Model Development of Large-Scale DoD Systems-of-Systems
Santiago Balestrini Robinson
[email protected]
2008 SCS Conference, June 17th, 2008 Edinburgh, Scotland
Guggenheim School of Aerospace Engineering Georgia Institute of Technology Atlanta, GA 30332-0150 http://www.asdl.gatech.edu
Model Development of Large-Scale DoD System-of-Systems
[email protected]
1
Outline Motivation Design for Performance Æ Capability-based Acquisition
The Evolution of Science: Towards the Problems of Organized Complexity Brief historic overview of Complexity Science
Modeling and Simulation Techniques for Complex Systems Network Modeling Analyzing Large-scale Systems-of-Systems using Network Models Quantifying DoD Architecture Framework Products The Principal Components of a Network Future Work Model Development of Large-Scale DoD System-of-Systems
[email protected]
2
Map the System to the System-of-system
Capability-Based Acquisition System requirements are implicit in the System-of-Systems (SoS) requirements Source: Courtesy of Kelly Cooper (SBE 2006 Presentation)
Need to assess the impact the system has on the System-ofSystems SoS Capabilities = f ( System Performance ) Image Source: http://www.clubs.psu.edu
Need the ability to estimate this transfer function
Model Development of Large-Scale DoD System-of-Systems
[email protected]
3
Characteristics that Complicate the Mapping Interactions
Emergent Behavior
The overall behavior of the SoS is sometimes more dependent on how the entities interact than their individual capabilities
Nonlinearity Small variations in the causes produce large variations in the effects, or viceversa
Intelligence Cognitive processes must be modeled to capture the effects of decision making
It is often difficult to predict the behavior of the overall system from the study of the parts in isolation
Adaptation Systems tend to learn from their environment and surrounding agents
Dynamic Behavior The system’s behavior occurs over time and it must be studied as a function of time
Hierarchies The system tends to organize itself into hierarchies (e.g., Command and Control)
These are the characteristics of complex systems
Model Development of Large-Scale DoD System-of-Systems
[email protected]
4
The Evolution of Science In 1947, Weaver analyzed the history of science from the 17th century and noticed a pattern He recognized that until then there had been two main efforts Between the 17th and 19th centuries science focused on problems with only a handful of variables In the 20th century statistical methods were developed to handle problems with large number of variables
This left a considerable range of the problems faced by science without solid foundations The advent of the computer enabled the study of the area in-between the two camps This in-between field has come to be known as complexity science Warren Weaver’s Portrait Source: http://osulibrary.oregonstate.edu/
Methods from Ecology, Psychology, Biology, etc…
Computer Image Source: http://www.nea.com/ Figure based on Weaver, W., “Science and Complexity,” American Scientists, 36(536), 1948.
Model Development of Large-Scale DoD System-of-Systems
[email protected]
5
The Evolution of Science In 1947, Weaver analyzed the history of science from the 17th century and noticed a pattern
Between the 17th and 19th centuries science focused on problems with only a handful of variables In the 20th century statistical methods were developed to handle problems with large number of variables
Newton
Detail of the Entities
He recognized that until then there had been two main efforts
Problems of Simplicity Problems of Organized Complexity
Boltzmann
This left a considerable range of the problems faced by science without solid foundations The advent of the computer enabled the study of the area in-between the two camps This in-between field has come to be known as complexity science Warren Weaver’s Portrait Source: http://osulibrary.oregonstate.edu/
Problems of Disorganized Complexity 2×100 2×102 2×104 2×106 2×108 Number of Elements Composing the System
Methods from Ecology, Psychology, Biology, etc…
Computer Image Source: http://www.nea.com/ Figure based on Weaver, W., “Science and Complexity,” American Scientists, 36(536), 1948.
Model Development of Large-Scale DoD System-of-Systems
[email protected]
6
Modeling Techniques for Complex Systems No ideal method
Technique Complexity
ABM most suitable, but most difficult to implement and validate NM are the easiest to implement but do not capture the dynamic behaviors or intelligence
Methods are not exclusive, but complementary Others techniques considered but not discussed: Markov Simulation, Petri Net Simulation, Dynamical Systems, Cellular Automata
System Dynamics Models
Agent-based Models
= Excellent
= Very Good
= Good
= Poor
= Very Poor
Network Models
Discrete Event Simulations
Nonlinearity
Interactions
Intelligent Agents Represent Hierarchies Emergent Behavior Adaptation
Dynamic Behavior
Ease of Creation
Ease of V&V
Legend:
Complementary: Coarse modeling can guide detailed modeling Model Development of Large-Scale DoD System-of-Systems
[email protected]
7
Modeling Techniques and the Paradigm of Complexity
how many entities?
in how much detail?
The modeling techniques map to a distinctive Pareto-front in the continuum
Notional plot based on literature reviews of different applications of the modeling techniques
As more entities are modeled, the techniques require that the entities be simplified − System Dynamics is an exception, because in its pure form it assumes that there are an infinite number of entities
Anything in the below or to the left of the line is a dominated solution
Network Models
Discrete Event Simulation
System Dynamics
Agent-based Models
Detail of the Entities
The question of how much can be modeled can be considered to be a question of
2×100 2×102 2×104 2×106 2×108 Number of Elements Composing the System
Large-Scale SoS Æ Network Models
Model Development of Large-Scale DoD System-of-Systems
[email protected]
8
Why not Start with Sea Basing? MPG
Ground Combat Element 3 Inf Bn 665 Personnel
2 Tank Co 47 M1A1
2 LAR Co 27 LAV-25
2 AA Co
ESG
47 EFV
3 Arty Btry 1 EFSS Btry 1 HIMARS Btry
CSG
2 Cbt Engr Co
Can See Can Kill
Aviation Combat Element
Can Supply Can Carry
4 AV-8B Sqdn 40 AV-8B
1 HMLA Sqdn 12 AH-1W / 12 UH-1N
CLF
Can Talk
2 F/A-18 Sqdn 24 F/A-18
1 EA Sqdn
2 CH-53 Sqdn 32 CH-53D/E
4 EA-6B
1 KC-130 Sqdn 6 KC-130
4 CH-46 Sqdn 48 CH-46E
LCAC
T-Craft?
