While
the RMS is an important
component of our space capability~
it is
limited to the proximity of the Shuttle and
to simple tasks such as the "fly swatter"
attempt to rescue the Hughes Leasat in April
1985.

In addition, during the summer of 1980, NASA

held a ten week workshop
to study the
feasibility of using robotics and automation
in future space missions-.
This
study
helped
focus
NASA's
modest
robotics research program on the development
of teleoperator technology to augment the
Space Shuttle's space operations capability.
The Remote Orbital Servicing System (ROSS)
concept served as the application focus.

The space program is now entered a new era

with routine access to space provided by the
Space Shuttle.
We have embarked on the Space
Station
program
to
provide
a
permanent
American
presence
in space.
While our
initial automation and robotics efforts were
focused on the narrow objective of replacing
human functions, our current program focuses
on the broader objectives of exploiting the
synergism of human and machine capabilities.

More
recently
automation
and
robotics
research has received renewed interest by
NASA and by Congress.
In 1982 the House
Cormuittee
on
Science
and
Technology,
Subcommittee on Investigation and Oversight
held hearings ~n the government role in
robotic research v. In July, 1984, Conference
report 98-867, was passed mandating NASA to:

Its objective is to exploit the potential of

artificial intelligence and robotics to:
I)
increase the probability of mission success;
2) decrease the cost of ground control; and
3) increase the capability and flexibility of
space operations.
Artificial intelligence
(AI) technology will be used to reduce the
size of the ground control contingent, and
telerobotics will be used to enable increased
space assembly, servicing, and repair.

"The conferees have included a total of

$205,000,000 for space station activities
including $5,000,000 which is available
only
for
automation
and
robotic
activities.
Specially, the Committee of
Conference directs that the $5,000,000 be
used for delivery of a flight telerobotic
system at the time of initial Space
Station
operational
capability
for a
mobile
remote manipulator
for station

I]

12

assembly and maintenance and a smart

front end in the orbital maneuvering
vehicle
for
remote
operations
and
servicing."

Human-machine interface must

the system from the start.

be

built

into

H U M A N INTERFACE

These reports were submitted to Congress on

April i, 19857 .
They identified the most
promising technology areas for advancement
and served to reinforce NASA's plans
to
exploit automation and robotics in the Space
Station and other Agency programs.

-- ~ FORMATS
HEAD.UP DISPLAY
ALERTING SYSTEM
- 1 ] ALEATS

KEY TECHNICAL CHALLENGES

~

The effective exploitation of automation and

robotics for aerospace applications requires
that we make significant technical advances
in three areas:

THROTTLEILEFT HANDJ
SW,TCHES
-,~,--~.

STICK GRIP
IRIGHT HANDI
- ~ MULTIFUm:nON
swsTc~s

Real Time Capability

FIGURE 2

Advanced automation and robotics application

require real-time computational capability.
As shown in Figure 19 traditional expert
system applications have been off-line~ nonreal time advisors.
As we move toward realtime automated control of aerospace systems,
delays in response could lead to overall
system failure.
The solution of this problem
will come through attention to control system
architecture
(e.g.
distributed
control)
through advances in matching software to
computer
architecture~
and
through
the
development
of
high-speed
embedded
processors.

Validation
Finally,
techniques
to validate automated
systems must be developed which will deal
with
system complexities
associated
with
fault tollerant and artificial intelligent
systems.
Figure 3 depicts the result of
inadequate
system
validation
techniques.
Conventional
aerospace
software validation
techniques,
which
strive
to
test
every
possible system state, are both infeasible
and unaffordable.

REAL T I M E

VALIDATION

....

PASSIVE

ACTIVE

CONSULTANTS

CONTROLLER

OFF-UNE, SLOW-TIME

ON-LINE. REAL-TIME

FIGURE 1

Man-Machine Interface
A second technical challenge is the ability
to design an effective man-machine interface
which will allow the system to transition
between operator controlled and automated
operation.
Figure 2 shows a view of the F-18
cockpit and lists the variety of displays
that it contains and the complexity of the
information
input.
One
can
see
that
overwhelming
the
operator
with
control
options is not an acceptable solution.
The
user
must
be
provided
with
sufficient
information
to effectively
supervise
the
system
in
the
event
of
unforeseen
circumstances.
User acceptance is essential
if the automated system is to be effective.

