brief introduction to data-driven application simulations craig c. douglas university of kentucky...

14
Brief Introduction to Data- Brief Introduction to Data- Driven Application Simulations Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University http://www.mgnet.org/~douglas with help from CS 612 (Fall 2006, UKy) Start with http://www.dddas.org Supported in part by the National Science Foundation: ITR-DDDAS and DDDAS programs

Upload: eustacia-marshall

Post on 01-Jan-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

Brief Introduction to Data-Brief Introduction to Data-Driven Application SimulationsDriven Application Simulations

Craig C. DouglasUniversity of Kentucky and Yale

Universityhttp://www.mgnet.org/~douglas

with help from CS 612 (Fall 2006, UKy)

Start with http://www.dddas.org

Supported in part by the National Science Foundation:ITR-DDDAS and DDDAS programs

Page 2: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

What is DDDAS?What is DDDAS?

It is a paradigm whereby application It is a paradigm whereby application (or simulations) and measurements (or simulations) and measurements become a symbiotic feedback control become a symbiotic feedback control system.system.

Dynamically incorporates additional Dynamically incorporates additional data into an executing application and data into an executing application and lets an application dynamically steer lets an application dynamically steer the measurement process.the measurement process.

Capabilities promise more accurate Capabilities promise more accurate analysis and prediction, more precise analysis and prediction, more precise controls, and more reliable outcomes.controls, and more reliable outcomes.

Page 3: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

What Benefits from DDDAS?What Benefits from DDDAS?

Applications in business, engineering Applications in business, engineering and scientific processes, analysis, and and scientific processes, analysis, and design.design.

Areas in manufacturing process Areas in manufacturing process controls, resource management, weather controls, resource management, weather and climate prediction, traffic and climate prediction, traffic management, systems engineering, civil management, systems engineering, civil engineering, geological exploration, engineering, geological exploration, social and behavioral modeling, social and behavioral modeling, cognitive measurement, and bio-sensing.cognitive measurement, and bio-sensing.

Page 4: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

How Is DDDAS Done?How Is DDDAS Done?

DDDAS creates a rich set of new challenges DDDAS creates a rich set of new challenges for applications, algorithms, systems’ for applications, algorithms, systems’ software, and measurement methods.software, and measurement methods.

DDDAS research typically requiresDDDAS research typically requires– strong, systematic collaborations between strong, systematic collaborations between applications domain researchers and mathematics, applications domain researchers and mathematics, statistics, and computer sciences researchersstatistics, and computer sciences researchers

– researchers involved in the design and researchers involved in the design and implementation of measurement methods and implementation of measurement methods and instruments.instruments.

Most DDDAS projects involve Most DDDAS projects involve multidisciplinary teams of researchers.multidisciplinary teams of researchers.

Page 5: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

A Few SamplesA Few Samples

Weather, climate, and ocean modelingWeather, climate, and ocean modeling FFllooding, typhoons or hurricanes, thunderstorms, tornadosooding, typhoons or hurricanes, thunderstorms, tornados FFires: wildland, forest, buildingsires: wildland, forest, buildings CContaminants: transport (water/air), identificationontaminants: transport (water/air), identification OOil/gas explorationil/gas exploration Traffic flow: lights and vehicular behaviorTraffic flow: lights and vehicular behavior TTerrorist attack follow upserrorist attack follow ups Passenger screening at transportation centers Passenger screening at transportation centers

(air/train/boat/bus)(air/train/boat/bus) MManufacturing and full plant managementanufacturing and full plant management Brain-machine interfacesBrain-machine interfaces MMedical treatment (cancer treatment)edical treatment (cancer treatment) HHigh energy physics device controligh energy physics device control Mining safety would be niceMining safety would be nice

Page 6: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

A Down to Earth Example:A Down to Earth Example:The Empty House DDDASThe Empty House DDDAS

Sensors along driveway and near houseSensors along driveway and near house Distinguish between animals, people, and Distinguish between animals, people, and vehiclesvehicles

Recognize regular delivery peopleRecognize regular delivery people– Possibly turn on lightsPossibly turn on lights– Greet them with prerecorded voices and messagesGreet them with prerecorded voices and messages– Direct receivables to box by doorDirect receivables to box by door– Move contents of box indoors regularlyMove contents of box indoors regularly

Tell sales critters and religious nuts to go Tell sales critters and religious nuts to go awayaway

Feed cats, put out garbage on correct days, Feed cats, put out garbage on correct days, etc.etc.

