Transcript Fundamental Research Project Proposal

Collaborative Research: An Interactive Situational
Awareness Simulation – A View from the Clouds.
Fundamental Research Project Proposal
Sheldon Brown, UCSD, Site Director
Milton Halem, UMBC Director
Yelena Yesha, UMBC Site Director
Tom Conte, Georgia Tech Site Director
Summary.
We will create a virtual world simulation, distributed across a
heterogeneous compute infrastructure, to facilitate dynamic
analysis of complex situations. The virtual world will be built
from a variety of data sources; including high fidelity, well
correlated data sources to data mining of internet feeds and
many-user updates with commodity devices such as cell
phones and digital cameras.
Integrating heterogeneous computing resources to enable
applications that can provide robust representations of dynamic
situations, including those with tragic human and environmental
impact, with data derived from a multitude of sources in a
timely manner for decision makers and for citizens at large.
Summary.
Working collaboratively with our industry partners, who have
joined with us in planning and committed to participating in
executing this project, will help extend our center’s capabilities
to begin addressing fundamental research problems of societal
relevance requiring longer term development through the
effective integration of our broad informational science base
with its rich variety of computing resources that are increasingly
heterogeneous in their architecture, distribution and
organization.
This project will allow us to synthesize best current practices in
divergent areas of multi-core computing, data assimilation
models and knowledge representation while exposing areas
requiring new approaches to better utilize emerging computing
paradigms to solve urgent broad social needs.
We propose to develop a test bed virtual world simulation which
will be computed upon a widely distributed, heterogeneous
compute platform; combining cloud services, virtual world
servers, compute accelerators and client systems ranging from
multi-core desktops to mobile phones.
We will develop the methods by which the multi-user virtual
world can have a coherent representation of assets which differ
by type and scale. We will utilize the distributed compute
services to integrate these divergent assets for effective real
time data assimilation of multi-media data information streaming,
meta-tagging, into real world socio- physical models, distributed
media databases, creating specific virtual world experiences for
distributed viewers and decision makers, whose activities will
interact with the virtual world to make more effective decisions
in planning event response by providing a scenario testing
platform.
Different aspects of the virtual world will be computed using the
best available resource to its particular needs, with a shared
database tracking resource availability and cataloging assets.
Each center site will bring specific expertise to bear: UCSD will
extend its work in virtual world client server architectures; UMBC
will develop new techniques in multi-media database
management to encompass the emerging types of data along with
a methodology for cloud hosting and network traffic analysis as
indicators of event conditions. Georgia Tech will undertake asset
analysis and network data mining. In this first year, we will begin
by collecting pre-event data bases to form the baseline virtual
world, and then update with static data at various time increments,
up to the point of integrating real-time data by year end.
The work will be readily applicable to develop future event
response situations, either natural or anthropogenic. We
will begin this work utilizing data from the recent Haitian
earthquake of January, 12, 2010, and the dynamic relief
efforts which followed and will continue for some time.
This recent event will allow us look across different
temporal windows of the situation and allow us to show
which methods have the most validity for further study and
development.
12 January, 2010 magnitude 7.0 earthquake
Cathedral of Port-au-Prince, August 25, 2009 – Google Earth
Partially Destroyed Cathedral of Port-au-Prince, January 14, 2010
As Seen From A U.S. Air Force RQ-4 Global Hawk Unmanned Aircraft
Cathedral of Port-au-Prince, January 24 2010 – Google Earth
The BLUR is a virtual
world where many
people interact and
update heterogeneous
datasets of dynamic
environments.
LiDAR scan Torrey Pines State Beach by Falko Kuester, UCSD
Data is mined from
communication
sites, and linked
into the evolving
virtual world, along
with portals for
mobile media
devices,
supplementing
more deliberate
methods.
"many newspapers have put precious resources into developing a people-finder system."
However, he explained, "This excellent idea has been undermined by its success: Within 24
hours it became clear that there were too many places where people were putting
information, and each site is a silo."
"Sharing common data and making it all accessible is key, because if Jean Q. Publique
enters a name of a missing loved one on one site, and another person says they have found
that loved one on a different site, they will not connect," Csikszentmihalyi told me by e-mail.
Others were thinking of a single, centralized site as well, including designer Tim Schwartz,
who createdHaitianquake.com, a registry for tracking missing persons.
"I realized immediately that in our Web 2.0 environment, with tons of social networking sites,
that missing people information was going to go everywhere on the Internet, and it would be
very hard to actually find people and get back to their loved ones if everything was
scattered," Schwartz said by e-mail. "So, my initial goal was to create a unified database
that would be the one repository for missing people data, and other online applications could
connect to it."
Over the weekend, the 22,000 entries collected at Haitianquake.com were moved to
Google's database, which Schwartz said was up to 30,000 records "and growing" as of early
Sunday evening. (I have not received any figures from Google on the number of records in
its system.)
Users interact with the data within the virtual world;
creating, correlating, validating and annotating data to
derive consensus scenarios. Visual semantics allow for a
wide range in data qualifications for scenario inclusion.
LiDAR scan at Yosemite National Park by J. Toby Minear, UCB
Developed in discussions with industry partners, who
have joined with us in planning and committed to
participating in executing this project:
Northrop Grumman – remote sensing, visualization,
data integration, command and control
NOAA – informal data collection and integration
LTS – cyberspace security applications
NIST – cloud computing integration
IBM – Smarter Planet initiatives
The virtual world simulation has five interrelated
operations:
1) It provides the scheme to correlate the various pieces of
data
2) A collaborative tool in which data elements are improved
and verified
3) An experiential tool to enable understanding of how the
different components of a complicated scenario unfold
4) Facilitates short term planning activities at the real site
5) A platform to simulate scenario forecasting for purposes
such as long-term planning or future incident mitigation.
Distributed computation
User participation,
extensibility and
coherency Distributed
Multi-Media Data Base
Management
Network Analysis
Asset Analysis, Integration
and Computation
Mobile Devices
Quarter 1: Establish an Ad Hoc CHMPR member industry planning and
evaluation team.
Quarter 1: Implement a Distributed Heterogeneous Cloud Computing system
based on the open Apache provided Hadoop system consisting of their File
system, H base system for large image data base querrying and the
MapReduce algorithm for content search and query.
Quarter 1 Milestone: Import large volumes of Haiti earthquake data for storage in
the H base system
Quarter 2 – 3: Extend the virtual scalable city world with real world data.
Quarter 2 Milestone: Visualize Haitian data static data from distributed cloud
Quarter 3 Milestone: Distribute visualization to simultaneous multiple users.
Quarter 4: Demonstration scenario with advanced GPS equipped multi-media
mobile devices
Quarter 4 Milestone: Test quality control features on human acquired event data.
Quarter 4: Test uploading of streaming data from variety of devices
Quarter 4: Implement user data refinement annotations within virtual world
Quarter 4 Milestone: Compare user refined and unrefined models through
external analysis. Quarter 4: Run simulation extraction scenarios for initial
demonstration of proof of concept. Road re-devlopment as initial test case.
.