“So What-Who Cares?” A Three Level Approach to Creating Wining High Stakes Initiatives ·...
Transcript of “So What-Who Cares?” A Three Level Approach to Creating Wining High Stakes Initiatives ·...
UTEP Distinguished Speaker Seminar SeriesUTEP Distinguished Speaker Seminar Series
“So What-Who Cares?”A Three Level Approach to
Creating Wining High Stakes Initiatives
Prof. Michael SilevitchNortheastern University
November 3, 2015
“So What-Who Cares?”A Three Level Approach to
Creating Wining High Stakes Initiatives
Prof. Michael SilevitchNortheastern University
November 3, 2015
The Three Level Approach Starts With A Top Down VisionThe Three Level Approach Starts With A Top Down Vision
“So What-Who Cares”
• What is the central vision?
• Why is it compelling?
• What “Grand Challenges” must be addressed?
• Why can’t this be done now?
Integrate the Research Thrusts
Integrate the Research Thrusts
Testbeds to Validate Research
Testbeds to Validate Research
Real World Applications to Validate Vision
Real World Applications to Validate Vision
The Three Level Approach Enables A Logical Development of Complex InitiativesThe Three Level Approach Enables A Logical Development of Complex Initiatives
Level 3System Goals and Challenges
Level 3System Goals and Challenges
Level 1 Fundamental Science
Level 1 Fundamental Science
Level 2 Controlled Environment Testing
Level 2 Controlled Environment Testing
System Goal: What is the Vision?
“So What-Who Cares?”
System Goal: What is the Vision?
“So What-Who Cares?”
Identify the Barriers That
Must be Overcome
Identify the Barriers That
Must be Overcome
Identify Research Thrusts to
Address Barriers
Identify Research Thrusts to
Address Barriers
Gordon-CenSSIS: A Case StudyNSF Engineering Research Center $38MGordon-CenSSIS: A Case StudyNSF Engineering Research Center $38M
ERCs are Similar to a Multi-Division Small CompanyPrestigious Program ~2% Success RateEach NSF ERC Must: Define a Compelling Vision and Mission Address “Grand Challenges” Utilize a 3 Level Approach to Structure a
10 Year Transformative Research Program Implement a Long Range Strategic Plan Create Testbeds to Validate Approaches Motivate Students & Develop Educational Programs Foster an Innovation Ecosystem and
Industry Involvement
ERCs are Similar to a Multi-Division Small CompanyPrestigious Program ~2% Success RateEach NSF ERC Must: Define a Compelling Vision and Mission Address “Grand Challenges” Utilize a 3 Level Approach to Structure a
10 Year Transformative Research Program Implement a Long Range Strategic Plan Create Testbeds to Validate Approaches Motivate Students & Develop Educational Programs Foster an Innovation Ecosystem and
Industry Involvement
100nm - 100 m100nm - 100 m
Subcellular BiologySubcellular Biology Tissues & OrgansTissues & Organs
10 m - 10 cm10 m - 10 cm
10 cm - 1 km10 cm - 1 km
OpticsOptics UltrasoundUltrasound
SonarSonarRadarRadar
1 cm - 100 m1 cm - 100 m
UndergroundDiagnosis
UndergroundDiagnosis
UnderwaterExplorationUnderwaterExploration
The CenSSIS Vision: Subsurface DiagnosisDiverse Problems– Similar SolutionsThe CenSSIS Vision: Subsurface DiagnosisDiverse Problems– Similar Solutions
Integrate the Research Thrusts
Integrate the Research Thrusts
Testbeds to Validate Research
Testbeds to Validate Research
Attack Real World Subsurface Problems
Attack Real World Subsurface Problems
Overview of The Three Level Approach Applied to The CenSSIS ERC.Overview of The Three Level Approach Applied to The CenSSIS ERC.
