2006 the itsimbw environment for simulation and decision support philipp hügelmeyer dr. timo...
TRANSCRIPT
2006
The ITSIMBW Environment for Simulation and Decision Support The ITSIMBW Environment for Simulation and Decision Support
Philipp Hügelmeyer
Dr. Timo Steffens
2006
Features of ITSimBW
Scalability
Focus on Cummunication
Freedom of modelling
Viewing alternatives graphically (decisions and/or events)
Basic principles of technical realization
Language for Agent-based Modelling of Processes and Scenarios (LAMPS)
Graph Metrics
Roadmap for the next steps (Gis, HLA …)
Features of ITSimBW
Scalability
Focus on Cummunication
Freedom of modelling
Viewing alternatives graphically (decisions and/or events)
Basic principles of technical realization
Language for Agent-based Modelling of Processes and Scenarios (LAMPS)
Graph Metrics
Roadmap for the next steps (Gis, HLA …)
OOutlineutlineOOutlineutline
2006
Global view Global view
Mission troups act as one single agent Mission troups act as one single agent
SScalability calability Multi-Scale Approach SScalability calability Multi-Scale Approach
Strategic levelStrategic level
HQ in Europe
Mission A-Land
SatCom
Time scale 1 minute Time scale 1 minute Local effects are modeled Local effects are modeled
More detailed agents More detailed agents
Time scale 1 second Time scale 1 second
Tactical levelTactical level
2006
SScalability calability Multi-Scale Approach SScalability calability Multi-Scale Approach
functionality (how detailed are functions/effects modeled/simulated ?)
functionality (how detailed are functions/effects modeled/simulated ?)
Scalable complexity: Scalable complexity:
time (in which time steps will the simulation proceed ?)
time (in which time steps will the simulation proceed ?) space (which spacial scale will be chosen for modeling?
space (which spacial scale will be chosen for modeling?
changable during run time changable during run time at different places different changes possible at different places different changes possible
system can be built in several stages (also depending on modeling efforts) system can be built in several stages (also depending on modeling efforts)
Granularity in modeling should be further refined only if it makes sense Granularity in modeling should be further refined only if it makes sense
2006
SScalability calability which level of accuracy is needed?which level of accuracy is needed? SScalability calability which level of accuracy is needed?which level of accuracy is needed?
Level of accuracy Level of accuracy
Level of significanc
e
Level of significanc
e
2006
Assumptions
/ models
Complexity
Plausibility
low
very high
Level of significanc
e
Level of significanc
e
SScalability calability which level of accuracy is needed?which level of accuracy is needed? SScalability calability which level of accuracy is needed?which level of accuracy is needed?
Level of accuracy Level of accuracy
2006
Results
Correctness Usefulness
very high
low
Level of significanc
e
Level of significanc
e
SScalability calability which level of accuracy is needed?which level of accuracy is needed? SScalability calability which level of accuracy is needed?which level of accuracy is needed?
Level of accuracy Level of accuracy
Complexity
Plausibility
low
very high
Assumptions/ models
2006
high
Level of significanc
e
Level of significanc
e
SScalability calability which level of accuracy is needed?which level of accuracy is needed? SScalability calability which level of accuracy is needed?which level of accuracy is needed?
Level of accuracy Level of accuracy
Results
Correctness Usefulness
very high
low
Complexity
Plausibility
low
very high
Assumptions/ models
high
high
high
2006
extremely high low
low
low
A simulation model must be adjustable to the appropriate scope A simulation model must be adjustable to the appropriate scope
This scope is changing from task to task This scope is changing from task to task
Level of significanc
e
Level of significanc
e
SScalability calability which level of accuracy is needed?which level of accuracy is needed? SScalability calability which level of accuracy is needed?which level of accuracy is needed?
Level of accuracy Level of accuracy
high
Results
Correctness Usefulness
very high
low
Complexity
Plausibility
low
very high
Assumptions/ models
high
high
high
2006
Language for Agent based Modelling of Processes and Scenarios
Based on high-level Petri-Nets
Actions describe the effects that agents apply to themselves or to other agents
Relations denote the links between places, actions, and agents.
