Welcome to the DDDAS2020 Conference Conference Co-Chairs: Dr. - - PowerPoint PPT Presentation

DDDAS – Dynamic Data Driven Applications Systems

Past, Present, Future

DDDAS2020 & Beyond

Frederica Darema, Ph.D., LF/IEEE (retired) Director – Senior Executive (SES) DDDAS2020 Conference October 2-4, 2020 www.1dddas.org

1

Conference Co-Chairs: Dr. Frederica Darema and Dr. Erik Blasch Program Co-Chairs: Dr. Alex Aved, Prof Sai Ravela, Dr. Murali Rangaswamy Program Committee:

• Robert Bohn (NIST)
• Newton Campbell (NASA and SAIC)
• Nurcin Celik (U of Miami)
• Ewa Deelman (USC/ISI)
• Salim Hariri (U of Arizona)
• Thomas Henderson (U of Utah)
• Artem Korobenko (U of Calgary)
• Fotis Kopsaftopoulos (RPI)
• Richard Linares (MIT)
• Dimitri Metaxas (Rutgers Univ)
• Jose Moreira (IBM)
• Chiwoo Park (FSU)
• Sonia Sachs (DOE)
• Ludmilla Werbos (U of Memphis)
• Themistoklis Sapsis (MIT)
• Amit Surana (UTRC)

Welcome to the DDDAS2020 Conference

Agenda: Keynotes, Plenaries, Posters, Panels

2

DDDAS Initiative Launched (March 2000)

DEFINTION of DDDAS (2000/NSF Workshop)

Instrumentation Observation/Actuation

Dynamic

Feedback & Control

Loop

DDDAS: ability to dynamically incorporate additional data into an executing application, and in reverse, ability of an application to dynamically steer the measurement process

• > Synergistic, Multidisciplinary Research

(2000/NSF Workshop) Dynamic Integration of Computation & Measurements/Data Unification of Computing Platforms & Sensors/Instruments (from the High-End to the Real-Time, to the PDA)

DDDAS – > architecting & adaptive management of sensor systems

DDDAS

(Dynamic Data Driven Applications Systems)

InfoSymbiotic Systems

Capabilities:

more accurate understanding/prediction systems characteristics/behaviors

• speeding-up simulation/modeling, by replacing computation with

instrumentation data in specific parts of the phase-space of the model

• improve accuracy of the model by augmenting the model with actual data

to improve analysis/prediction capabilities of application models

• dynamically manage/schedule/architect heterogeneous resources
• networks of heterogeneous sensors or controllers
• detect and mitigate sensor failures
• enable decision-support capabilities with simulation/modeling accuracy

Challenges: Application Modeling/Simulations Methods Algorithmic Stability Measurement/Instrumentation Methods Computing Systems Software Support

3

DDDAS Development

2010 2011 2014

QVO VADIMUS

2005

DDDAS NSF Program

2006

NSF Workshop DDDAS AFOSR-NSF Workshop AFOSR Program AFOSR-NSF Program Large-Scale-Big-Data Large-Scale-Big-Computing

1980 1984-86 1996 1998 2000 2003

Idea spawned NSF Workshop DDDAS IBM; UTC; ASME DDDAS NSF/ITR DARPA

Performance Engineering

NSF/NGS Program Gedanken Laboratory DDDAS

term coined (symbiotic systems)

1999 2016

4 InfoSymbiotic

Example Highlights of DDDAS Impact

• 2000 - Kelvin Droegemeier – Adverse Weather /Tornadic activity – LEAD project: Users INTERACTING with Weather

Experimental Dynamic

Observations Users

ADaM

ADAS

Tools

NWS National Static Observations & Grids Mesoscale

Weather Local Observations

Local Physical Resources Remote Physical (Grid) Resources Virtual/Digital Resources and Services

Interaction Level II: Tools and People Driving Observing Systems – Dynamic Adaptation

“Sensor Networks & Computer Networks”

