A High-Level Language for Military Fusion Problems

A High-level Language for Military
Fusion Problems
Richard Scherl
Computer Science
Department
Monmouth University
1
Knowledge Fusion
Research Workshop
October 20 - 22 2004
Cognitive Robotics/GoLog



2
October 20 - 22 2004
Integrating reasoning,
perception and action
within a uniform
theoretical and
implementation
framework.
Logic-based.
High-level language.
Knowledge Fusion
Research Workshop
The Situation Calculus



3
McCarthy and Hayes 1969.
A predicate calculus formalization of states,
actions, and effects.
Reiter 1991,2001.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
GOLOG







4
GOLOG -- AlGol in logic.
Sequences.
Nondeterministic choice of actions.
Nondeterministic choice of arguments.
Conditions,
While loops.
Recursion.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Plans Vs Computer Programs




5
There is a long tradition of viewing plans as
computer programs. (Green, manna and
Waldinger).
There are many problems with this view of
plans.
An agent may not know whether a test is true.
Agents may not know enough to execute the
action.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Incomplete Knowledge


6
Generally, agents do not have complete
knowledge of the world.
Formalism must distinguish between what is
true in the world and what the agent knows.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Incomplete Knowledge (Cont)



7
Agents must reason about:
Actions that produce knowledge --- perception,
reading, comunicative acts.
The knowledge prerequisites of actions.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
References



8
Levesque, Reiter, Lesperance,Lin, Scherl.
GOLOG. JLP, 1997.
Reiter. Knowledge in action: logical foundations
for specifying and implementing dynamical
systems. MIT press, 2001.
Scherl and Levesque. Knowledge, action, and
the frame problem. AIJ, 2003.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Goal

9
Can a similar sort of language be used for the
specification of high-level and flexible plans
useful in domains relevant to the military and
also homeland security?
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Example Scenarios


10
Silent prairie: agricultural bio-terrorism
exercise developed by the national strategic
gaming center at the national defense
university
Battlespace challenge problem: road to war
developed by signal solutions for ARL
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Silent Prairie


11
October 20 - 22 2004
Foot and mouth disease
suspected in north
Carolina and Kansas
Bio-terrorism considered
a possibility
Knowledge Fusion
Research Workshop
Steps



12
FBI notified – but the 82nd airborne recently
returned to north Carolina from Iraq.
Need to notify state officials and farm
companies. All leaders of states to which N.
Carolina cattle are shipped must be notified.
Send sample to plum island facility for analysis.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Steps (Cont)






13
Start initial containment strategy (quarantine
zones).
Notify governor and other state officials.
Notify USDA, FDA.
Enact regional containment strategy.
Notify DHS and DOD.
Check to see if there was “chatter” about FMD.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Sample Plan
While ¬knows(x State(x)Considered(x))
( X).state(x)?;
If ¬Kwhether(Shipto(f,a,x)) then
sense_ship(f,a,x) endIf;
If knows(Shipto(f,a,x) then
notify(f,a,x) endIf;
assert(Considered(x));
endWhile;
14
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Battlespace Challenge Problem




15
Signal Solutions
2015 North Korea
Kim Jung-Il dies
Civil War
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Forces Involved




16
U.S.
ROK.
DPRK Hardliners.
DPRK Reformers.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Determining Friend from Foe



17
Sensors(ELINT, SIGINT)
HUMINT
COP needs to indicate which forces are friends
and which are foes (DPRK Hardliners) based
upon both intelligence information and
inferences.
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Rules in Knowledge-Base
x ,y hardliner(x) engages(x,y)
 ¬ hardliner(y)
y,z
x supplies(x,y)  hardliner(y)
 supplies(x,z)
 hardliner(z)
18
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Identify Threats to Inchon Airport


19
Sensors and human intelligence are utilized to
identify threats.
Which U.S. Units can deal with the threats?
How long would it take them to arrive at
location of threat given current conditions?
October 20 - 22 2004
Knowledge Fusion
Research Workshop
20
Knowledge Fusion
Research Workshop
October 20 - 22 2004
Implementation



21
GoLog interpreter written in Prolog
Integration into Jade agent platform
Http://jade.tilab.com
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Further Topics






22
Concurrency (ConGolog, IndiGolog)
Exogenous actions
Probabilistic action occurrences and effects
Ability
Time
Integrating semantic web ontology languages
October 20 - 22 2004
Knowledge Fusion
Research Workshop
Theory of Actions (Cont)




23
Real time, resource bounded behavior
Belief revision
Execution monitoring and failure recovery
Automated plan construction
October 20 - 22 2004
Knowledge Fusion
Research Workshop

Download Report