Level 3System Goals and Challenges
Level 3System Goals and Challenges
Level 1 Fundamental Science
Level 1 Fundamental Science
Level 2 Controlled Environment Testing
Level 2 Controlled Environment Testing
Reveal Obscure Subsurface Regions: Diverse Problems-Similar Solutions
Reveal Obscure Subsurface Regions: Diverse Problems-Similar Solutions
Identify the Barriers That
Must be Overcome
Identify the Barriers That
Must be Overcome
Identify Research Thrusts to
Address Barriers
Identify Research Thrusts to
Address Barriers
The First Step: What Are The Fundamental Science Barriers?The First Step: What Are The Fundamental Science Barriers?
Barrier 1Barrier 1Inadequate understanding of the physics of subsurface sensing and imaging
Inadequate understanding of the physics of subsurface sensing and imaging
Barrier 2Barrier 2Unreliable inversion methods for inhomogeneous and cluttered subsurface media
Unreliable inversion methods for inhomogeneous and cluttered subsurface media
Barrier 3Barrier 3Lack of robust, physics-based recognition and sensor fusion techniques
Lack of robust, physics-based recognition and sensor fusion techniques
Barrier 4Lack of Real Time Computationally Efficient Modeling, Inversion and Image Processing Algorithms
Lack of Real Time Computationally Efficient Modeling, Inversion and Image Processing Algorithms
SubsurfaceSensing and
Modeling
Image & DataInformationManagement
Physics-BasedSignal Processing andImage Understanding
R1R1
R2R2R3R3
The Three Fundamental Science Thrusts Are Oriented Toward Overcoming Barriers 1-4The Three Fundamental Science Thrusts Are Oriented Toward Overcoming Barriers 1-4
FundamentalScienceFundamentalScienceL1L1
What Barriers Prevent The Developments of Testbeds to Validatee the Research?What Barriers Prevent The Developments of Testbeds to Validatee the Research?
Barrier 5Barrier 5 Lack of Meaningful Controlled Subsurface Testbeds to Effectively Test Research Results
Lack of Meaningful Controlled Subsurface Testbeds to Effectively Test Research Results
Lack of rapid processing and management of large image databasesLack of rapid processing and management of large image databases
Lack of optimal end to end sensor design methodsLack of optimal end to end sensor design methodsBarrier 6Barrier 6
Barrier 7Barrier 7
Four Validating TestBEDs Represent the Diverse Applications Inherent in the CenSSIS VisionFour Validating TestBEDs Represent the Diverse Applications Inherent in the CenSSIS Vision
BioBEDBioBED MedBEDMedBED
SoilBEDSoilBED SeaBEDSeaBED100nm - 100 m100nm - 100 m 10 m - 10 cm10 m - 10 cm
10 cm - 1 km10 cm - 1 km1 cm - 100 m1 cm - 100 m
Lack of real world case studies that leverage the results of the controlled testbeds
Lack of real world case studies that leverage the results of the controlled testbeds
Barrier 8Barrier 8
What Barriers Prevent The Ultimate Achievement of the Vision?What Barriers Prevent The Ultimate Achievement of the Vision?