"IT FIGURES. IF THERE'S ARTIFICIAL

INTELLIGENCE, THERE'S B O U N D TO BE
S O M E ARTIFICIAL STUPIDI'I'Y. '

FIGURE 3

PROGRAM DEVELOPMENT
Figure 4 depicts the planning assumptions
established to guide program development.
An
in-house technical committee was utilized to
develop an integrated technology development
plan which has been reviewed by several
external advisory groups.

13

Assm(PTIONS
o

L~*O
There

(i.e.,

will

be

two

foci~

talerobotlcs

and

system

A/~TCCnOLO~'SUMNAmV
L~,EI*=I

autonomy

,x,to[T
o

experta y s t m e s ) .

Each f o c u s w i l l
have = series of ground demonstrations
o f an e v o l u t i o n a r y
t e s t b e d t o show Incresslngc ~ p a b i l i t y
of integrated
technoloQles.

There will
Capabilities
sequences.

he s c o r e t e c h n o l o g y
program to develop
needed to enable the demonstration

L*.CU,~C

O,~.,TO.
J"T'"*CK
T,CDOLUCZlS

- VUSUAL. VOXCU. TOre"

PAI~LS

the

CONNUmlC,TIO"

TH"ST|XAL

TKCUmOLO~V
IO~OTS

O L'C.rW.~.~
.OTOUl
TI*"S"~S.IO"I

The r e s o u r c e b a l a n c e w i l l
1/3 for the demonstration

be 2 / 3 f o r
sequences.

Any flight
demonstrations
w i t h t h e NASA u s e r o f f i c e s

w i l l be f u n d e d
(Space S t a t i o n

Each o f t h e t w o g r o u n d
take place at a single,
technologies

demonstration
but possibly
will

core

technology

and

...................

be d e v e l o p e d

I'TC~RA*CB Ct. COIT ~IC""OLOCT

~o?ii???:!F;:::::~

in conjunction
and S h u t t l e ) .

sequences
different,
at

will each
Center.

The c o r e

Government, industry
~nd u n i v e r s i t y
research
leveraged
b y u s i n g 40 p e r c e n t o f t h e p r o g r a m
to sponsor out-of-house
research.

For each focus, the technology development will begin

with an early demonstration
of current capabilities
and
move ~orward i n a g g r e s s i v e b u t r e a s o n a b l e i n c r e m e n t s ,
leading to the development of a revolutionary
technology
for automation
and r o b o t i c s .

.o.oT r,oca,R.l.C

various

sites.

w i l l be
funding

FIGURE 6

will
be e s t a b l i s h e d
to l i n k r e s e a r c h
Centers, user Centers,NASA programoffices,and
apropria~eu n i v e r s i t i e s and industries.
Teaming arrangements

CORE TECHNOLOGY

FIGURE 4

The architecture for an automated system is

shown
in
Figure
5.
This
architecture
involves five general areas:
(a) operator
interface; (b) task planning and reasoning;
(c)
control
execution;
(d)
sensing
and
perception; and (e) systems architecture and
integration.
In all system implementations
an
operator
Interacts
with
the operator

u
Architecture

for

an

Automated

System

--1

,L_.
FIGURE 5

interface and effects state changes in the

world
through
control
execution.
Each
general
area
contains
a
number
of
subelements.
The task planning and reasoning
area plays a central role and is the most
complex.
It will
ultimately
contain
a
planner,
executor,
monitor,
simulator,
diagnoser,
and knowledge base.
NASA is
advancing core technologies in each of the
five technical areas given above.
The core
technology program has been developed with
careful attention to leveraging or exploiting
significant technology efforts of industry
and academia.
Figure 6 provides a listing of
technical areas in which NASA will lead,
leverage
or
exploit
National
research
efforts.

