AI for C2 Group at RMC

1
Artificial Intelligence (AI) in Command and
Control (C2)Part II new challenges

• Dr. Jean Fugère
• 2Lt Francois Robichaud
• Dr. Yawei Liang
• AI for C2 Group
• Dept. of Mathematics and Computer Science
• RMC

2
Contents

• AI in C2, where are we now?
• Potential Challenges
• An example Command Agent
• Conclusion

3
AI in C2, where are we now in MS perspective ?
a real world entity
modelling
model
simulation
4
Why does military want MS?

• affordability
• capability of simulating something rarely
  occurred
• reusability
• interoperability
• educational value and
• academic value.

5
About 500 B.C
Probably, the first recorded practitioner of
military operations research was Sun Tzu (also
known as Sun Zi in China), a fifth-century B.
C. (approximately) general, who wrote The Art
of War .
analysis
6
About 500 B.C
1916
In the 20th century, one of the first formal,
testable bodies of theory concerning military
engagements is that of F. W. Lanchester, who in
1916 published his insightful differential
equations for comparing the losses of opposing
forces in combat under simple assumptions.
F. W. Lanchester, Aircraft in Warfare,
the Dawn of the Fourth Arm, Constable and
Company Ltd., 1916
7
About 500 B.C
1916
1937-39
The formal initiation of military operational
research was given in the Second World War and
arose from a study of the effectiveness of
military operations, primarily in the Royal Air
Force, England.
D. J. White, Operational
Research, John Wiley Sons Ltd., UK, 1985
8
About 500 B.C
1916
1937-39
1940
The first operational modern computer was the
Heath Robinson, built in 1940 by Alan Turings
team for the single purpose of deciphering
German messages. S.
Russell and P. Norvig, Artificial Intelligence,
A Modern Approach, Prentice Hall, 1995, p.14
9
About 500 B.C
1916
1937-39
1940
1942
In October 1942, General Henry Arnold urged all
commands to establish operations research groups,
and, by the end of the war, 26 such groups were
in existence. The importance of operations
research is emphasised at the highest echelons of
the U.S. Army, Navy and Air Force.

G. Isac, Mathematical Methods in
Defence Analysis, RMC, 1998
10
About 500 B.C
1916
1979
1937-39
1940
1942
In my opinion, American Operations Research is
dead even though it has yet to be buried. I also
think there is little chance for its
resurrection because there is so little
understanding of the reasons for its demise By
the mid 1960s, most OR courses in American
universities were given by academics who had
never practised it. They and their students were
text-book products engaging in impure research
couched in the language, but not the reality, of
the real world As a result OR came to be
identified with the use of mathematical models
and algorithms, rather than the ability
to formulate management problems, solve them, and
implement and maintain their solutions in
turbulent environments.
Russuell L.
Ackoff, The Future of Operational Research is
Past,
J. Opl
Res. Soc., Vol.30, No. 2, 1979, pp. 93-104
11
About 500 B.C
1916
1979
1937-39
1988-89
1940
1942
There is now no aspect of military activity where
the simulator or trainer is not present in one
form or another, yet the growth of this
technological development has passed almost
without literary notice. It now seems strange
that it has taken so long for Janes Yearbook to
be compiled and provide military
training simulation with the attention they
deserve. A pilot can spend time in a flight
simulator and then fly a real aircraft, but a
staff officer cannot expect a real shooting war
to be laid on for him to complete his command
training experience. He has therefore to rely on
his own service experience and his command
training system to obtain some idea of what the
real thing would involve.
Terry J. Gander, Foreword, Janes Military
Training Systems, First Edition, 1988-1989
12
About 500 B.C
1916
1979
1937-39
1940
1988-89
1942
The term NDM (Naturalistic Decision Making) first
appeared in 1989, when a conference was
organised so that researchers who had stepped
outside of the traditional decision research
paradigms could discuss their findings, and then
publish them in a single volume. In the 1980s,
these researchers had begun to study how
experienced people actually make decisions in
their natural environments or in simulations
that preserve key aspects of their environments.
In fact, the foregoing can serve as a short
definition of NDM NDM is the way people use
their experience to make decisions in field
settings.
Caroline E. Zsambok, Naturalistic Decision
Making Where Are We Now?,
Naturalistic Decision Making,
edited by Caroline E. Zsambok and Gary Klein,


