Behaviour in OOTW Simulators - - PDF document

▶
behaviour in ootw simulators
SMART_READER_LITE
LIVE PREVIEW

Behaviour in OOTW Simulators - - PDF document

Mar 28, 2024 344 likes •427 views

Behaviour in OOTW Simulators zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA AI for OOTW Representing Plausible ~~ HVR C O N S U L T I N G S E R V I C E S LIMITED 0 0 1 4 2 0 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 0 0 1 4 2 0

slide-1
SLIDE 1

AI for OOTW

Representing Plausible Behaviour in OOTW Simulators zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

~~

HVR C O N S U L T I N G S E R V I C E S LIMITED

SELBORNE H O U S E . M I L L L A N E . A L T O N . H A M P S H I R E G U 3 4 2QJ 0 0 1 4 2 0 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

87977

0 0 1 4 2 0 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

8 9 8 1 9

@

w w w . H V R G r o u p . c o m

slide-2
SLIDE 2

,

Dr Stephen Curram zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

HVR Consulting Services Ltd steve.curramQ hvr-csl.co.uk

*

Dr Mark Cusack

QinetiQ, Malvern e

Mirek Gliniecki

HVR Consulting Services Ltd

@ zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

  • Background

Purpose of overatl study is to create simulators that allow commanders who may be involved In OOTW to experience the sorts of situations they may be faced with : zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0 Multiple factions and agencies 0 Highly political

environment

0 Restrictive ROE 0 Media coverage and public opinion 0 Multiple information sources

Background (2) 2 "

b

0 Purpose of slmulators is

to provide a challenging

0 The slmulators are 0 They will conform to

the High Level Architecture

0 Use Java and JavaBeansTM

technology

0 AI is

  • nly one

component zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

  • f

the study

* Pilot Is a peace-keeping

scenario and thought provoking envlronment tools (HLA) requirements designed to be anatysis

1

slide-3
SLIDE 3

@ zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Study Approach

~~
  • - - . ~ - ~ - ~

ldentlficatlon of behavlour types and requirements zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Review of technology

end case studies to

  • f zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Al

for zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA OOTW

slmulstors generate shortlist of approaches for further InVestIgatlofI Develop stand-alone AI oblects for analysis

0 T8M AI objects In an Integrated environment

Use

In pllot slmulstlon model

.

L

I

Alms of the AI

3

  • 0 Provide plauslble behaviour In simulation

from a player's perspective

  • computer generated actors

I

  • occurrence of events

0 Qive player somethlng to think about 0 Be responsive to player's actions

Not to be too transparent - player should be immersed In environment rather than trylng to be& the Al- zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

G

Itss ~ruficiai Intelligence, BUT

CI

I t la more Important that behaviour-

plausible to a player rather than for the commei

.

iter gamesthan traditional mllitary models 2

slide-4
SLIDE 4

Approaches Considered

2 = * ;

Random Sampling

  • doss not

take ICCOYIII zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

  • f

.

Knowledgbsued system8 (& mlaM)

~ u n

be

MO

  • condltloM. though not

mllmly

  • pmdlGbble. could

be comblned

WWh O l h W

apprOaChM pmdlst.ble ll N I

  • a n

simple, Whlb MM zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Of zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

S M n g

UP

falls rapldly

M rub8 become

mom COmplW

F q

Loglc- NIW

am gonerdly elmplar but mom

powMlul th.n @Se, made some experlma In

cholw

  • f hmy nglona, Ilnguletis

m a

can hdp h l m m

In

twaLlng SCMUIO~

.

CubBu.d Fmasonlng -

full

sy.1.ma can b t l m e

conaumlng to set up and have overhaad zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

  • t shell,

ahnplu wrsbna may be useful for Rmgnlalng when scr1pt.d went. u m

  • ccur

P

E

Approaches Consldered (2)

nnlm

S m M.chlMe - commonly

uwd

In computer games, elmple

lo zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

snl up. u n

go1 pndktsble

(or p1ay.r

F q

8uQ Machl-

  • adapmtlon
  • f

f l n b

Stale

Machlnaa urlng zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

funy NIM, h m l n g

popululn computer gMM. memb.nhlp funcllone could be wed for nndom aampllng

fauora on bhavlour, can b. drmcult to ml up,

dlMcully

d valldallon

mp"tatlon but Ums consuming and dllRcult to 9.1 up and vdldam Intemdive almulsllon B a y ~ l m

Bsll.1 W o r k s

  • pamrful

mpnamtatlon

  • f

Neural Natworke (varlow h/pw) - md.ptln, p

  • w

1 I ~ l Q e d c

Algorkhmr -

adaptlw but too SIW

(or

Short-llsted Approaches

2 " CI Finite State Machine (FSM)

  • for *Impls Whavlour
  • m

. 1 .

gensrlc FSM obJset to which "

t u and hnamons smsdded

  • MSybSdUp

3

slide-5
SLIDE 5

@ zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Short-listed Approaches zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Fuzzy State Machine (FuSM)

  • zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

for more complex behaviour

  • linguistic or numerical Inputs
  • smoother transltions between states
  • create generlc FuSM to which terms and fuzzy rules
  • state can be selected from centre of gravity or using

canbeadded random samollns from fuuv resion

I

I

L

@ Short-listed Approaches

"

0 Simplified Case-Based Reasoning

  • prescrlpted events or event generators
  • conditions for launching events, lncludlng time
  • random element
  • generlc object that llnks to event and allows

windows condltlons to be speclfled based on other objects present in simulation

Y

I @

Behaviour Type - Attitude 2 "

e Attitude of an actor towards other actors

  • may differ at different levels, e.g. factlon leaders
  • use Finite State Machlne or Fuzzy State Machine

towards zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA UN,

  • r factlon soldier towards UN

1 1 1 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

c- Y

1

4

slide-6
SLIDE 6

Behaviour Type zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

  • Negotiation

% E ! M zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

b

, . I l l l W U "I~"j,"l.-l~nMX*IXU> zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

"1"

0 Negotiation

between actors

  • player has role as a mediator, and suggests

compromise solutions

  • response o

f actors depends on the weighted dlstance

from thelr ideal solution and thelr attitude towards the negotiatlons

  • Use Fuzzy State Machine
  • likely to be iterative

process over time zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

6 % Behaviour Type - Negotiation zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

3liEZE;

L zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

  • u

_ _ _ 1

L

C I Y

\

/

I @

Behaviour Type - Event zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0 Scripted events need

to occur at plausible times

  • e.g. negative newspaper stories on refugee camp

conditions only when conditions are bad

  • I n k *

tkiLb*om

5

slide-7
SLIDE 7

. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

aP zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Current Work

IC zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

6