modeling and simulation methodology: the challenge of complex endeavors bernard zeigler arizona...
TRANSCRIPT
Modeling and Simulation Methodology: The Challenge of Complex Endeavors
Bernard ZeiglerArizona Center for Integrative
Modeling and Simulation,University of Arizona,
Tucson, AZzeigler @ece.arizona.edu
AI and Computing in Countering Terrorism INFORMS General MeetingOct 13. 2008
Outline
• What are Complex Endeavors?• We need adequate models of
– humans– human-human interactions
• What such models might be based on• Complex Endeavors as Systems of Systems• M&S Environment to Support SoS • Levels of Interoperability• SOA-based Integration and Testing of SoS
Marvin Minsky, The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind, Simon Schuster
Richard E. (Dick) Hayes, Complex Endeavors as Challenges to the Modeling and Simulation Community, Military Modeling and Simulation Conference, Singapore
Suiping Zhou, Human Behavior Modeling and Simulation For Military Operations, Military Modeling and Simulation Conference, Singapore
Complex Endeavors (Richard Hayes) • Formed when a large number of disparate entities form an association for a
limited time to achieve a shared objective• No single leader or commander
–Neither unity of purpose nor unity of command–Composed of independent entities Different traditions, cultures, goals,
priorities, and processes• Interdependence
–No single actor is capable of achieving its relevant goals independently–Actors bring different expertise and resources to the endeavor
• Increasing need for international peace operations• information technology enables collaboration• multinational, interagency, governmental, non-governmental organizations,
private industry, and local authorities
Complex Endeavors are characterized by Human-Human Interactions
•Perceptions of actors about othersotrustocompetenceocross-cultural biases
•Interoperability: share oinformation and knowledgeoawareness (situation characterization)ounderstanding (cause and effect, temporal dynamics)
•Collaboration about purposes, decisions, planning, and execution
ocoalitions without common doctrine oinvolving a variety of actors (e.g. Tsunami, Katrina relief)
r
Limitations of Current Models and Reuse
Models
• Classic Rule Based and Algorithmic Models -- ignore soft factors
• Human in the Loop Models —generalization limited to the types of people who participate
• Simulation Models – Systems Dynamics, Agent-Based, etc., difficult for a policy or decision maker to comprehend, must have faith in black box
Problems in Reuse:• must know the original purposes and assumptions (Experimental frame)•models operate at different levels of abstraction – they cannot communicate with each other• built in biases of developers, new forms e.g., cultural biases
Behavior Modeling Principles (Suiping Zhou)
• Humans are social animals. The social aspect and the animal aspect of a human being are inhibitory to each other.
• Behavior is largely determined by experiences rather than by complex decision rules.
• Behavior is greatly affected by social context, family, friends, colleagues, etc
• Human’s decision-making process consists of multiple layers of micro-level/macro-level interactions.
• Decision making and perception are heavily influenced by emotion and culture
Layered Model of Mind (Marvin Minsky)
Self-Conscious Reflection
Self-Reflective Thinking
Reflective Thinking
Deliberative Thinking
Learned Reactions
Learned Reactions
Values, Censors, Ideals, Taboos
Innate, Instinctive, Urges, Drives
Multiple, Concurrent Ways to think(Learning Processes)
• We are born with many mental resources.• We learn more from interacting with others.• Emotions are different Ways to Think.• We learn to think about our recent thoughts.• We learn to think on multiple levels.• We accumulate huge stores of commonsense
knowledge.• We switch among different Ways to Think.• We find multiple ways to represent things.• We build multiple models of ourselves.
Federations of Models • No single model or approach to modeling will be adequate to meet the
needs for validity, reliability, and scalability.• Federations of models will be needed for different:
– Levels of Analysis– Functions (Communications, Logistics, Decision Making, etc.)
• Models in Federations should:– Be developed and tested together– Be modular and inform one another
• Be based on compatible underlying assumptions and parameters • Be transparent
Interoperation vs Integration*
Interoperation of components• participants remain autonomous and
independent• loosely coupled• interaction rules are soft coded• local data vocabularies persist• share information via mediation
Integration of components• participants are assimilated into whole,
losing autonomy and independence• tightly coupled• interaction rules are hard coded• global data vocabulary adopted• share information conforming to strict
standards
* adapted from: J.T. Pollock, R. Hodgson, “Adaptive Information”, Wiley-Interscience, 2004
NOT Polar Opposites!
reusabilitycomposability
efficiency
syntactic
semantic
pragmatic
Linguistic Levels of Interoperability
LinguisticLevel
InteroperabilityDemonstrated if:
Example
Pragmatic – How information in message is used
The receiver reacts to the message in a manner that the sender intends
A commander’s order is obeyed by the troops in the field as the commander intended. (This assumes semantic interoperability.)
