Proceedings of EASCON 1984: 17th Annual Electronics and Aerospace Conference

Washington DC, 11-13 September 1984, pp 215-219

CURRENT TECHNICAL RESEARCH ISSUES OF AUTONOMOUS ROBOTS EMPLOYED IN COMBAT

S. Y. Harmon & D. W. Gage

Code 442, Naval Ocean Systems Center

San Diego, CA 92152

ABSTRACT here are robot reliability, combat robot

security and coordination of multiple
The recent upsurge in interest in robots. These issues are, for the most
autonomous robots for combat applications part, mainstream issues in computer science
has focused considerable attention on and engineering. However, full
several of the obvious technical issues consideration of their impact upon the
(e.g, target recognition, autonomous implementation of combat robots is critical
navigation, route planning). However, to the introduction of these autonomous
several technical issues exist which remain combat robots into the defense inventory.
unapproached and, in some cases, even The final section of this paper discusses
unacknowledged by the robotics community. the risks of and opportunities for
This paper explores three such issues: (1) deployment of autonomous robots in combat
robot fault tolerance, (2) robot security situations.
and (3) multi-robot coordination. These
issues are discussed in terms of the
technology limitations and the research TECHNICAL ISSUES
issues associated with those limitations.
A common message which occurs several times Reliability
during this discussion denotes the
importance in modular implementation and System reliability is no new problem for
well defined interfaces between subsystems the military. Military operations are
in the development of autonomous combat often critically dependent upon timing.
robots. Failure of a single piece of equipment at
the wrong time (i.e., just before or during
a mission) could cause significant losses
INTRODUCTION of human and materiel resources as well as
the potential failure of the mission
Interest in combat applications of itself. Of course, the wrong times are
autonomous and semi-autonomous robots has precisely when maximum system performance
risen recently with the spreading is demanded and expected. The wrong times
realization that recent developments in a are also when enemy actions are most likely
number of hardware and software technology to reduce system capability.
areas (sensors, processors, knowledge based
programming techniques, complex system The word reliability can be used in a
control) will soon make demonstrations of number of senses. It can mean
autonomous combat robots for various susceptibility to failure as a result of
missions feasible. flaws introduced during system manufacture.
These are the failures to which consumer
Autonomous robots will eventually change warranties apply. Reliability can also
the face of combat as much or more than any mean resistance to adverse external
other single technology. However, before influences during operation. These
that can happen the user community must failures are common in the stressful
have confidence that autonomous robots are environment of military operations and are
reliable, secure and cost effective options not protected by manufacturers warranties
which can be successfully integrated into but are common in military operations.
existing and future command, control and Military users often do not discriminate
communications (C3) systems. This paper between these different types of failures
attempts to begin bringing into focus since they often create the same problems.
several technical issues that must be Also, common with complex systems is the
confronted to transform the feasibility reliability problem which arises when the
demonstrations of combat robots of the system designer defines the operational
mid-1980s into the operational systems of problem differently than the system user.
the 1990s and beyond. The issues discussed
 Form Approved
Report Documentation Page OMB No. 0704-0188

Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it
does not display a currently valid OMB control number.

1. REPORT DATE 2. REPORT TYPE 3. DATES COVERED

SEP 1984 N/A -
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Current Technical Research Issues of Autonomous Robots Employed in 5b. GRANT NUMBER
Combat
5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

5e. TASK NUMBER

5f. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION

REPORT NUMBER
Naval Ocean Systems Center Code 442 San Diego, CA 92152
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)

11. SPONSOR/MONITOR’S REPORT

NUMBER(S)

12. DISTRIBUTION/AVAILABILITY STATEMENT

Approved for public release, distribution unlimited
13. SUPPLEMENTARY NOTES

14. ABSTRACT

15. SUBJECT TERMS

16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE
UU 5
unclassified unclassified unclassified

Standard Form 298 (Rev. 8-98)

