A Practical Process for Simulation Component Reuse Dissertation Proposal Presentation by Robert G. Bartholet 27 May 2005
Dec 30, 2015
A Practical Process for Simulation Component Reuse
Dissertation Proposal Presentation by
Robert G. Bartholet
27 May 2005
2
Committee Members
Worthy N. Martin, Chair
Paul F. Reynolds, Jr., Advisor
John C. Knight
David C. Brogan
Harsha K. Chelliah
Ernest H. Page
3
Thesis Statement
Component selection, applied to reusable simulation components, can be enhanced significantly by including considerations for the utility of component adaptation, simulation specific attributes, and other features that have not been considered in traditional approaches to component selection.
4
Thesis Statement
Component selection, applied to reusable simulation components, can be enhanced significantly by including considerations for the utility of component adaptation, simulation specific attributes, and other features that have not been considered in traditional approaches to component selection.
SIMULATIONCOMPONENT
DATABASE
SIMULATIONCOMPONENTDEVELOPERS
S1
S2
Sn
. . .
COMPONENTSELECTION
TOOL
IDEALFEDERATION
REQUIREMENTS
TH
EO
RY
BESTPRACTICES
M&S Reuse in the Ideal World
6
What is Available?
ExecuteFederation
andPrepareResults
6
DevelopFederationConceptual
Model
2
DesignFederation
3Develop
Federation
4
DefineFederationObjectives
1
AvailableResources
ProgramObjectives
FederationObjectivesStatement
Federation Requirements
FederationConceptualModel
FederationScenario
Initial PlanningDocuments
Allocated Federates
FederationDevelopmentPlan
FOM
FED file
Modified Federates
Scenario Instance
RTI RID File
TestedFederation
TestingData
Test EvaluationCriteria
Reusable Products
UserFeedback
IntegrateandTest
Federation
5
7
Software Reuse
SCAVENGINGSCAVENGING
S/W ARCHITECTURESS/W ARCHITECTURES
TRANSFORMATIONTRANSFORMATION
VERY HIGH LEVEL LANGUAGES
VERY HIGH LEVEL LANGUAGES
APPLICATION GENERATORS
APPLICATION GENERATORS
S/W SCHEMASS/W SCHEMAS
COMPONENTSCOMPONENTS
Krueger, 1992
8
Related Work
• Metrics and Models
• Component Representation
• Selection Techniques
9
Reuse Exemplars
),( ytfdt
dy
10
Motivation
11
Simulation ComposabilityComponent Selection (CS)
r1
r3
r4
r2
r5r6
Rx1
x3
x4
x8
x2
x6x7
x5
r8r7
X
CS: Is there a subset of X of cardinality k or less that covers R?
Example instance when k = 3.
CS is NP-complete.
Proof: reduction from SAT (Page and Opper 1999) and MSC (Petty et al. 2003).
CS can be approximated using GREEDY (Fox et al. 2004).
REQUIREMENTSCOMPONENTS
12
Composability Assumptions
• Components are immutable.
• There exists a master set of components from which all possible sets of requirements can be satisfied.
• Requirements are known a priori and do not change.
Component selectionin the context
of simulation composabilityis inflexible.
13
Thesis Statement
Component selection, applied to reusable simulation components, can be enhanced significantly by including considerations for the utility of component adaptation, simulation specific attributes, and other features that have not been considered in traditional approaches to component selection.
14
Applied Simulation Component Reuse (ASCR)
• Leverage simulation specific characteristics in component reuse
• Exploit adaptability of components to satisfy requirements.
• What have we gained?– We no longer have to assume the existence of a master set
of components.
– We can more flexibly react to changing requirements.
– We can pre-select components based on over-arching simulation specific requirements.
• But…
Exploiting simulation specific characteristics and adaptationchanges component selection!
15
ASCR Model
RCIZUIZU ,
ECI 22:
REZ 22:
MU RR 22:
BEHAVIOR DETERMINATION FUNCTION
BEHAVIOR MAPPING FUNCTION
UTILITY FUNCTION
16
Assumptions and Notation
• θ: Upper bound on the number of requirements that can be satisfied by any one component.
• β: scaling factor which captures the change in utility encountered when a component satisfies multiple requirements
• uxyz: the utility of the xth component satisfying requirement y while satisfying z-1 other requirements.
17
Adapting Components to Satisfy Requirements
r1 r2 r3 r4 r5 r6 r7
x2x2 x2
r1
r3
r4
r2
r5r6
Rx1
x3
x4
x8
x2
x6x7
x5
r7
X
CS CS-ASCRθ=3
18
Computing Utilities
r1
r3
r4
r2
r5r6
Rx1
x3
x4
x8
x2
x6x7
x5
r7
X
CS CS-ASCR
θ=3
r1 r2 r3 r4 r5 r6 r7
u211
u212
u213
u241
u243
u242
u253
u252
u251
β reduces the number of computed utilities.