Semantic – Shared understanding of meaning of messages
The receiver assigns the same meaning as the sender did to the message.
An order from a commander to multi-national participants in a coalition operation is understood in the same manner despite translation into different languages.
Syntactic – Common rules governing composition and transmitting of messages
The consumer is able to receive and parse the sender’s message
A common network protocol (e.g., IPv4) ensures that all nodes on the network can send and receive data bit arrays while adhering to a prescribed format.
Fundamental Research in M&S
• Discrete Event System Specification (DEVS )
• Provides sound M&S framework
• Derived from Mathematical dynamical system theory
• Supports hierarchical, modular composition
• System Entity Structure: ontology framework for M&S
• Distributed simulation, web-based, SOA-based
• Linguistic levels of interoperability (syntactic, semantic, pragmatic)
• M&S Simulation interoperability standards
Heterogeneous-Formalism Modeling agents
Discrete-event,Models
landscape
Discrete-time, Cellular Automata Models
Knowledge Interchange
Broker
interactions
Knowledge Interchange Broker (KIB) provides its own distinct formalism and realization
Separately accounts for domain-neutral and domain-specific modeling
Removes the need for composed models to have detailed knowledge of each other
NSF ERE Biocomplexity in the Environment program
NSF Science of Design Program
Design of Adaptive Service-based Software Systems with Security and Multiple QoS Requirements
• Develop a SOA-based DEVS simulator to aid design and evaluation of flexible and configurable SOA-based software systems
• support design of SOA systems able to adapt to changing tradeoffs among timeliness, throughput, accuracy, and security
QoS Adaptation
QoS Monitoring
SBS
Sim
ulat
ion
& Q
oS
me
asur
emen
ts
QoS Expectations
Adaptationcommands
ProduceEvents
Resources
ExtrageneousEvents
AffectQoS
ConsumeResources
Measure changes of resource states
[Adaptable Service Based Software system]
Fundamental Research in M&S (Cont’d)
Background: DEVS M&S Framework
Discrete Event Systems Specification (DEVS)• Based on mathematical formalism using
system theoretic principles• Separation of Model, Simulator and
Experimental Frame• Atomic and Coupled types• Hierarchical modular composition
Level Name System Specification at this level
4 Coupled Systems
System built from component systems with coupling recipe.
3 I/O System Structure
System with state and state transitions to generate the behavior.
2 I/O Function
Collection of input/output pairs constituting the allowed behavior partitioned according to initial state of the system. The collection of I/O functions is infinite in principle because typically, there are numerous states to start from and the inputs can be extended indefinitely.
1 I/O Behavior
Collection of input/output pairs constituting the allowed behavior of the system from an external Black Box view.
0 I/O Frame Input and output variables and ports together with allowed values.
Source System
Simulator
Model
Experimental Frame
SimulationRelation
ModelingRelation
message
DEVS/SOA Federation Support Infrastructure
LiveTest
Player
DEVS Observer
Agent
Service Discovery: UDDI
Sevice Description: WSDL
Packaging:XML
Messaging:SOAP
Communication: HTTP
SOA
Mission Thread
DEVS Modeling and Simulation Infrastructure supports simultaneous testing at multiple levels
Syntactic LevelTests
Semantic Level Tests
Pragmatic LevelTests
network probes return statistics and alarms to DEVS transducers/acceptors
Mission Thread Test Agents Control and Observe collaborations
Semantic Level agents activate probes at Syntactic Level
DEVS acceptors alert higher layer agents of network conditions that invalidate test results
Pragmatic Level agents inform Semantic Level agents of the objectives for health monitoring
Semantic Level agents observe message exchanges between collaboration participants
Middleware (SOAP, RMI etc)-
Net-centric infrastructure
DEVS Simulator Services
DEVS Modeling Language (DEVML)
DEVS Simulation Concept•Specifies the abstract simulation engine that correctly simulates DEVS atomic and
coupled models
•Gives rise to a general protocol that has specific mechanisms for:
•declaring who takes part in the simulation:o format for referencing federates (participants)
•declaring how federates exchange information: oformat for their message exchange patterns
•executing an iterative cycle that•controls how time advances:
oupdating the clock based on next event times•determines when federates exchange messages:
othe point in the cycle when all interchange takes place•determines when federates do internal state updating
othe point in the cycle when next event times are collected
Note:If the federates are DEVS compliant then the simulation is provably
correct in the sense that the DEVS closure under coupling theorem guarantees a well-defined resulting structure and behavior.