Prescribed by ANSI Std Z39-18
As a result, the system behaves as designed redundancy a device's likelihood of failure
but differently than the user intends and is proportional to its complexity.
expects. Finally, the treatment of "edge However, a complex device can be made
effects" of autonomous robots as a significantly more reliable than an
situation encountered differs increasingly immensely simpler device through
from the designed task domain becomes redundancy. Redundancy can be used to
critical when developing systems for actual detect faults (e.g., though voting), to
deployment. In military operations quite isolate faults and to recover from faults
bizarre situations can be encountered. For which are discovered or occur during
example, how should an infantry robot be operation. Analytical techniques are
programmed to discriminate between a deaf available to determine the level of
nun and a Soviet soldier pretending to be a redundancy necessary to provide the
deaf nun? Infantry robots may be faced required system reliability for a specified
with 10,000 soldiers dressed like nuns mission. Redundancy is necessary not only
should such a vulnerability be discovered in the computing elements of a robot but
by the enemy. also in the mechanical components (e.g.,
two arms, six independently driven wheels).
A reliable system must perform its assigned In order to take advantage of such
task within the expected time when the task redundancy, the robot must be able to
is within the robot's capabilities. revise its normal strategies when faults
Furthermore, the robot must let the user occur, as it is often possible to use
know when it no longer has the ability to system components for tasks for which they
perform a requested task. Thus, the system were not originally intended (e.g., using
must be able to recognize when it can no arms to drag a damaged vehicle a short
longer perform its task and must be able to distance, using pliers as a hammer). This
communicate that knowledge to the user. issue of employing conventional resources
This type of reliability can be achieved for unconventional purposes is well beyond
through systematic and accurate design, the capabilities of existing techniques in
implementation using reliable components, automated planning.
coordinated redundancy, fault minimization
and self repair. The addition of self repair capability can
further enhance long term reliability and
The good design and implementation reduce maintenance and repair costs. This
procedures necessary to ensure a fault free is an option available only to robots and
nature are, for the most part, available as it simplifies the support of the system.
computer aided design tools which make At the very least, the robot should provide
exhaustive and systematic design of extensive self diagnostic capability to
mechanical, electronic and software assist field repair. Unfortunately, beyond
components a reality. For hardware, automated diagnosis, robot self repair is
extensive methodical design practices and beyond the state of existing technology.
careful choice of components improves the
robot's fault resistance. For software,
modern design practices coupled with Security
program specification and verification
tools provide the only hope of implementing The use of complex information handling
reliable software in complex systems. Good devices always presents a threat to
robot design begins with careful system military security. For this reason
specification. Care is needed in system considerable resources are invested in
specification to insure that the delivered establishing and maintaining the computer
capability corresponds with the desired and communications security of military
capability. These observations about robot systems. Robots present special security
design are true for any complex system. challenges in combat situations. They
However, designers of manned systems have must interact with many elements of the
taken advantage of the inherent flexibility hostile environment in many different ways
of humans. Autonomous systems designers to accomplish their missions. These
will have no such luxury. A successful channels of interaction add to the commonly
operational autonomous combat robot will recognized channels of compromise.
require all the computer aided design Furthermore, autonomous robots present
capability available. significantly more opportunities for
compromise because they must interact with
System fault tolerance can ensure that a the battlefield environment in ways which
complex piece of equipment will function cannot be predicted when the robots are -
reliably throughout a mission of prescribed first programmed. For a robot, security
duration. Without a high degree of means prevention of compromise of the
information stored within the robot and for future autonomous robot designers. How
minimization of enemy ability to alter the much of a robot's processing resources
behavior of the robot. How the should be devoted to analyzing system
interactions between an autonomous robot security? Can the robot identify
and the enemy can be monitored and situations with high security risk and
controlled without sacrificing the robot's avoid them? The alternative to take action
effectiveness has yet to be determined. against security violations is an
Elements of classical communications and alternative available only to a robot and
computer security can be applied to this makes robot security unlike computer or
problem but existing techniques do not communications system security. Just one
provide a complete or even satisfactory step from action to prevent compromise is
solution to robot security. the most intriguing security question
related to autonomous robots. Will a robot
Autonomous robots must be built upon a ever be able to be a double agent (i.e.,
reliable foundation to be secure. make the enemy think that it is
Autonomous robot designers must also take compromising itself when it is actually
advantage of existing secure system design trying to gather information about or to
techniques. Much of the design care that affect enemy capability)?
is required for a reliable system is also
necessary for secure system design.
Security and reliability both emphasize the Multi-Robot Coordination
need for modular implementations. Computer
aided design must also be used to insure Autonomous robots will not be used alone in
that the foundation system behaves as combat. They will always be used to
specified. A secure kernal of proven complement available human controlled
functionality can be used while it is still resources including fully manned systems as
infeasible to mathematically prove the well as remotely manned devices such as
correctness of all software components. remotely piloted vehicles (RPVs).
Secure autonomous robot design requires the Considerable investment has already been
adoption of security models but existing made in existing C3 assets and any new
models inadequately represent the processes system must be integrated into these C3
of robot systems. Such techniques as systems if it is to be accepted and