ALDS?
HSV?
Too many elements, too many interactions Æ too complex Difficult to demonstrate trends, computationally expensive Æ Can’t experiment Currently developing a representative model Model Development of Large-Scale DoD System-of-Systems
[email protected]
9
Modeling Systems as Networks
Communication
Engage
ge Eng a
ure n at
S ig
Eng age
ge
Signature
Signature
Signature
r de
Ord er
Or
ga
ge ga En
En
ation unic m m Co
Order
Graphs are ideal for representing pairwise relations between large numbers of objects One of a few modeling tools that enables truly holistic quantitative analysis Emphasis is placed on the structure of the macro system, rather that in decomposing the entities that make the system
System behavior can be inferred from studying the structure of the function-based network Model Development of Large-Scale DoD System-of-Systems
[email protected]
10
Simulating an Engagement with Networks Proto-Engagement One of each, in matrix form it becomes the Graph Generation Matrix (GGM)
How many of each? age
Engage
Be Detected by Engage Order Communicate
En g
En g age
atu re
Sig n
Signature
Can Can Can Can
Signature
Signature
r
Ord er
de Or
ge ga En
Contains the probability that any two entities are related by the given function Depends on the individual system’s capability, the theater of operations, tactics, etc.
ge ga En
Graph Generation Matrix (GGM)
Communication
on icati m un Com
Order
Proto-Engagement
Graph Generation Matrix
Force Structure
Engagement Matrix
Force Structure Specifies how many systems of each type are involved
Engagement Matrix Represents a functional network in a possible engagement Model Development of Large-Scale DoD System-of-Systems
[email protected]
11
Incorporating Standard Architecture Products OV-5
SV-7
What set of activities constitute a capability
SV-5a
What functions constitute an activity
What systems perform which functions
Can Be Detected by Can Engage
Who can exchange non-automated data
SV-6
Graph Generation Matrix
How good are the entities at communicating and performing their functions
OV-3
SV-4
Who can exchange automated data (e.g., Link16)
OV-5
SV-4a
OV-3
Activity Sequences
Systems-Functions
Info Exchange
SV-5a
SV-7
SV-6
Activities-Functions
Performance Metrics
Info Exchange
Can Order Can Communicate
Model Development of Large-Scale DoD System-of-Systems
[email protected]
12
Testing through Experimentation Create an agent-based simulation of an engagement including the same elements
Address spatial and time-domain complexities addressed explicitly
Plan
Modify the force structure and capabilities of the assets in both models to
Compare force loss ratios
NetLogo Simulation of the Engagement between:
Agents intelligence not trivial
AWACS vector friendly fighters to closest detected targets
Ground radars must be able to communicate with SAM sites
Shooters must lock on for a given period of time before being able to shoot
Available Online at: http://sanbales.googlepages.com
Model Development of Large-Scale DoD System-of-Systems
[email protected]
13
The Principal Components of a Graph 20%
100%
0%
0%
0%
20%
20%
20%
0%
20%
λ1 = 0.00
λ1 = 1.00
17%
19%
14%
20%
14%
25%
Perron-Frobenius Eigenvector (PFE) Node’s contribution to the cycles 27%
23%
λ1 = 1.17
25%
19%
λ1 = 1.32
Associated Eigenvalue Number of Autocatalytic Cycles
Model Development of Large-Scale DoD System-of-Systems
[email protected]
14
Using the Principal Component to Measure a Capability 1.8
1
1.6
0.9
0.8
1.4
0.7
0.6
λPFE 1,net
NET
1 0.5 0.8 0.4 0.6 0.3 0.4
0.2 Blue Capability(PFE λ1,net)
0.2
0.1
Red Casualties 0
0 0.0
0.3
0.5
0.8
1.0
1.3
1.5
1.8
2.0
2.3
2.5
2.8
3.0
3.3
3.5
3.8
4.0
4.3
4.5
4.8
5.0
5.3
Time (hours)
Model Development of Large-Scale DoD System-of-Systems
[email protected]
15
5.5
Fraction of Red Casualties
1.2
Future Work: The rest of the Spectrum?
2.5
Blue Force 2
1.5
Imaginary
1
0.5
0
-0.5
-1
-1.5
-2
Red Force
-2.5
-8
-6
-4
-2
0
2
4
6
8
10
12
Real Model Development of Large-Scale DoD System-of-Systems
[email protected]
16
Conclusions Agent-based Simulation Can capture many characteristics of complexity Very expensive to create, execute, and verify/validate
DoDAF Products OV-5
SV-7
Activity Sequences
Performance Metrics
SV-5a
SV-6
Activities-Functions
Info Exchange
SV-4a
OV-3
Systems-Functions
Info Exchange
Network Models
Network Modeling Captures fewer characteristics of complexity but allows for more encompassing modeling
0.6
Robustness
0.4
Vulnerability
0.2
0
Used in other fields successfully to infer behavior from structure
Focused Higher-Fidelity Models
Capability
-0.2
-0.4
Self-Synchronization
-0.6
0.2
0.4
Can be used to focus the higher fidelity models Model Development of Large-Scale DoD System-of-Systems
[email protected]
17
0.6
0.8
1
1.2
1.4
1.6
1.8
Questions?
Model Development of Large-Scale DoD System-of-Systems
[email protected]
18