Agents correspond to the conditions of Petri nets
Graphical and textual representation
Not restricted to a timing model
Time scaling is possible
Language for Agent based Modelling of Processes and Scenarios
Based on high-level Petri-Nets
Actions describe the effects that agents apply to themselves or to other agents
Relations denote the links between places, actions, and agents.
Agents correspond to the conditions of Petri nets
Graphical and textual representation
Not restricted to a timing model
Time scaling is possible
LAMPSLAMPSLAMPSLAMPS
2006
LAMPSLAMPSLAMPSLAMPS
Enemy spottedResistance broken
Combat
1
Inf A
2006
Agent Agent
Autonomous active elements Autonomous active elements
What are the active system components ? What are the active system components ?
Persons Machines, vehicles Objects, Bridges, …
Agents carry out the system processesAgents carry out the system processes
…
AAgent Architecture gent Architecture FTA AAgent Architecture gent Architecture FTA
2006
Agent Agent
Tag Tag
Agents cause effects by producing tagsAgents cause effects by producing tags
How do active elements interact ?How do active elements interact ?
Causing effects Causing effects spoken messages physical interaction effects of weapons …
AAgent Architecture gent Architecture FTA AAgent Architecture gent Architecture FTA
2006
Agent Agent
Tag Tag
Agents are self-responsible for their own behavior, there is no direct action from outsideAgents are self-responsible for their own behavior, there is no direct action from outside
Taking effects Taking effects reading a message realizing physical effects taking orders …
AAgent Architecture gent Architecture FTA AAgent Architecture gent Architecture FTA
How do active elements interact ?How do active elements interact ?
2006
Agent Agent
Tag Tag
Tag types describe the form, tag instances contain the concrete data Tag types describe the form, tag instances contain the concrete data
Tags have a typeTags have a type
order form general message physical power (effects)
…
AAgent Architecture gent Architecture FTA AAgent Architecture gent Architecture FTA
How do active elements understand each other ?
How do active elements understand each other ?
2006
Agent Agent
Tag Board Tag Board
Tag boards serve as models for regions of interaction Tag boards serve as models for regions of interaction
What is the radius of an interactionen ? What is the radius of an interactionen ?
Regions of local interaction Regions of local interaction rooms, streets
radius of physical effects communication links …
Tag Tag
AAgent Architecture gent Architecture FTA AAgent Architecture gent Architecture FTA
2006
Agent Agent
Agents can link together several regions of local interaction Agents can link together several regions of local interaction
What is the system‘s topology ? What is the system‘s topology ?
Tag Board Tag Board Tag Tag
AAgent Architecture gent Architecture FTA AAgent Architecture gent Architecture FTA
2006
Real topologies can be modeled 1:1Real topologies can be modeled 1:1
Agent Agent
Tag Board Tag Board Tag Tag
AAgent Architecture gent Architecture FTA AAgent Architecture gent Architecture FTA
What is the system‘s topology ? What is the system‘s topology ?