capability addressing, system partitioning, effective. Ideally, autonomous robots can
encryption, identification friend or foe be configured to respond similarly to
and more are necessary to realize practical manned systems. This strategy reduces the
autonomous combat robots. An awareness of alterations necessary to make them an
system security must be designed into the integral part of a combat C3 system.
autonomous robot as part of its task. This However, this mimicry only solves part of
awareness can shift some of the burden of the problem. Considerable research is
security to the robot itself and, thus, still necessary to determine how manned and
make design simpler. autonomous resources can best be employed
cooperatively. Incremental introduction of
Use of imperfectly secure systems for autonomous robots into the combat
limited applications is possible but environment means more complex
autonomous robots for widespread deployment communications are required to facilitate
in combat must be proven secure against the mix of manned and autonomous force
enemy penetration and corruption. elements evolving over time. The best
Autonomous robots which control firepower strategy is to design communications
or electronic countermeasures are between robots that humans can always
particularly sensitive to the security understand. This strategy also provides an
issue because they could inflict inherent debugging capability. An
significant damage upon friendly forces if autonomous robot should, like an expert
compromised. That is, they could not only system, be able to explain the reasoning
compromise information but they could also behind its actions.
actually adversely affect friendly force
elements which have not been compromised. Autonomous robots offer the opportunity to
streamline the operations of multiple
The very complexity of autonomous robot combat systems by providing well defined
systems makes them formidable security responses to known situations. This
risks. In addition, as experience with streamlining could reduce system response
secure computer systems has demonstrated, time enough to gain significant advantage
security imposes considerable overhead. over an adversary with superior numbers.
Considering this cost, the security issues Multiple robots used cooperatively as
pose some of the most interesting questions distributed sensors gain improved range,
accuracy and resistance to errors as funding is decreased then the rate at which
compared with a single robot. Distributed autonomous robots are introduced into the
robots can also be used in a variety of operational inventory decreases. This link
tactical roles to improve the ability to between user confidence and funding makes
bring coordinated fire upon a single target maintaining high user confidence paramount.
or a series of distributed targets in Premature introduction of autonomous robots
coordinated attack. Distributed autonomous for combat could have a disastrous effect
robots could provide an ability to upon the future development of combat
coordinate military operations with a robots. At best, premature fielding of
precision unknown today if they are secure autonomous robots would sour the user
and reliable as individuals. community on this new technology and result
in inefficient and wasteful application to
Permitting complex cooperation between noncritical missions. At worst, it could
manned and automated systems further lead to significant and unanticipated
exacerbates the security situation by battlefield losses. Poor user perception
providing many more complex interaction of the effectiveness of these systems would
mechanisms which could be penetrated and certainly dramatically affect their future
compromised. Well defined interfaces development and deployment.
between autonomous robots are necessary for
their communication and cooperation. There is no reason to suppose that current
operational doctrines incorporate the
Multi-robot cooperation also raises several optimal modes of deployment for battlefield
questions. For instance, how should robotic systems, which will, for the
function be allocated between the various foreseeable future at least, be more
systems to accomplish a single mission? expendable than manned systems, more
How should the cooperating systems precise in their response to anticipated
communicate and how much should they situations, and more unpredictable in their
communicate? Should redistribution of response to situations unanticipated by
functional roles occur during the actual their developers. Operational commanders
execution of the mission? If so, who will have to develop a sense of the
should be able to coordinate that capabilities of the systems, and,
redistribution of function and how should commanders being human (at least so far)
they decide? How should the command this will most probably be done initially
structure change when the mix of manned in terms of manned system equivalents. If
systems and automated systems changes (what a commander's model of one squadron of 6
if ultimately all manned systems were robotic microtanks is that it is equivalent
eliminated from active combat roles?) to one platoon of infantry, he will use it
in the same way that he would use a platoon
of infantry, which may not be fully
DEPLOYMENT OPPORTUNITIES AND RISKS appropriate. The implication of this is
that it is not enough that a robotic system
Advances in computer security, fault should offer spectacular capabilities that
tolerance and multi-system cooperation will will certainly change the shape of the
have operational relevance only if the battlefield in the future; if a system is
users have an accurate understanding of and to be accepted today it must be capable of
well founded confidence in autonomous playing a contributing role in the
robots' capabilities. Considerable battlefield of today, using doctrines of
experience with these systems will be today.
required before this desirable state is
developed. Implementation experience will Introduction of autonomous robots to combat
come first through feasibility will be hastened as existing manned assets
demonstrations, then through limited are retrofitted. This retrofit can occur
applications and, only later, through incrementally as different autonomous
widespread application. As user confidence subsystem technologies develop. However,
in autonomous robots increases so shall the flexible system architectures which
application opportunities (and the facilitate incremental implementation will
corresponding funding of autonomous robot have to be developed and demonstrated for
development efforts) increase. As user retrofit to be possible. This requirement
confidence decreases so shall the translates into the need for modular
application opportunities (and the subsystems with well defined interfaces
corresponding robot development funding) between subsystems. Furthermore, new
decrease. Implementation experience manned systems should be designed to
depends upon development activity which accommodate this retrofit.
depends upon funding. If development
 CONCLUSIONS