The core technology program is responsible

for developing
the component
technologies
which
are
then
transferred
to
the
demonstration programs.
Current projects in
each of the five areas are described.
Sensing and Perception
The sensing and perception area will develop
hardware/software systems for all types of
sensing
including:
vision,
tactile,
force/torque, and proximity.
At the present
time there are two projects, both in vision
sensing.
The
objective
of research
in
machine vision is to develop the capability
to recognize, acquire and track objects and
to verify actions in space operations.
A. Programmable Image Feature Extractor~
PIFEX
(JPL).
In
1986
an
end-to-end
demonstration of the acquisition and tracking
of
simple
unlabelled
objects
will
be
performed.
Also underway is the testing and
validation of a scheme
for locating and
tracking simple labelled objects. Development
of
a
real-time
vision
processor
will
continue,
culminating
in
1987
in
an
implementation of a 120 module advanced image
processing system, PIFEX, capable of some ten
billion
operations
per
second
on
image
data.
PIFEX
will
allow
real-time
identification of complex object features,
stereo correlation, and other computationintensive vision functions.
B. Focal Plane Array Processor (LaRC).
The
second
element
of
the
sensing
and
perception area is the development of a focal
plane array processor which will reduce the
numerical
computation
load
on a machine
vision system by performing some of the data
reduction optically at the sensor.
Task Planning and Reasoning
This area covers the artificial intelligence
R&D which form the basis for intelligent
monitoring,
planning,
operating
and
diagnosing of systems both in ground control
and in telerobots.
It has three elements.

14

A. D e ision
Makin$ (JPL).
This task
includes
interactive
goal-driven
planning,
spatial planning for multi-arm telerobots and
planning with uncertainty.
The approach on
interactive, goal driven planning will be to
integrate DEVISER and PLAN-IT.
The FAITH
diagnoser program will begin extension to
allow multiple, temporal and spatial failure
reasoning
with
intelligent
search-space
reduction and the ability to reason about
permissible
execution
deviations
due
to
uncertainty.
Plan-driven
execution
monitoring will continue with the development
of plan simulation capability.
The
approach
in
knowledge-based
system
development tools is to complete a prototype
of the Multiple Reasoning Engine (MRE).
The
Multiple Reasoning Engine is composed of a
Blackboard, Conditions Model, Memory Model,
Process
Model,
Reasoning
Engine
Design
Language (REDL), Graphics Debugging Tool, and
a Time Representation Model.
The approach on
integration
of
knowledge-based
subsystems
will be to use the blackboard interface of
the MRE for integrating subsystems into a
cooperating
structure.
The
blackboard
architecture allows data, task requests, and
knowledge
to be shared among the various
knowledge-based subsystems.
B. Computer
Assisted
Desisn
(CAD)
Planner (GSFC).
CAD based telerobot planning
consists
of using the detailed
computerreadable geometric descriptions of spacecraft
and payloads that result from the computer
aided design (CAD) process, and transforming
them
into
a knowledge
base
useable
to
automatically plan the robot motions needed
to accomplish servicing tasks.
Two basic
types of robot plan can be built by AI
programs
operating
against
this geometric
knowledge base:
the macro-plan that defines
the sequence of operations and the gross
motions needed to get the tool and use it to
remove a bolt.
An important consideration in
either type of plan is that the execution of
the
plan
has
to
involve
real
world
uncertainties
and
the
consequent
modifications of the plan to accommodate them
when necessary.
C. Knowledse Based Systems (ARC).
The
primary
focus
of
the
research
is
the
development
of AI technologies
leading to
advanced
machine
intelligent
systems
for
imagery
and
pattern
recognition
applications.
The
critical
research
component
centers
around
knowledge
engineering and includes technology elements
such
as:
knowledge
extraction
and
understanding
from multiple
data sources;
representation of that knowledge; maintenance
of data base consistency; automated software
development, verification and
validation to
minimize
the need
for
skilled
knowledge
engineers; and machine learning algorithms.
Research
products
include
expert
systems
development tools for planning, scheduling,
fault diagnostics,
monitoring and control,
world
simulation,
systems
analysis/
interpretation/configuration,
and
training;
executive
controllers;
and
machine
intelligent systems.

