DEVS Today: Recent Advances in Discrete Event - Based Information Technology Bernard P. Zeigler Bernard P. Zeigler Professor, ECE Professor, ECE Arizona Center for Integrative Modeling and Arizona Center for Integrative Modeling and Simulation Simulation University of Arizona University of Arizona Tucson Tucson www.acims.arizona.edu www.acims.arizona.edu Keynote Talk to Ma Majestic
30
Embed
Bernard P. Zeigler Professor, ECE Arizona Center for Integrative Modeling and Simulation
DEVS Today: Recent Advances in Discrete Event - Based Information Technology. Bernard P. Zeigler Professor, ECE Arizona Center for Integrative Modeling and Simulation University of Arizona - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
DEVS Today:
Recent Advances inDiscrete Event -
Based Information Technology
Bernard P. ZeiglerBernard P. Zeigler
Professor, ECEProfessor, ECEArizona Center for Integrative Modeling and SimulationArizona Center for Integrative Modeling and SimulationUniversity of ArizonaUniversity of ArizonaTucson Tucson www.acims.arizona.eduwww.acims.arizona.edu
Coupled: composed of one or more atomic and/or coupled models
Atomic
Atomic
Atomic
Hierarchical construction
7
DEVS Theoretical Properties
• Closure Under Coupling• Universality for Discrete Event
Systems• Representation of Continuous
Systems– quantization integrator approximation– pulse representation of wave equations
• Simulator Correctness, Efficiency
8
Atomic Models
OrdinaryDifferentialEquationModels
Spiking NeuronModels
Coupled Models
Petri NetModels
Cellular Automata
n-Dim Cell Space
PartialDifferentialEquations
Self Organized Criticality
Models
Processing/Queuing/
Coordinating
ProcessingNetworks
Networks,Collaborations Physical
Space
DEVS Expressability
can becomponents in a coupled model
MultiAgent
Systems
Discrete Time/
StateChartModels
QuantizedIntegrator
Models
Spiking Neuron
Networks
Stochastic
Models
ReactiveAgent
Models
Fuzzy Logic
Models
9
Cell Space
Wind
Water
Ignite
Coupled model structure
N
E
NENW
W
SW
S SE
Potential neighbor cells to ignite by fire from center cell.
10
unburned unburned_wet
burning
burned_wetburning_wet
burned
Fire suppressant
Burning delay = 0
Ignition and
(fireline intensity > Threshold)
Fire suppressant
delay = 0
Fire suppressant and fire fighting rule
satisfied
Forest Cell State Transitions
Atomic model structure
Time advance
input
Make a transition
elapsed time
Time advance
input input
Make a transition Make a transition
elapsed time
Phase “unburned” If (FI > Threshold)
holdIn (“burning”,
else passivateIn( “Unburned)”
Compute new spread ( using
Rothermel’s eq)
Compute remaining distance to reach center of neighbor cell Compute time delays
Fireline IntensityFI
Phase “burning”
11
Experimentation
WindFlowModel
Fire Fighting
Model
Forest CellIgniter
Cell Space Display
Transducer
DisplayAverage Rate of Spread &
Direction
DisplayActive Cells
Vs.Total Cells
DisplayOther Stats
Cell Space
Wind
Water
Ignite
experimental frame
12
10°15°Wind
Wind Wind N
NFFL-Fuel-Model
5: Brush (2 ft)
NFFL-Fuel-Model 11:
Light logging slash
NFFL-Fuel-Model 7:
Southern rough
wind across valley floor experiments
13
water meets fire experiment
14
M&S Framework Implications for Current Practice
• Separate Models From Simulators • Separate Models From Experimental Frames• Use the DEVS Formalism for Developing Models,
Experimental Frames, and Simulators• Experimental Frames Support Defense Certification
Testing• Maintain Repositories of Reusable Models and Frames
15
Separate Models From Simulators
• Models are goal oriented abstractions of reality.
• Simulators are the computational engines that drive the models to obtain results.
In the M&S-Framework-based approach..
• Models and Simulators are treated as distinct entities with their own software representations.
• There are simulators for different kinds of models that can be selected according to the needs of the simulation,
• For example, a simulator might be chosen for its efficiency on a single host, or for its ability to execute the model on multiple hosts (distributed simulation)
Currently…Simulation software tends to encapsulate models and simulators in
tightly coupled packages.
16
Separate Models From Experimental Frames
• Experimental Frames are specifications of the experimentation to be done on a model
• Frames represent the objectives of the experimenter, tester, or analyst
In the M&S-Framework-based approach..
• Models and Experimental Frames are treated as distinct entities with their own software representations.
• Since the experimental frames appropriate to a model are distinctly identified, it is easier for potential users of a model to uncover the objectives and assumptions that went into its creation.
Currently…Simulation software tends to encapsulate models, simulators and
experimental frames into tightly coupled packages.
17
Use the DEVS Formalism for Developing Models, Experimental Frames, and Simulators
• The DEVS formalism enables users to develop models separately from experimental frames .
• Models and frames can then be coupled together and given to an appropriate simulator to execute.
In the M&S-Framework-based approach..