The most important conclusion one can draw

from the present state of autonomous robot
development is that modular subsystem
implementation with well defined interfaces
between subsystems is necessary to robot
reliability, robot security, multi-robot
coordination and, eventual, operational
deployment. Furthermore, modular design
readily facilitates system evolution as
well as simplifying troubleshooting and
repair. Both qualities are necessary for
operational deployment in combat.

As a final word of caution against the

rising enthusiasm about autonomous combat
robots, potential military users should be
careful not to interpret near term
feasibility demonstrations of autonomous
robots for limited combat situations as
near term opportunities for operational
deployment. In the same light, autonomous
robot developers should be careful not to
oversell the capabilities of their systems.
Premature deployment of autonomous robots
will slow the overall development of combat
robots. As discussed in this paper, many
more critical issues need solution and
resolution before operational combat robots
can be deployed.

ACKNOWLEDGEMENTS

The material in this paper evolved from

discussions with Dr. John Clark and Dr.
William Whelan of the Rand Corporation, Dr.
Azad Madni of Perceptronics Inc., David
Smith of the Naval Ocean System Center and
Dr. David Mizell of the Office of Naval
Research. Their contributions are
gratefully appreciated. The most important
contribution has been the funding for this
work and for that we are grateful to Mr.
Gerald Clapp, manager of the USMC 6.2
Surveillance Block Funding.