Control Execution
The objective of this research is to develop,
evaluate, and apply telerobotics guidance and
control technology for space applications,
and to advance the state of the art in
manipulator
control.
The approach is to
investigate cooperative human/machine tasks
and to augment teleoperator functions through
the
application
of
advanced
computer-and
sensor-based control technology, to automate
the system and to elevate the operator to a
higher level of supervisory control.
There
are three elements.
A. Telerobot
Guidance
and
Control
(LaRC).
The 1986 basic research in adaptive
control of manipulators will be investigated
both in the ROBSIM robotics simulation and on
actual
manipulator
hardware
in
the
Intelligent Systems Research Lab (ISRL).
The
primary emphasis is the implementation and
evaluation of adaptive control algorithms, to
handle varying loads and inertias, and to
address
the interaction
of a manipulator
mounted on a moving base.
Algorithms for the
coordinated control of multiple manipulators
performing
a
cooperative
task
will
be
developed and evaluated in early 1986.
A
joint program with NBS and the Army will
result
in
a
prototype
laser
scanning/
designator system which will be evaluated in
the ISRL.
A high accuracy proximity sensing
design based on the laser system is being
developed.
These basic research results will
be implemented in late 1986 to accomplish a
realistic space servicing task.. Fairchild
has developed a satellite refueling connector
which will be tested on a future shuttle
flight.
The refueling task will be automated
in
the
ISRL
so
that
the
task can be
accomplished
faster,
and
the
human
can
function
as
a
supervisor,
with
manual
(teleoperator) control available as a back up
or contingency option.
B. Teleoperator
Control
(JPL).
The
objective of this work is twofold.
(i)
Development
and evaluation of modular and
expandable distributed microcomputer hardware
and
software
system matching
the natural
needs
of
real-time
mechanization
of
manipulator
control in space applications.
(2)
Development
and
evaluation
of
new
prototype
smart
end
effectors
with
microcomputers
integrated
into
the
end
effectors
for
sensor
and
control
data
handling
and
interfaced
to
the
overall
distributed
real-time
manipulator
computer
control
system.
The
natural
needs
of
advanced
manipulator
control
in
space
include:
(a)
distribution
of
real-time
control computing between control station and
remote manipulators equipped with smart end
effectors
and tools and (b) the use of
alternative, interchangeable and interactive
control
techniques
like
(i)
generalized
force-reflecting
hand
controller
equipped
with
force-reflecting
hand
trigger,
(ii)
sensor-referenced automatic control and (iii)
supervisory control, including interface to
task planning expert systems.
The notion of
expandable control mechanization includes the
capability
of
extending
the
distributed
microcomputer
system
to
the
coordinated
control of multiple-arm systems.

15

C. Limber
Manipulator
Control
(ARCStanford). The long term objective of this
research
is
to
develop
methods
for
controlling
satellite
based
manipulators
during the real time performance of orbital
assembly and handling tasks.
The research
focus is on fast, precise control of the
endpoints
of
manipulators
using
direct
spatial measurements of endpoint position and
target position, and development of control
strategies for teleoperation at a supervisory
level,
i . e . giving
the astronaut
cogent
dynamic
insight
and
task
management
authority.
Included
with
this research
effort is the demonstration of air cushion
vehicled
equipped
with
flexible
(limber)
manipulator
systems.
These vehicles are
being used to obtain precise data on the
dynamics and control of service spacecraft
intended to interact with target spacecraft
via
flexible
manipulators.
Problems
involving the real-time control and execution
of autonomous systems are part of the overall
research
effort.
Research
in the Task
Planning and Reasoning element of the core
technology program is being integrated into
this effort.
Operator Interface
The goal of the operator interface research
is to develop the capability to evolve human
control
of
remote
manipulation
from
teleoperation
(manual
control
of
remote
manipulator) to supervisory control (giving
task-level connnands and letting the computer
generate
the implementation plan).
This
includes being able to monitor the telerobot,
to aid it in doing what it is not yet capable
of doing automatically, and to take over when
the automation fails or degrades into a
telepresence mode (i.e. teleoperation with
rich sensory feedback).
There are three
elements:
A. Operator
Station
Human
Factors
(JPL).
New
operator
control/information
interface
concepts
will be designed
and
tested in a stand alone mode and in an
integrated
control
station
environment,
focusing the development and data-gathering/
modelling efforts on human factors issues
related to operator interface with dual arm
telerobots.
Experimental investigation will
be carried out on:
(i) the effects of
alternative display techniques of visual and
non-visual sensor information on operator's
perceptlve/cognitive
performance,
(2)
operator's manual control performance using
generalized Task-level and force-reflecting
control techniques, including the effect of
microgravity on operator performance, and (3)
language-like
interface
methods
to
supervisory control of telerobots.
Function
allocation
between
operator
and
sensor/
computer/based-automation
will
be
investigated for various task and operational
constraints,
including
time delays,
using
appropriate task boards.
A feasibility study
is also carried out for automating stereo
vision systems.
B. Visual/Tactile
Feedback
(JPL-NOSC).
The objective of this research effort is to
develop and evaluate tactile display system
suitable for integrating tactile information