Lawrence Erlbaum Associates, Inc., 1997
13
About 500 B.C
1916
1979
1937-39
1988-89
1940
1992
1942
The first Conference on Computer Generated Forces
and Behavioural Representation (CGFBR)
conference took place in the States. It becomes
an annual event providing a forum for the
exchange of information on computer generated
forces and the representation of behaviour within
these forces.


http//www.sisostds.org/index.htm
14
About 500 B.C
1988-89
1916
1979
1937-39
1995
1940
1992
1942
In the early 1990s there were so many military
simulations that not one had an accurate count
of them. Each organisation that saw a need for
simulation rolled their own, with little reuse
of what already existed. There was a clear need
for simulations (built by different
organisations) to be interoperable, and for
simulation components to be reusable for
multiple purposes. In 1995, the High Level
Architecture (HLA) supported by DoD became an
IEEE standard of assuring interoperability and
reusability of defence models and simulations.
F. Kuhl, et al,
Creating Computer Simulation Systems, Prentice
Hall PTR, 1999
15
About 500 B.C
1988-89
1916
1979
1937-39
1995
1998
1940
1992
1942
In 1998, I detect a change. Either the dinosaurs
are dying off, or are evolving into a species
which accepts what simulation can do. The
young, brought up in a virtual reality
environment, expect it. What was often resisted
by an older generation (we must have the real
thing or it is no good) is now replaced by an
expectation of the employment of simulation
technology as a necessarily step in the
operation of the equipment itself.
Ian Strachan, Foreword, Janes
Military Training Systems, First Edition,
1998-1999
16
About 500 B.C
1988-89
1916
1979
1937-39
1995
1998
now
1940
1992
1942
? getting wider awareness in DND ? establishing
new administrative group within DND/CF ?
starting new facilities ? having industry
support within Canada ? attracting more
investment ? creating more powerful tools or
components ? being accepted by the younger
generation, e.g., cadets at RMC ? becoming an
component of C2IS infrastructure ? creating
corporation within DND, and ? pushing Canada to
catch up
17
About 500 B.C
1988-89
1916
1979
1937-39
1995
1998
now
2020
1940
1992
1942
Twenty years from now, it is predicted that the
Integrated Single Battlespace will be a reality.
This is a battlespace in which military
operations will no longer be characterised
as land, sea or air. Instead there will be a
single battlespace in which land, maritime and
air forces will be directed, targeted and
supplemented by a new generation of
intelligence, surveillance, information and
communications systems offering a step change in
military capability.
MGen K. R. Pennie, et al, Foreword,
Symposium on Creating
the
Canadian Forces of 2020, April 26-28, 2000
18
Potential Challenges

• Interoperability
• Updating models within war game simulators,
• Taking advantages of high performance computers
• Command agents
• Education

19
An example command agents
To replace the live command entity (operators)
with a software command entity (a command agent).
Software command entity (command agent)
Real world C2 network
ModSAF
Live command entity(operator)
DIS
20
Kleins team of researchers studied 156 decision
points where fire-fighting commanders made
decisions. The result is presented as
follows. Gary Klein, Sources of Power, How
People Make Decisions, the MIT Press, 1998
Klein explains this decision-making strategy as
singular evaluation approach, i.e., the decision
maker looked at several options yet never
compared any two of them. He/she will mentally
simulate each in turn, rejected it, and turned to
the next most typical option. This decision
making process is known as recognition-primed
decision model, which can be simplified here.
21
d6, ?0º
In theory The attack could take many forms. In
practice aggressive attack see the shaded
arrow. conservative attack see the blank
arrows.
2 km
4 km
6km
8km
22
A neural networks approach to simulate the ways a
real commander making battle field
decisions. (Francois takes over from here.)
23
Conclusion Artificial intelligence is changing
the way of how military command and control
systems work. But one thing will unlikely be
changed only human intelligence commands.