VASA: Visual Analytics for Simulation-based Action

Post on 29-Jun-2015

276 views 1 download

Tags:

description

Slides from our IEEE VAST 2014 talk at IEEE VIS on VASA, a visual analytics system for interactive computational steering of pipelines of asynchronous simulation models.

transcript

VASAInteractive Computational Steering of Large

Asynchronous Simulation Pipelines for Societal Infrastructure

SUNGAHN KO JIEQIONG ZHAO JING XIA SHEHZAD AFZAL XIAOYU WANGGREG ABRAM NIKLAS ELMQVIST LEN KNE DAVID VAN RIPER KELLY GAITHER

SHAUN KENNEDY WILLIAM TOLONE WILLIAM RIBARSKY DAVID S. EBERT

Society is under threat from many sources…

Power grid

Power plants

Motivating scenario slide:Picture of critical infrastructure

Highways

Supply chains

Simulation slide:Simulation is the answer

DISASTERstrikes…

When

HOW CAN WE PREPARE?

WHEN REAL-WORLD EXERCISES ARE COSTLY AND DANGEROUS

simulationI S T H E A N S W E R

IDEA: pipeline of asynchronous simulations

Weathersimulation

Powergrid model

Road networks

Criticalinfrastructure

System-of-systems:multiple heterogeneous systemscombined into a unified systemgreater than its individual parts

CHALLENGES

C1Monolithic simulation

C2Complex relations

C3Non-

standard data

C4Long exec. times

C5Certainty

+ fidelity

VASAVISUAL ANALYTICS FOR SIMULATION-BASED ACTION

Distributed component-based framework for computational steering of systems-of-systems simulations for societal

infrastructure based on a visual analytics approach

CASUAL EXPERTS:Deep expertise in domain

No expertise insimulation + data science

RESILIENCE + RESPONSE:Understand and trace events

Identify vulnerabilities“What if?” scenarios

COMPLEX SYSTEMS:Supply chain logistics

Public safetyCybersecurity

USERS TASKS DOMAIN

DESIGN GUIDELINES

• Avoids integration of a monolithic design with another • Provides a data exchange format (C1, C3)• Enables parallel execution of distributed models (C4)

G1Simulation services

• Provides approximated results for interactive response• Enables real-time response hiding long execution times

G2Simulation proxies

• Help to simplify configurations for non-experts• Provides a data exchange format (C1, C3)

G3Visual Relations

DESIGN GUIDELINES (2)

• Partial and interruptible computational steering

G4Computational steering

• Uncertainty visualization (C5)• Propagation of errors

G5Visual representations

• Main focus of VASA is mapsG6Spatiotemporal focus

VASA WORKBENCH

Interactive desktop tool fora distributed system

Visual analytics dashboard w/multiple coordinated views

Configure + steer + explore(simulation models)

Control distributedsimulations

using REST API

Simulation proxyprovides real-time response

VASA COMPONENT:WEATHER

MODEL PROXY

NOAA: National Oceanic andAthmospheric Administration

ADCIRC: AdvancedCirculation

Prepares eventdatasets from server

Historic data(Irene, Sandy, etc)

Visualizes hurricane coneover time using slider

Generates inputs todownstream components

VASA COMPONENT:CRITICAL INFRASTRUCTURE

MODEL PROXY

E.g. Electric grids, telecomnetworks, gas distribution

Simplified connectivitygraph of important structures

Vu environmentwith submodels

Example: show impact ofhurricanes on restaurants

VASA COMPONENT:ROUTING

MODEL PROXY

Inputs: barriers and closuresOutputs: new transport routes

Simulation engine:ArcGIS Server

Approximates disabledroutes and facilities

Maintains road networkfor critical infrastructure

Visualizes disabledroutes and facilities

VASA COMPONENT:SUPPLY CHAIN

MODEL PROXY

Discrete event simulation onchain in geolocated facilities

Our models: poultry firm + fast food = farm to restaurant

Accepts externalinputs (weather and roads)

Supply chains dependon business and goods

Supports road closures, powerless stores and flooding (polygons)

Food contaminationalso modeled and visualized

Optimizes distribution and even redistributes

products

EXAMPLE: U.S. HURRICANE SEASON

Hurricane Irene hits NorthCarolina on August 27, 2011

34-knot winds batter the coast; criticalinfrastructure proxy estimates impacted

restaurants

EXAMPLE: U.S. HURRICANE SEASON (2)

Complete power grid simulation is run;a shaded polygon shows actual

power outage

Supply chain simulation runshows that some routes are no

longer completing deliveries

CASE STUDIES + FEEDBACK

Regional FEMA•Unprecedented work•Visual investigation• Instant approximations

• “Whole Community”•Meets missions needs•Enable informed decisions

•Suggestion: real-time weather data

U.S. Coast Guard•Dire need with no current solution

•VASA could drastically change their operations

•Potential interface for emergency response

•Great potential

CONCLUSION

•VASA: Visual Analytics for Simulation-based Action•Systems-of-systems approach•Multiple heterogeneous systems into a unified system

•Case studies•Hurricane impact on societal critical infrastructures• Feedback by FEMA and U.S. Coast Guard

FUTURE WORK

Advanced simulation:Energy infrastructures, transportationnetworks, societal infrastructure

Visual representations:Configurations, proxies, intermediate,and final results from simulations

QUESTIONSWork supported by the U.S Department of Homeland Security’s

VACCINE Center 2009-ST-061-CI0001-06.

We thank our analysts and partners for feedback and advice during the project.

Iconography created by designers from the Noun Project.

Niklas Elmqvistelm@umd.edu

David Ebertebertd@purdue.edu