into the direct human and supervised control

of remote manipulators.
The effort covers
both unimodal
and cross-modal information
feedback possibilities and methods.
Also
included
in
this
research
work
is
the
development of techniques (i) for combined
tactile and visual information displays, (ii)
for mixing tactile (cutaneous) sensor data
with
robot
hand
internal
state
(kinesthetic/proprioceptive) sensor data to
create
haptic
information,
that
is,
to
recognize shapes of objects by grasp, and
(iii) for mixing tactile data with robot hand
motion
to
create
information on shapes,
c o n t o u r s , etc.
C. Supervisory Control (JPL-MIT).
The
objective of this research is to develop a
quantitative understanding of human factors
parameters involved in supervisory control of
remote space manipulators.
The supervisory
control concept covers a broad spectrum of
human
involvement
in the remote control:
bilateral task-level manual control exercised
through an adjustable computer loop, shared
and traded manual and automatic computer
controls referenced to task models and to
sensor information,
use of task planning
expert systems
for on-line reconfiguration
and
monitoring
of
automatic
computer
controls,
language-like
interface
to
intelligent computer control of manipulators,
etc.
Included in this research effort are
issues
of
human
operator
control
and
information interface to the operation of
dual-arm robots equipped with dexterous and
effectors.
System Architecture and Intesration
This area focuses on computing and telerobot
architectures
for
real-time
execution
of
autonomous
functions.
System
control
architectures enable smooth integration of
these functions.
There are three elements:
A. Spaceborne
Symbolic
Processor
(ARC).
The focus of this research is a space
borne VHSIC symbolic processor capable of
handling a minimum of 22,000 rules with an
execution rate of 8,000 rules per second
(equivalent
to
8
mega-instructions
per
second).
Functional characteristics of the
processor include 40-bit data-tagger parallel
architecture
capable of acconrmodating new
evolving architectural designs without impact
on
the
existing
hardware
or
software
environment; vendor-independent data and bus
interfaces capable of accommodating evolving
peripheral
subsystems
such as an optical
read-write
disk;
operation
in a vendorindependent
distributed
computer
environment;
control
strategies
for
maintaining
data-base
consistency;
faulttolerant
capabilities
for
fault
testing,
identification, isolation, and resolution via
software
control
and
implementation;
and
radiation resistance (minimum of 10n to the
5th rads).
The research effort also includes
the programming environment for a parallel
architecture and will support common LISP,
Prolog,
Ada,
and C.
Several potential
architectures are currently being evaluated
and
include
state
of
the art machines
developed by Symbolics, Inc. (NASA sponsored
effort in conjunction with IR&D), TI (DARPA

16

sponsored effort)~ DEC (IR&D) and

(DARPA and NASA sponsored effort).