• The DEVS formalism Is employed for all simulation software development.
• DEVS simulators are employed to perform single host, distributed and heterogeneous real-time execution as needed.
• DEVS simulators exist that run over various middleware such as MPI,HLA, CORBA,P2P, and MOM.
Currently…Programming languages such as Fortran, C, C++ or Java are used to
develop software packages of strongly coupled models, frames and simulators.
18
Maintain Repositories of Reusable Models and Frames
• Models and Experimental Frames can be stored in organized repositories to support reuse under well specified conditions
In the M&S-Framework-based approach..
• Repositories of models and frames are created and maintained.
• Such repositories foster reuse of existing models and frames to serve as components for constructing new ones.
• When new models or frames are developed they are deposited in the repositories with appropriate information to enable their reuse with high confidence of success.
Currently…There are relatively few examples of storing previously developed
simulation infrastructure commodities in such a way that they can be easily adapted to developing interoperability test requirements
19
Managed Modeling in Lockheed’s “System of Systems” M&S Environment
• DEVS (Discrete Event Modeling Formalism) – Separates Model and Simulators
– Defines Couple Models and Atomic Models
– Modularized via Ports and Defined Events
• SES (System Entity Structure) – Provides a well defined structure for model reuse
• Architecture based on SES/DEVS supports component
model reuse evolved during last decade
20
Project
Model
Critical Mobile Target
Global Positioning System
III
Arsenal Ship
Coast Guard Deep Water
Space Operatio
ns Vehicle
Common Aero Vehicle
Joint Compos
ite Tracking Network
Integrated
System Center
Space Based Laser
Space Based
Discrimination
Missile Defense
(Theater / National)
RAD x x x x x x x
IR x x x x x x xMISMIS x x x x xLAS x x x x
Comm x x x x x xCC x x x
Earth x x x x xWC x x
Component Reusability in Lockheed’s DEVS M&S Environment
21
DEVS framework for knowledge based control of steel production
Sachem = large-scale real-time monitor/diagnose control system for blast furnace operation
Usinor -- world’s largest producer of steel products, Sachem saves it millions of euros annually
Problems for conventional control and AI:•Experts’ perception knowledge is implicit, concerns dynamic physical processes •Difficult to model the reasoning of a control process expert. •Lack of mathematical models for blast furnace dynamics
Solution:• time-based perception and discrete event processing for dealing with complex dynamical systems
22
quantization
signalevents
signalphenomena
processphenomena
Large Scale:•Conceptual model contains 25,000 objects for 33 goals, 27 tasks,etc.•Approximately 400,000 lines of code. •14 man-years: 6 knowledge engineers and 12 experts
One advantage of DEVS is compactness: 50,000 reduction in data volume
Effective analysis and control of the behavior of blast furnaces at high resolution
DEVS framework for knowledge based control of steel production (cont’d)
23
University of New Mexico Virtual Lab for Autonomous Agents
V-Lab-a virtual laboratory for autonomous agents-SLA-based learning controllers El-Osery, A.I.; Burge, J.; Jamshidi, M.; Saba, A.; Fathi, M.; Akbarzadeh-T, M.-R.; Systems, Man and Cybernetics, Part B, IEEE Transactions on , Volume: 32 Issue: 6 , Dec. 2002 Page(s): 791 -803
Physics Terrain Dynamic
SimEnv
Control Agents SimMan
Computer Network
Middleware (HLA,CORBA,JMS)
DEVS Simulator
IDEVS SimEnv
V-Lab developed on top of DEVSJAVA includes a simulation environment for robotic agents with physics, terrain and dynamics. It extends DEVS to provide a layer for specifying intelligent automation and soft computing algorithms (IDEVS).
24
nD
(n-1)D
X>0
X<0
D
ta(nD) = |D/x|
nD
D
ta(q) = ((n+1)D-q)/x
e
X>0
X<0
q
ta(q) = |q-nD/x|
(n+1)D
Mapping Differential Equation Models into DEVS Integrator Models
DEVSinstantaneous
function
DEVS Integrator
d s1/dt s1f1x
d s2/dt s2f2
d sn/dt snfn
sx
sx
sx
...
d s1 /dt s1f1x
d s2 /dt s2f2
d sn /dt snfn
sx
sx
sx
...
DEVSSDEVS
DEVS
F
F
F
25
)(tf
t
quantum
Number of crossings = Activity/quantum
Activity – a characteristic of continuous models
dttfdq
)(
Activity = |f(t1) – f(t0)|
26
DEVS Efficiency Advantage where Activity is Heterogeneous in Time and Space
Time Period
T
time stepsize
# time steps
=T/
tt
activityA
quantumq
# crossings=A/q
Potential Speed Up=
#time steps /# crossings
X
numberof
cells
diffusion
activity
27
Activity as unifying continuous and discrete paradigms
Heterogeneous activity in
time and space
Quantization allows DEVS to naturally focus computing resources on high activity regions
DEVS represents all decision making and continuous dynamic components in the
scene
28
Modeling and Simulation as a Bridging Discipline (3)
Continuous Systems• Analog• Control theory• Linear/Non Linear • ODE/PDEs