DEVSSimulator
DEVSModel
DEVS Protocol
Concept of DEVS Standard
DEVS
CoreSimulatorInterface
Single
processor
Distributed
Simulator
Real- Time
Simulator
C++
Non
DEVS
DEVS
Model
Interface
Java
Other
Representation
DEVSSimulationProtocol
Virtual- Time
Simulator
DEVSML
Integrated Development and Testing Methodology
Define Requirements
Define Requirements
InterpretStructural Aspects
InterpretStructural Aspects
Capture Requirements
Capture Requirements
GenerateAtomic
DEVS Models
GenerateAtomic
DEVS Models
Generate System Entity
Structure
Generate System Entity
Structure
Prune Entity
Structure (PES)
Prune Entity
Structure (PES)
Transform PES to hierarchicalDEVS Models
Transform PES to hierarchicalDEVS Models
Create Test Models
Create Test Models
Insert Models into Test Platform
Insert Models into Test Platform
SimulateSimulate
InterpretBehavioral
Aspects
InterpretBehavioral
Aspects
ImplementSystem
ImplementSystem
Simulation-Based
Testing
Simulation-Based
Testing
DEVS/SOA Infrastructure: Supports Deployment and Execution of DEVS Models on the Web
WEBSERVICECLIENT
Middleware (SOAP, RMI etc)Net-centric infrastructure
DEVS Simulator Services
DEVS Modeling Language (DEVML)
DEVSJAVA
DEVSAgent
( Virtual User)
DEVSAgent
(Observer)
WEBSERVICECLIENT
Run Example
• Service Oriented Architecture (SOA) consists of various W3C standards
• Client server framework
• XML Message encapsulated in SOAP wrapper
• Machine-to-machine interoperability over the network based on WSDL interface descriptions
Search
find_xxxPost
save_xxx
Content/Service
Catalogs/Registries
Content/Service
Consumer
Content/Service
Provider
ServiceSOAP
XML
Schema
WSDL
Client
Access (& Use)
(Bind)
XML
Payload
Simple Object Application Protocol
Verification/Validation relative to service
Testing for Organization and Ontology quality
Assessment of content for pragmatic, semantic, syntactic correctness
Measurement of timeliness of information exchange
Content discovery accuracy and effectiveness
Requirements for Testing and Data Collection
DEVS/SOA Infrastructure for GIG Mission Thread Testing
1. MAJ Smith tasks Intell to reconnoiter objective area and provide threat estimate
2. Posts taskings using Discovery and Storage
6. MAJ Smith pulls estimate from Storage
3. Intell Cell initiates high priority collection against objective, and collectors post raw output
4. Intell posts products via Discovery and Storage
NCES GIG/SOA
DEVS/SOA Infrastructure for GIG Mission Thread Testing
1. MAJ Smith tasks Intell to reconnoiter objective area and provide threat estimate
2. Posts taskings using Discovery and Storage
5. Intell Cell issues alert via messaging 6. MAJ Smith pulls estimate from Storage
3. Intell Cell initiates high priority collection against objective, and collectors post raw output
4. Intell posts products via Discovery and Storage
Observing Agent for Major Smith
Observing Agent for Intell Cell
NCES GIG/SOA
• Test agents are DEVS models and Experimental Frames
• They are deployed to observe selected participant via their service invokations
notes time of posting
Observing Agent alerts other Agent
Computes Time for Task,Measure Performance
sends time to other Agent
Negotiation Modeling Approach
Domain-dependentstructure
Domain-independentbehavior
FD-DEVS
SES
~ phases~ message types
message specializations
FD-DEVSMarket Place
ReceivemessageInterpret message
Sendmessage
Language of EncounterClassification of the Negotiation’s Primitives
Abort Initiators Reactors Completers informative
Terminate ContractQuery Offer Reject Busy
NotMet CapabilityQuery
CounterOffer Accept LinkEstablished
ItemRequest Decline BestProvidor
CapabilityStatement ProvidorsChosen
DomainName
Item
ItemCheckResult
• Negotiation Scenario 1Language of Encounter Structure
devsworld.org acims.arizona.edu Rtsync.com
Books and Web Links
Backup