Transcript VanhookLL100799

Dynamic Declarative Networking
Exploiting Declarative Knowledge To Enable
Energy Efficient Collaborative Sensing
Daniel J. Van Hook
[email protected]
MIT Lincoln Laboratory
Lexington, Massachusetts
SensIT PI Meeting
October 7-8, 1999
MIT Lincoln Laboratory
SensIT Dynamic Declarative Networking
NEW IDEAS
• Declarative subscription/publication protocols drive ad
hoc network organization
• Sensor applications subscribe and publish (request to
consume/offer to produce data) using abstract
specifications of type, resolution, range, rate, reliability,
quality, size, etc. via a Declaration Services API
• Exploit sensors’ subscriptions and publications at
multiple levels of the communications stack via
Declaration Services layer
IMPACT
• Employing sensor applications’ abstract data declarations at
multiple layers of the network stack facilitates optimization while
maintaining modularity
• Energy constraints of sensor networks counterbalanced through
declarative subscription and publication control of routing
• Simulation framework and militarily relevant scenarios provide a
methodology and structure for comparing and experimenting
with alternative protocols and approaches
• Hardware experiments pit algorithms against departures from
simulation models; feedback improves fidelity of models,
robustness of algorithms
SCHEDULE
•Declarative Routing
Protocol (DRP)
•Develop simulation
framework
•Develop sensor
simulation
scenarios
•Implement DRP
in WINS nodes
and demo
•Sensor system
simulation
experiments
•Extend declarative
routing concepts
•Extend sensor
system simulation
experiments
MIT Lincoln Laboratory
Declarative Approach to Networking:
Subscription and Publication
Making declarative knowledge explicitly available allows the
infrastructure to make better resource tradeoffs.
MIT Lincoln Laboratory
Declarative Routing Protocol (DRP)
• •• • • • •
•
•
•
•
•
•
••
•
• ••
•
• • •• • •
•
•
•
•
•• • • • •
•
• ••
• •
• • • • • •
• •• •
• • • •
• •
PUB
PUB
SUB
PUB
IP_Bound
Global_Flood
IP_P2P_Bound
IP_All_P2P
DRP
MIT Lincoln Laboratory
Battlefield Acoustic Detection and
Tracking Scenario
•
•
•
Acoustic “dust”
sensor array
•
Acoustic “dust” sensors detect,
classify, identify, and track threats
Hierarchical sensor fusion
Network self–organizes based on
sensor data declarations
(subscriptions and publications) via
Declarative Routing Protocol (DRP)
10,000 node simulation completed
Report
Track
Estimate Bearing
Detect, Classify, Identify
MIT Lincoln Laboratory
Secure Building Scenario
•
•
•
•
•
•
Acoustic
Imager
Acoustic (alarmed)
Imager (alarmed)
X
Acoustic and imaging sensors
deployed after building is swept
to detect and report on intruders
Acoustic sensors cue imaging
sensors; images are relayed to a
command post
Building used is the South Lab at
MIT Lincoln Laboratory: 4 floors,
~300,000 Sq. Ft.
Threat enters on first floor, up a
staircase, and then continues on
the third floor across the building
414 nodes
240 seconds duration
Threat
MIT Lincoln Laboratory
Summary of Progress
• Declarative Routing Protocol (DRP)
– Prototype of DRP is running
– Initial evaluation of DRP complete; but more to do
– Declaration Services API defined
– Ready to adapt to WINS for demonstration
• Simulation framework for sensor system protocol/algorithm
experimentation developed
– GloMoSim/PARSEC–based
– Visualization tool for observing scenario evolution
– Instrumented for energy, throughput, losses, collisions, etc.
• Scenarios to support experimentation developed
– Secure building scenario
– Battlefield acoustic detection and tracking scenario
– Random node placement/traffic (for scalability testing)
MIT Lincoln Laboratory