Deploying%20Long-Lived%20and%20Cost-effective%20Hybrid%20Sensor%20Networks

1
Deploying Long-Lived and Cost-effective Hybrid
Sensor Networks
Nirupama Bulusu, Portland State University joint
work with Wen Hu, Chun-tung Chou, Sanjay
Jha (University of New South Wales)
2
Our Sensor Net App

• Cane Toad Monitoring
• Cane toads will progressively shell
• shock an unsuspecting Kakadu environment,
• in particular they will come close to wiping
• out native quoll populations, poison large
• masses of goannas and disturb the food
• supply of many native animals.
• Requirements
• Cheap, wide-area sensing coverage
• but sophisticated processing, communication for
  target localization and vocalization

Cane Toad Explosion
Co
3
Hybrid Sensor Networks


Platforms
A Hierarchical View
4
Previous Work Data Anycast

• Deliver data to preferably nearest micro-server,
  which can
• Forward it using out of band, broadband
  (802.11b, 802.11g) communications link
• Store/process it
• Perform the desired actuation
• Build up a reverse-tree at each sensor
  efficiently
• (scalability, adaptability/mobility, little
  over-head/energy efficient, distributed,
  simplicity)
• Reference
• Wen Hu, Nirupama Bulusu, and Sanjay Jha, A
  communication paradigm for hybrid sensor/actuator
  networks, in Proceedings of the IEEE
  International Symposium on Personal, Indoor and
  Mobile Radio Communications (PIMRC 2004),
  Barcelona, Spain, October 2004.

5
This Paper Deployment Issues

• What is the maximum network lifetime possible for
  a given number of micro-servers?
• What is the optimal micro-server placement of
  micro-servers to achieve maximum lifetime?
• Is the deployment cost-effective? How to achieve
  optimal cost benefit?

6
Cost Model

Energy consumption limitation of sensor

Energy consumption limitation of
Micro-server (assuming sensors, micro-servers in
2D grid layout)
7
A Tabu Search Algorithm

• The model is a huge combinatorial problem.

• An approximation tabu-search algorithm
• Algorithm benchmark (a 20 grid network)

8
Network lifetime of a 100 grid network


9
Micro-server Placement
10
Cost-effectiveness (network size100, vary k)


11
Cost-effectiveness (k50, vary network size)
12

Concluding Thoughts

• A methodology for studying hybrid sensor
  deployment (cost model tabu search algorithm)
• spatiotemporal event rates captured as discrete
  variables
• learn from inference, use to reconfigure
  micro-server deployment
• applicable to storage, computation etc.
• Use as engineering guideline, not as a guarantee