Stanford

B. Satellite
Desisn
for
Servicin$
(GSFC).
This work addresses the problem of
how
satellites
and
payloads
need
to be
designed to facilitate their servicing by
robots.
This work has the added benefit that
such design guidelines tend to also make this
equipment more easily serviceable by humans
on the ground or in space.
Considerations
cover such areas as the design of fasteners;
electrical/gas/fluid
connectors;
the
size,
function and number of replaceable modules;
visible
markings
for
automatic
identification; and the design of tapered
guides,
etc.
to decrease
robot
accuracy
requirements.
The 1986 work will involve
collecting existing information on the design
of
satellites
for
servicing.
Existing
information exists in the Goddard MMS, GRO,
ST operations and Space Station programs.
This information will be used to develop
point designs of space payloads that reflect
robot friendly design characteristics.
C. Beam Assembly Telerohot
(HD~-MIT).
The HIT Space Systems Laboratory has a grant
to develop technologies for increasing the
capability of telerobots to perform on-orbit
operations such as assembly, and to evaluate
those capabilities using a neutrally buoyant
telerobot
called
the
Beam
Assembly
Teleoperator
(BAT)
in
the
MSFC
Neutral
Buoyancy Facility as a simulation of the
space
environment.
Using
capabilities
previously developed under this grant, they
also evaluate the allocation of tasks to EVA
astronauts and to telerobots.
Using BAT
together
with
the
Multi-Mode
Proximity
Operations Device (MPOD); which is operated
with a human on-board rather than remotely,
as
the BAT,
the relative advantages and
limitations
of
near
versus
remote
telerobotics will be evaluated.
Also covered
under this grant is the interaction of EVA
and telerobotics with the design of satellite
mechanisms such as latches~ connections and
interchangeable modules.
Students in this
program spend periods of time at JPL both as
internships
and as a technology transfer
mechanism.
To
focus
our
core
research
efforts
a
comprehensive set of ground based and flight
demonstrations has also been planned.

DEMONSTRATION PROGRAMS
Each
focus of the A&R program,
aircraft
telerobotics
and systems autonomy,
has a
planned
demonstration
sequence.
More
demonstrations will integrate technologies,
and permits evaluation of the overall state
of the art in each field.
It provides a
magnet for relevant component technologies
being
developed
outside
the
program.
Finally~ and perhaps most importantly,
it
permits
potential
users
to
provide
the
program with feedback as to its potential
usefulness.

Telerobotics
The thrust of the telerobotics focus is to
evolve the technology of remote manipulation
from its current state of teleoperation (i.e.
direct manual control of a remote manipulator
by humans) to teleroboties (i.e. supervisory
or
task-level
control
of
a
remote
manipulator)
OAST
is
planning
to evolve
teleoperation to telerobotics.
The
Space
Station
Office
was
tasked by
Congress with developing a remote manipulator
flight-article to be ready for use at Initial
Operating Capability (IOC).
The OAST telerobotics ground demonstration
sequence is shown in Figure 7.
It begins
with a two-armed telerobot that can perform
servicing (e.g. module exchange) on robotfriendly satellites, and has some limited
autonomous capabilities.
This will be the

NASA SPACE TELEROBOT -LAIIOIIATOIIY OEMONS'IIIAIION

SEQUENCE
o~

~0p~mc

sir

e~LESF~X~LlU~E~tA~
Ol,N~l

~ j

IP&ClSll

n ~ T o ~ TELl RHOTS P l ~ t OAU C ~ L [

i ~ t l C

TI ~ L ~ I

~ T ~

~lPa~ Al~Lmml

I A n ~ q N I O ~ I mY

a 7I~IAY

A~ PInMA~

RIPNN~

I x TLUlLVl C,~LDDATA ILL46~

FIGURE 7

first telerobot ever built with its own

intelligence
for
planning
and
control
execution.
More importantly, it will have an
architecture
to
permit
its
level
of
intelligence to be increased in subsequent
demonstrations.
As
the
demonstration
sequence
progresses,
the
telerobot
will
become more autonomous and more flexible in
handling different types of tasks, and it
will become more robust in terms of detecting
and working around anomalies autonomously.
In the second demonstration, the testbed will
be able to grapple and de-spin a tumbling
satellite.
In the third, it will be capable
of local mobility.
Then comes the capability
for fabrication, and finally at about the
turn of the century, the goal is to have
autonomous cooperating robots.
Figure 8 shows a concept drawing of the FY
1987 demonstration.
The target FY 1987
demonstration is a satellite servicing task
in which highly structured coordination-level
activities (e.g. grasp, move, open, etc.)
will be enabled by an autonomous run-time
control implemented through the teleoperative
interaction by a human supervisor.
Prototype
tasks include Orbital Replaceable Unit (ORU)
replacement
and
fluid
transfer.
Major
milestones in FY 1986 include:
completion of