19
Building a Bin View
r1
r2
r3
r4 r5r6
r7R
x1
x2
x3
x4
X
r1 r2 r3 r4 r5 r6 r7u111 u221u112 u222
u223u113
u321u322
u323
u331u332
u333
u432
u431
u433
u441u442
u443
u251u252
u253
u461u462
u463
u261u262
u263
u171u172
u173
u371u372
u373
REQUIREMENTS
COMPONENTS r1 r2 r3 r4 r5 r6 r7x1 x2
x3
x3x4
x4 x2 x4x2
x1x3
θ=3
20
Building a Set View
u111 u171u112u172
u221 u251 u261
u371 u331 u321u433u443u463 u432
u442u223
u253u263
u252u262
u322u372
u323u333u373 u431 u461 u441
u222u252
u222u262
u332u372
u332u322
u442u462
u462u432
C1C2
C3 C4
Cr1 r2 r3 r4 r5 r6 r7u111 u221u112 u222
u223u113
u321u322
u323
u331u332
u333
u432
u431
u433
u441u442
u443
u251u252
u253
u461u462
u463
u261u262
u263
u171u172
u173
u371u372
u373
Scenario for x2
21
CS-ASCR Definition
r1r2
r3r4 r5
r6
r7
Ru111 u171
u112u172
u221 u251 u261
u371 u331 u321u433u443u463 u432
u442u223
u253u263
u252u262
u322u372
u323u333u373 u431 u461 u441
u222u252
u222u262
u332u372
u332u322
u442u462
u462u432
C1C2
C3 C4
C
CS-ASCR (Informal): Is there an exact cover of R, constructed by choosing no more than 1 element from each Ci, with a total utility greater than k?
CS-ASCR is NP-complete (Bartholet et al. submitted to ACM/IEEE WSC 2005).Proof: By reduction from X3C.Optimization problem is NP-hard.
22
Interesting Effects of θ and β
• θθ = 1, CS-ASCR is in Pθ = 2, complexity of CS-ASCR is openθ >= 3 but bounded, CS-ASCR is NP-completeθ is unbounded, CS-ASCR is exponential
• ββ=1, CS-ASCR is in P
23
Generalizing the Result
r1 r2 r3 r4 r5 r6 r7 r8
θ=3
x1 x3x6
x6x4 x2
Modified definition of θ: utility can be dependent on selection of other components.
x6 x7x6
CS-ASCR-X: ASCR component selection with the modified θ.
Corollary: CS-ASCR-X when θ >= 3 is at least NP-complete (by reduction from X3C).
24
Leveraging Simulation Properties
• Stochastic sampling
• Time
• Event generation
25
Requirement 1: Model ground combat.
Requirement 2: Model air combat.
Requirement A: Provide up-to-date conflict adjudication data no less than once per minute.
Example of Leveraging Time
XX
X
MODEL GC1
MODEL GC2MODEL AC1
LANCHESTER ATTRITION
CALCULATED EVERY HOUR OF LOGICAL
TIME
STOCHASTIC ATTRITION
AGGREGATED EVERY 10 MINUTES OF LOGICAL TIME
STOCHASTIC ATTRITION
AGGREGATED EVERY 5 MINUTES OF LOGICAL
TIME
LOW
UTILITY
HIGH
UTILITY
HIGH
UTILITY
26
Example of Leveraging Time
PRE-SELECTION
MODEL AC1MODEL GC2
TIME SCALE
FACTORED INTO
COMPONENT SELECTION
27
Research Areas of Focus
• Define the problem
• Define the process
• Characterize complexity of ASCR
• Component selection
Sim Specific Characteristics
Adaptation
28
Measures of Success
• Accurately formalized ASCR problem
• Defined a practical ASCR process
• Built practical methods for component selection
• Developed useful utility functions
• Analyzed complexity of critical algorithms in ASCR
29
Expected Contributions
• Creation of a methodology that significantly improves state of simulation component reuse and provides practical methods for component selection
• Improved understanding of complexity of component selection
• Demonstration of how simulation specific properties can be leveraged in component selection
30
Publication Efforts
Bartholet, Brogan, Reynolds, Carnahan. In Search of the Philosopher's Stone: Simulation Composability Versus Component Based Software Design. Proceedings of the Fall 2004 Simulation Interoperability Workshop, Orlando, FL, September 2004.
Brogan, Reynolds, Bartholet, Carnahan, Loitière. Semi-automated Simulation Transformation for DDDAS. Proceedings of the 5th International Conference on Computational Science, Atlanta, GA, May 2005.
Bartholet, Reynolds, Brogan. The Computational Complexity of Component Selection in Simulation Reuse. Submitted to the ACM/IEEE 2005 Winter Simulation Conference, Orlando, FL, December 2005.
Bartholet, Kuang, Son. Intelligent Decentralized Update Management in Real-Time Embedded Applications. Working Draft Completed. Submission in August 2005 to conference TBD.
31
Conclusion
Component selection, applied to reusable simulation components, can be enhanced significantly by including considerations for the utility of component adaptation, simulation specific attributes, and other features that have not been considered in traditional approaches to component selection.
Discussion