Download Oppurtunities in High-Rate Wireless Sensor Networking

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Theoretical computer science wikipedia , lookup

Data analysis wikipedia , lookup

Hendrik Wade Bode wikipedia , lookup

Corecursion wikipedia , lookup

Data assimilation wikipedia , lookup

Multidimensional empirical mode decomposition wikipedia , lookup

Transcript
Opportunities in High-Rate
Wireless Sensor Networking
Hari Balakrishnan
MIT CSAIL
http://nms.csail.mit.edu/
Today’s WSN Monitoring Applications
• Periodic monitoring
repeat:
wake up and sense
transmit data
sleep for minutes
Pic: Sam Madden
• Event-based monitoring
• Transmit data on external event
• Low data rates & duty cycles
Pic: Sam Madden
High-Rate WSN Applications
• High sensing rates: O(102 – 105) Hz
• Non-trivial analysis of gathered data
• Frequency analysis, correlation analysis
• Many domains
• Industrial monitoring, civil infrastructure,
medical diagnosis, process control,…
• What are the reusable components of a
general architecture for high-rate WSNs?
Industrial Monitoring
• Preventive maintenance of fabrication plant
equipment (Intel)
• Done manually today, offline processing
• Sense vibration (acceleration)
• 100 machines, >10 observation
points per machine
• 10-40 kHz frequency band
• Aggregate data rate about 10 – 100
Mbits/s
Pic: Wei Hong
Intel Fab’s “20 Questions”
• Is energy in [f1, f2] > E?
• Compare energy in [f1, f2] with past activity
• Which frequency bands have highest energy?
• What is the phase relationship between
samples at different locations
• Provide high-resolution view of last T mins of
samples at location L
Pipeline Pressure Monitoring
• Preventive maintenance of (aging) water and
sewage infrastructure
• Leaks are precursors to bursts
Pic: Rory O’Connor (MIT)
• Monitor pressure and flow at 0.5 to 2 KHz
• Done manually today
Thames Water’s “20 Questions”
(Thanks to Kevin Amaratunga & Ivan Stoianov)
• What’s the flow / pressure at location L?
• Is pressure / flow at location L different
from dynamic state estimator?
• Has there been a significant pressure drop
between locations L1 and L2?
• How long does it take pressure wave to
travel from L1 to L2?
Constraints
• Wireless communication rates
• Total required raw data rates exceed nextgeneration radio rates
• Energy
• Sensing and communication consume energy
• Want months of operation on batteries
• Unreliable sensor nodes
• “In-the-net” processing essential
Challenges
• High-level programming abstractions
• Distributed signal and data processing
operators
• Collaborative data acquisition
• High-performance network delivery
High-Level Programming
• Users won’t (can’t) write embedded signal
and data processing code
• Generalized stream processing: continuous
query processing + signal processing
• Develop a declarative stream processing
interface
• Support iterative refinement
Generalized Stream Processing
• Application-independent
• Continuous query processing (“TinyDB++”)
• Distributing wavelet, Fourier operators
• “Boxes and arrows” program specification
• Connect up processing operators
• Specify high-level sampling rate
• Specify energy/lifetime constraints
• Support iterative refinement
Supporting Iterative Refinement
Collaborative Data Sampling
• Sampling rates too high for single sensors
• Sensing may not be fast enough, or
• Consumes too much energy
• Group of sensors subsample, collaboratively
produce desired sampling rate
• Spreads processing and energy burden
• How should sub-sampled signals be aligned?
High-performance Data Delivery
• WSNs today have per-node delivery rates
that are 10x worse than they should be
• Obtain 5-10x improvement in throughput
distribution without physical layer changes
• Traditional stack layers considered harmful
• Physical, link+MAC, network layer
decomposition bad for wireless
Traditional Layering has Problems
• With wires, links are shielded from one
another
• Sharing starts only at network layer
• Wireless networks do not have such shielding
• No “links” over the air
• Increasing traffic degrades channel quality
• MAC protocols are too local to resolve
contention correctly
Dismal Throughput Distribution
[HJB, Sensys04]
A Different Layering May Help
• Replace current link+MAC and network layer
decomposition
• Local channel control layer
• Traffic-based rate control, no per-packet
contention resolution
• Has info about other nodes in “region”
• Take advantage of path diversity
• Global topology control layer
• Large-scale routing
Summary
• Many WSN applications require high sampling
rates
• Want general distributed “in-the-net”
processing primitives
• High-performance wireless data delivery
with different layered decomposition