Slide courtesy Droegemeier

Tornado

March 2000 Fort Worth Tornadic Storm Local TV Station Radar

(Slide – Courtesy K. K. Droegemeier)

NEXRAD CASA 2010 - Tinsley Oden – showed predicting on-set in materials damage before visible 2011 – Nurcin Celik - Powergrids Real-time Decision Support - Renewable Resources – multiple consumer classes&priorities 2012 – Karen Willcox – Real-Time Decision Support for aerial platforms - Structural Health Monitoring and Mission Planning Potential examples (DDDAS-based solutions):

• Future aircraft designs (Dutch V-shaped – DelftU -KLM)
• Tree-tomography

(WashDC Sculpture Garden)

5

Examples of Areas of DDDAS Impact

from the nano-scale, to the tera-scale, to the extraterra-scale

AFOSR-NSF 2010 Report (www.1dddas.org)

• Materials – Fundaments & Design
• Structural Health Monitoring
• Advanced Manufacturing
• Smart Civil Infrastructures
• Transportation
• Power-grids
• Water Distribution
• Smart Cities
• Ecological Systems
• Smart Agriculture
• Atmospheric Weather
• Adverse events
• Hurricanes
• Tornadoes
• Earthquakes
• Environmental Disasters
• Wildfires
• Oil Spills
• Earthquakes
• Space Weather
• Land, Air, Space
• Emergency Response
• Resource Planning
• Supply-Chain Logistics
• Model-based Real-time Decision Support
• > Autonomic Systems

DDDAS/InfoSymbiotics drives:

• Foundational methods
• Filtering, Estimation,
• Machine Learning
• Uncertainty Quantification
• Applications approaches
• systems-of-systems
• representation models
• network control
• sensor management

DDDAS has influenced extensions:

• Data Assimilation
• Digital Twin
• GANs

Recent/emerging ML algorithms

apply and/or adopt the essence the DDDAS paradigm

• Informative Sensing, Estimation, Planning
• Targeted Observation, Active Learning
• Reinforcement Learning RelevanceFeedback
• Stochastic Modeling, Feature Selection
• Recommender Systems, etc

Other initiatives, such as:

Cyber-physical Systems (NSF 2006) can benefit from the more comprehensive approaches of the DDDAS paradigm

6

QUO VADIMUS

DDDAS/InfoSymbiotics has shown:

• comprehensive, principle-based models rendered more efficient and accurate
• understanding, prediction, and real-time decision support with the accuracy of full-scale models
• dynamic and adaptive coordination of heterogeneous resources abilities, and ensuring fault tolerance

DDDAS/InfoSymbiotics - timely now more than ever:

• increased emphasis in complex systems multi-scale/multi-modal modeling/algorithmic methods
• ubiquitous sensing, networks of heterogeneous collections of sensors/controllers
• large-scale-big-computing – large-scale-big-data;
• increased computation and communication capabilities
• enable and exploit these capabilities

Some New Opportunities Areas:

Test&Evaluation:

• embedded sensors&actuators (physical and software); additive manufacturing
• DDDAS-based adaptivity to improve performance, system evolution adaptivity – interoperability - maintenance;
• No longer limited to the design (“breadboarding”) cycle – T&E become a lifecycle process

5G&Beyond:

• DDDAS-based methods for adaptive coordination of heterogeneous resources; optimization of performance, energy, QoS

Data alone is not enough Data is not the 4th paradigm… - Data is the primordial paradigm Data Analytics is not enough - We need Systems Analytics ML alone is not enough

• > …. Models Data ….

Moving into the future there is a confluence of needs and technological advances:

Increasingly we deal with systems-of-systems & systems/environments that are complex | heterogeneous | multimodal | multiscale | dynamic Need to understand characteristics/behaviors: design – operation – evolution – interoperability – maintenance -- lifecycle Ad-hoc methods are not enough – need modeling not only for design but entire file-cycle

7