Give a complete lived in appearance to houseGive a complete lived in appearance to house Two way secure communication with residentsTwo way secure communication with residents

Page 7: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

SensorsSensors

Motion activated cameras that cover the Motion activated cameras that cover the premisespremises– When motion detected, try object recognitionWhen motion detected, try object recognition

Intercom for recording people at the Intercom for recording people at the door outputting sounddoor outputting sound

Page 8: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

Object DetectionObject Detection

The object detection system is pre-The object detection system is pre-trained to distinguish between trained to distinguish between vehicles, people, and animals while vehicles, people, and animals while ignoring other objects (such as leaves, ignoring other objects (such as leaves, swaying trees, etc.) using a swaying trees, etc.) using a classification system.classification system.

The system also trains to recognize The system also trains to recognize regular delivery people and common regular delivery people and common visitors.visitors.

Page 9: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

System BehaviorSystem Behavior

If the system detects a vehicle, it If the system detects a vehicle, it checks to see if the vehicle is a common checks to see if the vehicle is a common arrival, such as a delivery truck.arrival, such as a delivery truck.

Cameras track the occupant as he Cameras track the occupant as he approaches the door.approaches the door.

If he does not approach the door, the If he does not approach the door, the behavior is logged as suspicious and the behavior is logged as suspicious and the owner is contacted.owner is contacted.

If the occupant is recognized at the If the occupant is recognized at the door, then he/she is greeted with an door, then he/she is greeted with an appropriate message.appropriate message.

Page 10: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

System Behavior (continued)System Behavior (continued)

If it is a delivery-person then the If it is a delivery-person then the system instruction the him/her to leave system instruction the him/her to leave the package in a box by the door. the package in a box by the door.

The system then brings the contents of The system then brings the contents of the box into the house.the box into the house.

Page 11: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

System Behavior (continued)System Behavior (continued)

If the person is not recognized, he/she is If the person is not recognized, he/she is asked to leave a message, which is later asked to leave a message, which is later communicated to the owner. The owner can communicated to the owner. The owner can then establish a secure remote connection then establish a secure remote connection with the system and flag the person as a with the system and flag the person as a desirable or undesirable. The system will desirable or undesirable. The system will attempt to memorize the newcomer’s face attempt to memorize the newcomer’s face for a time duration set by the owner.for a time duration set by the owner.

If the owner flags the person as If the owner flags the person as undesirable, such as a salesman, their undesirable, such as a salesman, their face is stored long term and the person is face is stored long term and the person is told to leave the next time they show up.told to leave the next time they show up.

Page 12: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

System Behavior (continued)System Behavior (continued)

Automated TasksAutomated Tasks– If the owner has pets, the system maybe If the owner has pets, the system maybe configured to operate a food dispenser to configured to operate a food dispenser to feed them.feed them.

– The system learns the manner in which the The system learns the manner in which the house is lit when it is lived in and tries house is lit when it is lived in and tries to replicate it in the owner’s absence.to replicate it in the owner’s absence.

Page 13: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

ArchitectureArchitecture

The system architecture is two tiered.The system architecture is two tiered.– The first tier is responsible for object The first tier is responsible for object detection and classification.detection and classification.

– The second tier receives information about a The second tier receives information about a detected object (recognized friendly, detected object (recognized friendly, recognized undesirable or unrecognized) and recognized undesirable or unrecognized) and chooses an appropriate response.chooses an appropriate response.

The two tiered approach reduces the The two tiered approach reduces the load on the second tier so that load on the second tier so that resources are available for decision resources are available for decision making and communication with the owner.making and communication with the owner.

Page 14: Brief Introduction to Data-Driven Application Simulations Craig C. Douglas University of Kentucky and Yale University douglas with

Some Possible ExtensionsSome Possible Extensions

MilitaryMilitary– Replace cats with our troopsReplace cats with our troops– Replace sales critters with hostile troopsReplace sales critters with hostile troops– Religious nuts still a threatReligious nuts still a threat

Far Out ExtensionsFar Out Extensions– K-12 School SecurityK-12 School Security

Difficult due to number of teachers and studentsDifficult due to number of teachers and students– HR DepartmentsHR Departments

Deal with new hires and employee problemsDeal with new hires and employee problems– Walk in IT Help DesksWalk in IT Help Desks

Routine problems versus cranksRoutine problems versus cranks– Emergency Room Walk InsEmergency Room Walk Ins