2006
3D voxel model for environment 3D voxel model for environment
landscape model landscape model
MModelling odelling Environment MModelling odelling Environment
2006
obstruction of radio communication obstruction of radio communication electronic countermeasures electronic countermeasures
3D voxel model for environment 3D voxel model for environment
landscape model with radio beam at a hilllandscape model with radio beam at a hill
weather, clouds, mist, smoke weather, clouds, mist, smoke visibility visibility
shapes of units shapes of units
computationally fast computationally fast
scalable in resolution scalable in resolution
appropriate for multi-scale features appropriate for multi-scale features
… …
MModelling odelling Environment + Communication MModelling odelling Environment + Communication
2006
MModes of use odes of use Nondeterministic Wargaming MModes of use odes of use Nondeterministic Wargaming
ITSimITSimManeuver
Control
Red
Persons CGF
Blue
Persons CGF
+ Data Farming
ITSimITSimPlanner
Szenario Data Base
1. Maneuver mode 1. Maneuver mode
4. Szenario Data Base 4. Szenario Data Base
Data Minig
2. Get empirical knowledge about the szenario
2. Get empirical knowledge about the szenario
3. Planner / Strategy-Builder
3. Planner / Strategy-Builder
Variety of
possible traces for
different outcomes
Variety of
possible traces for
different outcomes
Evaluated
non-deterministic
szenario
Evaluated
non-deterministic
szenario
„To know what might happen“„To know what might happen“
2006
ITSimITSimManeuver Control
Red
Persons CGF
Blue
Persons CGF
Data Farming
ITSimITSimPlanner
Data Minig
Szenario Data Base
WargamingSzenario
Description
C4I with szenario data base C4I with szenario data base
more coherent behavior more coherent behavior „no improvisation needed“ „no improvisation needed“
ITSimITSim
HQ
Armed Forces
Combat Support
MModes of use odes of use Combat Support MModes of use odes of use Combat Support
2006
1. Multi-scale simulation of scenarios
1. Multi-scale simulation of scenarios environment environment
communication communication reporting structures reporting structures order structures order structures effects effects
2. Analysis of variables
2. Analysis of variables communication load communication load
risk measured risk measured probabilities of failure probabilities of failure
3. Scenario analysis
3. Scenario analysis alternative scenarios alternative scenarios
critical paths critical paths demands, costs, … demands, costs, … time-dependent factors time-dependent factors operation orders, bill of materials operation orders, bill of materials training maneuvers training maneuvers
… …
… …
1
2
4. Online-C2System (decision support)4. Online-C2System (decision support)
comparison between stored and current mission graph based on similarity measures. comparison between stored and current mission graph based on similarity measures.
offl
ine
pre
para
tion
onli
ne
mis
sion
scenario database
4
3
Data
Farm
ing
MModes of use odes of use Combat SupportMModes of use odes of use Combat Support
2006
Scalable in space, and functionality
Finishing the LAMPSSys-Core
Finishing of the HLA-Impementation
Adapter for statistical Analysis and GIS
Scalable in space, and functionality
Finishing the LAMPSSys-Core
Finishing of the HLA-Impementation
Adapter for statistical Analysis and GIS
OutlookOutlookOutlookOutlook
2006
In a simulation the time is run in disc cycles
A state S in a point in time t is described by the states of all the individual agents
A simulation run is fully described by the set of all states S in all time steps
In a simulation the time is run in disc cycles
A state S in a point in time t is described by the states of all the individual agents
A simulation run is fully described by the set of all states S in all time steps
Graph-MetricGraph-MetricGraph-MetricGraph-Metric
mttt ...21
},...,,{ 21 nt sssS
},...,,{ 21 hrrrR
2006
Graph-MetricGraph-MetricGraph-MetricGraph-Metric
A1t A2t A3t
A1t’ A2t’ A3t’
2006
For all attributes of all agents in all time steps this is an overwhelming amount of information
=> we pick interesting attributes beforehand:
The position (in the 3D space)
The loss of own troops
The time to accomplish the mission goal
The amount of effects to act upon the opponent
For all attributes of all agents in all time steps this is an overwhelming amount of information
=> we pick interesting attributes beforehand:
The position (in the 3D space)
The loss of own troops
The time to accomplish the mission goal
The amount of effects to act upon the opponent
Graph-MetricGraph-MetricGraph-MetricGraph-Metric
2006
Graph-MetricGraph-MetricGraph-MetricGraph-Metric
St
r
B
C
D
E
F
G
I
H
A
BA
DC
E F
G H I
100 30 70
80 20 100
100 45 55
Clusters over time and runsClusters over time and runs
Markov tree with propabilities at the edges
Markov tree with propabilities at the edges
2006
TTechnical Realization echnical Realization Agent Technology TTechnical Realization echnical Realization Agent Technology
The world consists of concurrent interacting agents (also non-human parts of the game are modeled as agents)
The real-world behavior is simulated in detail on a computer as an artificial reality
Humans and real systems outside the simulator are coupled with this
artificial reality by special communication agents
The world consists of concurrent interacting agents (also non-human parts of the game are modeled as agents)
The real-world behavior is simulated in detail on a computer as an artificial reality
Humans and real systems outside the simulator are coupled with this
artificial reality by special communication agents
2006
TThank you for your attention !hank you for your attention !TThank you for your attention !hank you for your attention !