Lack of a unified framework for subsurface sensing and imaging applicable to diverse problems
Lack of a unified framework for subsurface sensing and imaging applicable to diverse problems
Barrier 9Barrier 9
The System Level Applications Are Built Around Important Real World ProblemsThe System Level Applications Are Built Around Important Real World Problems
Biological-Medical Applications3D Imaging of Cellular Structure
4D Image Guided Therapy
3D Multi-Mode Breast Imaging
Environmental-Civil ApplicationsRemote Assessment of Benthic Habitats
4D Underground Assessment
Biological-Medical Applications3D Imaging of Cellular Structure
4D Image Guided Therapy
3D Multi-Mode Breast Imaging
Environmental-Civil ApplicationsRemote Assessment of Benthic Habitats
4D Underground Assessment
S1
S2
S3
S4
S5
The System Applications Require Collaboration With Domain ExpertsThe System Applications Require Collaboration With Domain Experts
ImportantBiological& Medical Problems
(S1, S2, S3)
ImportantBiological& Medical Problems
(S1, S2, S3)
Important Environmental
& Civil Problems(S4, S5)
Important Environmental
& Civil Problems(S4, S5)
Real Time Imaging and Sensor Fusion: Augmenting X-Ray Mammography Real Time Imaging and Sensor Fusion: Augmenting X-Ray Mammography
Important ProblemImportant Problem
Reliable Detection And Biopsy of
MalignantBreast Tumors
Reliable Detection And Biopsy of
MalignantBreast Tumors
TomosynthesisMGH PlatformTomosynthesisMGH PlatformImpedance
OpticalMicrowave
ImpedanceOptical
Microwave
ElastographyElastography
SensorFusion
S3
X-Ray Tomosynthesis Fused with CenSSIS Developed Techniques: Better ScreeningX-Ray Tomosynthesis Fused with CenSSIS Developed Techniques: Better Screening
TomosynthesisX-Ray
+ComputationalElastography
Electrical impedance Tomography
Diffuse Optical Tomography
S3
The S Level Requirements Spur Our Fundamental Science EffortsThe S Level Requirements Spur Our Fundamental Science Efforts
CenSSIS R3Tomo Speedup
From ~ 3 Hours- Single PCTo ~2 Minutes – 64 PC
Project to ~ 10 secondsWith GPU Hardware
CenSSIS R1-R2Fusion of Tomo
Model + Prior Info With DOT
Leads to Better Localization
0
180
360
540
720
900
1080
1260
1440
64 32 16 8 4Number of processors
Tim
e (s
ec)
PhantomPatient
S3
Clinical Data from EIT Holds Promise of a Non-Invasive Signature of MalignancyClinical Data from EIT Holds Promise of a Non-Invasive Signature of Malignancy
S3
Satellite Or AirbourneHyperspectral Remote Sensing
Coral Reef Health AssessmentVia Multi-Sensor FusionCoral Reef Health AssessmentVia Multi-Sensor Fusion
Benthic Habitat quantitative imaging needs multimodal/multiplatform capability SeaBED
AUV
Remote Sensing isLimited to 20 meters
S4
New Hyperspectral Remote Sensing Algorithms Enables Coastal Mapping at a New LevelNew Hyperspectral Remote Sensing Algorithms Enables Coastal Mapping at a New Level
Quantitative AssessmentUnmixing AlgorithmsChange DetectionModeling and Inversion
S4
The SeaBED AUV Enables Quantitative Imaging of Deep Coral Structures (> 200 Meters) The SeaBED AUV Enables Quantitative Imaging of Deep Coral Structures (> 200 Meters)
Mosaicing and Registration AlgorithmsUsed in Multiple Applications
CenSSIS Students Built the AUV
Reef Health Classification of Bottom Composition
Using AUV Imagery
Structural Complexity Derived from AUV Pencil Sonar Data
S4
SubsurfaceSensing and
Modeling
Validating Testbeds
Environmental-CivilApplications
Bio-MedicalApplications
Image & DataInformationManagement
Physics-BasedSignal Processing andImage Understanding
R1R1
R2R2R3R3
The CenSSIS Program Is Structured Around the Implementation of the Three Level ApproachThe CenSSIS Program Is Structured Around the Implementation of the Three Level Approach
FundamentalScienceFundamentalScience
EnablingTechnologiesEnablingTechnologies
System Goals and Challenges
System Goals and Challenges
L1L1
L2L2
L3L3
The CenSSIS 3 Level Strategy Enables the Solution of Diverse Relevant ProblemsThe CenSSIS 3 Level Strategy Enables the Solution of Diverse Relevant Problems
L1L1
L2L2
L3L3DOD-DOE-DHS
ApplicationsBiological-Medical
ApplicationsSuicide Bomb
DetectionBreast
Imaging
FunctionalImaging
Radiationtherapy
Littoral Zone Monitoring
Retinalsurgery
MineDetection
TunnelDetection
ATR
Computational ModelingInverse Techniques
Image Processing andUnderstanding
etc
MosaicingMultiview Tomography
Multispectral DiscriminationSensor Fusion
etc
Diverse Problems
SimilarSolutions
How did we make it happen?How did we make it happen?