17

detailed testbed functional and mechanization

deslgns~
taskboard
design
and
facility
preparation~ and fabrication of assemblies of
the sensing and operator control subsystems.

1917 DEMONSTRATION

- tOW-LEVEL AUTONOMT ANO

TELIOIqtAIION IN SAIGUIIE
~tvJoNO

fill

- $IA~NMV

t~V

Ills
lq~I
II~m

II
Mm

mmen

im*~ m lo

I~aanan

Development and integration of generic

software tools for the management and
operation of complex dynamic systems.

2)

Development,
test
and
validation
of
system and subsystem planning and control
technologies for automation of ground and
on-board operations.

Ames Research Center is responsible for the

systems
autonomy
ground
demonstration
sequence.
The site of the demonstrations
will
be
the
Johnson
Space
Center.
Demonstrations will be held in FY 1988, 1990,
1993, and 1996, as shown in Figure 9.

.~~I

..

I)

m ~ i

w e ~ ~ I c m a u 5 m e n u
s ~ m s ~ eo i m
ill-.m I

SYSTEMSAUTONOMY
DEMONSTRATIONPROGRAt'I
igel

~ n c. aI nr o l ,
Cl u ~ m 5 I i I

l-ul

iI~m

~ l I

~tqINN
Clat rll M
F~IUtI! ~ I l I i
~mtelll~lnI A~WIlll KI1

k~lIlled
(Wrll
Of
mlIll~ ~ll'Illl~I
liIl~llII

~I m * . m .

trois

imRm~

mill

I I ~

FIGURE 8
mmlcl!

c~t~iI

1996
Oletrlkte4 c 4 o l ~ t O f
mltlple S * b s y s t e m s
rlRilllllmlt AII~ IMe')

it

Multiple M q I t O e S
rlMllllpt A.IIINI')

The telerobot will have a generic control

architecture which incorporates process-level
planning~ trajectory planning~ and execution
monitoring into an embedded run-time control
system capable of dual-arm ORU exchange.
This
task will
be
typical
of EVA ORU
replacement tasks such as those that occurred
on the Solar Max repair mission.
Coordlnation-level
autonomy
will
be
demonstrated
in FY 1987.
Process-level
autonomy will be demonstrated in FY 1990.
Process-level refers to a structured sequence
of coordlnation-level tasks. An example of a
process-level task would be "remove module
from
bay,"
which
is
comprised
of
coordinatlon-level elements as: acquire task
frame, detach panel, remove r stow~ etc.
The Jet Propulsion Laboratory is responsible
for the telerobotics ground demonstration
program, and will be the site of the testbed.
System Autonomy
The thrust of the system autonomy focus is to
develop and integrate artificial intelligence
technology
for
the
intelligent
automated
control of complex dynamic systems.
This
means more than the ability to control the
system
automatically
during
nominal
operation.
It means
that the automated
control must be robust to both small and
large anomalies~ some of which have standard
malfunction procedures and some of which are
unanticipated
or
at
least
have
no
precedent.
At
the
present
time,
all
unanticipated
problems,
as well
as many
anticipated problems~ and often even system
monitoring during nominal system operation,
are handled by humans.
The goal of the
systems autonomy focus is to reduce the size
of
the
ground
control
contingent
by
developing and applying the techniques of
artificial intelligence.
The systems autonomy
has two goals:

demonstration

program

AdIelI
v m i r iir~iI i t i m

la'~

I~"e~

& ~mI~ I r ~ w l I

pl~L~q * m r ~ v m . ' . l v
A i m , q ~.w.amrpc~; I

~ u
~ ,we p ~ , . W ~
urvn.~
m-lrg
~
i~ I r . ~

FIGURE 9

As can be seen in Figure 9, the capabilities

of artificial intelligence technologies for
control of complex dynamic subsystems will
evolve from control of single subsystems in
1988, to control of multiple subsystems in
1990~ to hierarchical control of multiple
subsystems
in
1993,
and
to
distributed
control of multiple subsystems in 1996.
As
the capability of artificial
intelligence
increases, the role of the remaining human
supervisory
controller
changes.
The AI
demonstration
capability will evolve from
what might be termed an intelligent "aide" in
1988, to an "apprentice" in 1990, to an
"associate" in 1996.
The implication is that as AI capability
evolves, and as confidence in it increases,
fewer people will be needed as controllers.
However some controllers will remain.
They
will be fewer in number, have qualitatively
different roles and responsibilities (i.e.
higher levels of supervisory control), but
they will retain the top-level authority and
responsibility.
They will insure that the
changing
needs
and
wants
of
the
user
communities are taken care of as well as
possible, using available automation as a
tool
to
see
that
changing
tasks
are
accommodated as effectively, efficiently and
safely as possible.
The initial demonstration (1988) will be of
the
"Integrated
Communications
Officer"
(INCO) subsystem for the Shuttle.
The INCO
is a "front room" control position in Mission
Control which manages shuttle communications

18

and instrumentation systems.

There are three
support
personnel
in
the
"back
room"
assisting the INCO.
Candidates for the 1990 demonstration are
being evaluated.
They include:
the electrical/environmental/consumables
mechanical
engineer (EECOM), the propulsion subsystem,
the
data-processing
subsystem,
and
the
payload subsystem.

CONCLUSIONS
The NASA automation and robotics program has
undergone dramatic change over the past eight
years.
We have a strong cow~nitment to
automation and robotics in future Agency
programs
such as space
station and the
orbital maneuvering vehicle.
In part this is
due to the congressional mandate to increase
our reliance on automation and robotics, but
more significantly this increased commitment
is
a
result
of
our
pilot
efforts
in
automation and robotics research which show
promise of significant savings in operations
costs and the potential for enhancing our
space operations infrastructure.
NASA also
believes the spin off of our investment in
automation and robotics will have a profound
impact on the productivity of U.S. industry.

REFERENCES
i.

2.

Sagan, C., et el. "Machine Intelligence

and Robotics: Report of the NASA Study
Group", (Tech. Rep. N80-30086), Pasadena,
California,
Jet Propulsion Laboratory,
September 1979
"NASA Computer Science Research Program
Plan", Technical Memorandum 85631, NASA,
Washington, DC, March 1983

3.

"The
Second
Annual
CASIS
Workshop
Meeting",
Department
of
Electrical
Engineering
ISEC, Stanford University,
Stanford, CA May 31, 1984

4.

Denning,
P.
"Research
Institute
for
Advanced
Computer
Science,
Annual
Report",
NASA
Ames
Research
Center,
Moffett Field, CA, January 31, 1985

5.

Long, J.E., and Healy, T.J. "Advanced

Automation for Space Missions", A Report
of the 1980 NASA/ASEE Summer Study on the
University of Santa Clara, Santa Clara,
CA, September 15, 1980

6.

Rolcomb, L.B. "Statement on Robotics",

Hearings
before
the
Subcommittee
on
Investigations
and
Oversight
of
the
Committee on Science and Technology, U.S.
House of Representatives~ pp. 228-336,
U.S.
Government
Printing
Office,
Washington, DC, June 1982

7.

"Automation and Robotics for The National

Space Program"
Automation
& Robotics
Panel,
California
Space
Institute,
University of California,
February 25,
1985

8.

"Advancing
Automation
and
Robotics
Technology for The Space Station and for
The U.S. Economy"~ National Aeronautics
and
Space
Administration 9
TM-87566,
Advanced Technology Advisory Committee,
Submitted to the United States Congress,
March 1985