• Aspired to win an ERC (1994)• Defined the initial vision• Northeastern was the sole partner• The 3 level diagram was not real• Submitted to NSF in 1996 • Moved to Full Proposal Stage: No Site Visit
• Recalibrated vision and mission (1997)• Brought in multiple university partners• Took the 3 level diagram seriously• Resubmitted in 1998• Was one of two proposals funded in 2000
out of 89 submissions
The Three Level Approach Starts With A Top Down Vision:The Three Level Approach Starts With A Top Down Vision:
“So What-Who Cares”
• What is the central vision?
• Why is it compelling?
• What “Grand Challenges” must be addressed?
• Why can’t this be done now?
Other Examples
ALERT: Awareness and Localization of Explosives-Related Threats
ALERT: Awareness and Localization of Explosives-Related Threats
A Department of Homeland Security
Center of Excellence
For Explosives Detection Mitigation & Response
~$40M in Funding To Date
A Department of Homeland Security
Center of Excellence
For Explosives Detection Mitigation & Response
~$40M in Funding To Date
InnovativeResearchResults
InnovativeResearchResults
ValidatingFacilities & Testbeds
ValidatingFacilities & Testbeds
FieldableProducts and
Systems
FieldableProducts and
Systems
A Top Down Approach Enables ALERT To Create A Coherent Strategy of OperationsA Top Down Approach Enables ALERT To Create A Coherent Strategy of Operations
Level 3System Goals
Level 3System Goals
Level 1 Fundamental Science Level
Level 1 Fundamental Science Level
Level 2 Testing Level
Level 2 Testing Level
Grand Challenges: ComprehensiveDefense Against
Explosive Threats
Grand Challenges: ComprehensiveDefense Against
Explosive Threats
IntegratedMultidisciplinary
Approach Required:No Silver Bullet
IntegratedMultidisciplinary
Approach Required:No Silver Bullet
Identify CrosscuttingResearch Thrusts
Identify CrosscuttingResearch Thrusts
C1: Characterization & Elimination of Illicit Explosives
C2: Actionable Remote Trace and Vapor Chemical Detection
C3: Ultra-Reliable Screening
C4: Effective > 50 meter Stand-Off Discovery and Assessment
C5: Seamless Transition of Research to the Field
ALERT is Guided by Grand Challenges Relevant to the Homeland Security Enterprise
The ALERT Fundamental Science Program: Requires Multidisciplinary Integration
R4 : Video Analytics & Signature Analysis
R2 : Trace & Vapor Sensors
R3 : Bulk Sensors & Sensor Systems
R1 : Characterization & Elimination of Illicit Explosives
ALERT Testbeds, Tools & Facilities Enable Meaningful Transition & E2E Initiatives
Multisensor Explosives Trace Detection
Video Analytics Based Anomaly Detection
Multisensor Whole Body Imaging
Multimode Standoff Detection
CT Luggage Data Sets to Test Algorithms
Level 1:Fundamental Research
Level 3:Grand Challenges
Level 2:TransitionAnd E2E
Explosives Database and Performance
TestingVideo Analytics‐Based Anomaly
DetectionMultisensorWhole Body Imaging
Stakeholders:DHS Other Gov’tAgenciesNat’l LabsIndustryK‐14 Students & TeachersUniversity Students & FacultyDomestic & International
Requirements for safety and securityin various sectors
Video Analytics and Signature Analysis (R4)
C5: Seamless Transition of
Research to the Field
ALERT deliverables include strategic studies reports, testbeds, datasets, simulations, models, courses, trained students
The ALERT Structure is Directly Related to the Trade‐offs Between Systems and Science
Multimode StandoffDetection
CT Luggage Data Sets to Test Algorithms
Bulk Sensors and Sensor Systems
(R3)Trace and Vapor
Sensors (R2)Characterization of Illicit Explosives
(R1)
C1: Characterization & Elimination of Illicit Explosives
C3: Ultra‐ReliableScreening
C4: Effective Stand‐Off Threat Discovery and Assessment
C2: Actionable Remote Trace and Vapor Chemical
Detection
Education programs are integrated seamlessly with each level of the ALERT
research program
Level 1:Fundamental Research
Level 3:Grand Challenges
Level 2:TransitionAnd E2E
Explosives Database and Performance
TestingVideo Analytics‐Based Anomaly
DetectionMultisensorWhole Body Imaging
Stakeholders:DHS Other Gov’tAgenciesNat’l LabsIndustryK‐14 Students & TeachersUniversity Students & FacultyDomestic & International
Requirements for safety and securityin various sectors
Video Analytics and Signature Analysis (R4)
C5: Seamless Transition of
Research to the Field
ALERT deliverables include strategic studies reports, testbeds, datasets, simulations, models, courses, trained students
The ALERT Structure is Directly Related to the Trade‐offs Between Systems and Science
Multimode StandoffDetection
CT Luggage Data Sets to Test Algorithms
Bulk Sensors and Sensor Systems
(R3)Trace and Vapor
Sensors (R2)Characterization of Illicit Explosives
(R1)
C1: Characterization & Elimination of Illicit Explosives
C3: Ultra‐ReliableScreening
C4: Effective Stand‐Off Threat Discovery and Assessment
C2: Actionable Remote Trace and Vapor Chemical
Detection
Education programs are integrated seamlessly with each level of the ALERT
research program
ALERT is Creating a Realistic Video Analytics Testbed to Enable Successful Transition at Cleveland Airport
Informs DHS Planning and Identifies Transition Gaps
Topics have Included: Segmenting Volumetric CT Data Advanced Imaging Technology (AIT) Explosives Detection Using CT
Image Reconstruction Automated Threat Recognition
Outcomes: A Collaborative Community Identification of Technology Gaps Strategic Studies Reports Comprehensive Data Sets to Test
New Approaches Future Directions for DHS Research
ALERT Algorithm Development for Security Applications (ADSA): Builds a Strong Third Party Community
ALERT Trained Students: The Next Generation of Professionals Who Will Help Safeguard Our Nation
Our Students Are Our Most Important Product
An Innovation Ecosystem: A Key Component of Many Major Initiaitives
The ALERT Transition Team
www.neu.edu/protect
The project described was supported by Award Number P42ES017198 from the National Institute Of Environmental Health Sciences.The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute Of Environmental Health Sciences or the National Institutes of Health.
Puerto Rico Testsite for Exploring Contamination
Threats (PROTECT) NIEHS SRP P42 Research Program
Co-Directors:Akram N. AlshawabkehJose F. Cordero
36
www.neu.edu/protect
PROTECT Center
• Supported by the Superfund Research Program (SRP) of NIEHS of NIH
• Started in April 2010 – $9.9 M from 2010 to 2014– $13.5 M from 2014 to 2019
• Multi-university involvement• Holistic source to outcome approach• Diverse expertise
– engineers, biochemists, electrochemists, toxicologists, epidemiologist, biostatisticians, pediatricians, agronomist, hydrogeologists, and social scientists.
37
www.neu.edu/protect
Babies born before 37 completed weeks of gestation are considered preterm.
Puerto Rico has the highest rate (17.7%) of any U.S. jurisdiction
Below only Malawi (18.1%) globally.
Preterm Births, United States and Puerto
RicoMarch of Dimes 2013 Premature Birth Report Card
www.neu.edu/protect
Preterm Birth is A Major Cause of Infant Mortality: Is it Linked to Contaminated Groundwater Exposure?
• Preterm birth: The #1 cause of neonatal mortality
• A public health concern that costs the US ~$26 billion a year
• In 2005, 12.5 percent of births in the United States were preterm, a 30 percent increase over 1981
• The known risk factors for prematurity cannot explain the data
www.neu.edu/protect
Contamination in Puerto Rico
• 200+ Hazardous Waste Sites• 16 Sites listed on the National Priority
List (NPL)• Many sites include unlined landfills
above aquifer in karst geologic formations
• Aquifer is primarily limestone with highly permeable karst aquifers from which most of the wells draw water
www.neu.edu/protect
PROTECT 3-Level Approach
41
The Three Level Approach Can Also be Used For Single investigator Projects
The Three Level Approach Can Also be Used For Single investigator Projects
Diverse Problems ‐ Similar Solutions
DoE CAREER proposal: Petrophysical characterization and dynamic imaging of flow transport using multi‐physical wave fields and multi‐scale sensing models
Level C: SystemChallenges and Real World Tests
Level C: SystemChallenges and Real World Tests
Level T: Models Testbeds & Tools
Level T: Models Testbeds & Tools
Level R:Fundamental Research
Level R:Fundamental Research
R3: Coupled EM‐AC/S imaging at reservoir‐scale (tens of meters): Dynamic consensus‐based imaging
using ADMM and compressive sensing
R1: Modeling of wave
propagation physics: 2D/3D‐EM, AC/S, TA
Barriers:B1: There is a need to develop a new science and technology base
that will enable a better understanding of the
interaction of disturbances in physical‐fields (Electromagnetic ,Acoustic/Seismic and Thermoacoustic) with
fluid‐filled porous media at different spatial and
temporal scales. B2: There is no thorough understanding of the characterization of natural variability
of geologic media at multiple spatial scales. B3: There is a need for refinement of existing geophysical methods
that provides enhanced imaging capabilities by
fusing data from multiple sensors at
multiple scales, in quasi‐real‐time (in‐situ), and
with limited data availability.
C1:Petrophysical Assessment of Hydrocarbon Reservoirs
Real World tests and datasets developed through partnerships with industrial manufacturers, customers, government agencies , and policy stakeholders
T1‐ Scaled Multi‐Physical‐Field Testbed:
• Controlled dynamic imaging• Core characterization
MULTI‐SCALE, MULTI‐PHYSICAL‐FIELD DESIGN & SIMULATION ENVIRONMENT
C2‐ Detection and Monitoring of Anthropogenic CO2 Storage
Underground
R2: Petrophysical characterization at pore‐scale (millimeters): wettability,
morphology, fluid distribution and saturation, salinity and
clay presence
The 3 Level Approach: A means To Enhance The Winning Of High Stakes EndeavorsThe 3 Level Approach: A means To Enhance The Winning Of High Stakes Endeavors
Try it for a few “Meaningful” Problems
Biomedical Advances Resilient Infrastructure Big Data to Knowledge Smart Power Grids and Energy Sources Inadequate K-12 Education in Science & Math
Compete for New Federal Funding Opportunities
A Good Way To Structure Your Response
Try it for a few “Meaningful” Problems
Biomedical Advances Resilient Infrastructure Big Data to Knowledge Smart Power Grids and Energy Sources Inadequate K-12 Education in Science & Math
Compete for New Federal Funding Opportunities
A Good Way To Structure Your Response
Two Last ThoughtsTwo Last Thoughts
The name of a Center is VERY important It sends a “message” to the reviewers.
“CenSSIS”“ALERT”“PROTECT”“SENTRY” (A new ERC pre-proposal)
There MUST be joy and a sense of adventure in creating high stakes initiatives.
The name of a Center is VERY important It sends a “message” to the reviewers.
“CenSSIS”“ALERT”“PROTECT”“SENTRY” (A new ERC pre-proposal)
There MUST be joy and a sense of adventure in creating high stakes initiatives.