Real Problems. Real Research

1
Real Problems.Real Research?

• Vicraj (Vic) Thomas
• vic.thomas_at_honeywell.com
• 22 March 2007

2
Background

• Staff Scientist at Honeywell Labs, Minneapolis
• Research Interests
• Distributed embedded systems
• Sensor networks
• Wireless networks for flight-essential aircraft
  systems
• Dependable system design
• Recent sensor networking related projects
• Wireless sensor networks for vehicle health
  management
• Wireless sensors for avionics systems
• Networked MEMS optical sensors for battlefield
  situation awareness
• Quality of information from sensor networks

Im not young enough to know everything Oscar
Wilde
3
Observations

• Fielding a sensor network is non-trivial
• But not for reasons youd expect if you delve
  into the research out there
• Its the mundane realities of life that get you

Practical wisdom is only to be learned in
the school of experience Samuel Smiles
4
Real Problems

• True Stories from Sensor Network Deployments

5
1. The Grass Grows Fast in Florida

• System Sensor network to detect an event
• Our focus
• Minimize false positives and negatives
• Maximize lifetime of network
• Observed system behavior
• Large number of false positives
• High energy consumption lot more messages than
  expected
• Problem Noisy PIR sensors
• Cause Tall grass blowing in the wind
• Solution Get facilities to mow the grass short

For every problem there is a solution which is
simple, clean and wrong. Henry Louis Mencken
6
2 Its a Cold December Morning in Albuquerque

• System Sensor network for vehicle health
  management
• Our focus
• System reliability
• Maximize battery life
• Easy of use
• Observed system behavior
• Everything worked great night
• before demo
• Nothing worked next morning
• Problem
• Sensor nodes behaving erratically
• Cause One of the crystals was rated for gt0C
• Solution Postpone demo to the afternoon

There are three intolerable things in life cold
crystals (coffee), lukewarm champagne, and
overexcited women. Orson Welles (with apologies)
7
3 Radio Waves Cant Read Do Not Disturb Signs

• Situation
• Multiple teams in the same hotel for sensor
  network demonstrations
• Teams worked late into the night fine-tuning
  their software
• Problem
• Next day one team finds its sensor network
  behaving very strangely
• Cause
• This teams sensor nodes had gotten accidentally
  reprogrammed by different team working in nearby
  room
• Both teams were using the same default system
  ID used by the over-the-air programming
  software
• Solution Change system ID

Whenever evil spreads, it is only by default.
Ayn Rand
8
Real Research?

• Observations on Sensor Network Research

9
Sensor Network Research

• Is this a hot area of research?
• 42 conferences/workshops in 2007 with interest in
  sensor networks
• http//www.cs.virginia.edu/adw5p/conferences.html
  
• Common research topics
• Sensor network services
• Routing, time-synchronization, localization,
  management
• Cross-layer protocol design
• Sensor network security
• Programming
• Operating systems, simulation tools
• Is this research important?
• Yes, but need proper perspective on importance
• Too much old wine in new bottles
• Wireless aspect of these sensor networks often
  ignored

vs.
When we try to do something new, we don't know
what we're doing. That's the biggest challenge.
Jeffrey Kalmikoff
10
Research Needs

• Compensating for the poor quality of the sensing
• Sensor malfunction most likely failure mode
• Sensing channel attacks most likely security
  threat
• Can we make up in quantity what weve lost in
  quality?
• More realistic failure models
• Account for faults induced by the environment
• Scientific bases for designing, deploying and
  operating sensor networks
• Estimates of confidence that can be placed on the
  data
• Optimal data fusion architectures
• Quantifying trade-offs
• Alternate energy sources
• Dont dream up problems to be solved
• Plenty of real problems waiting to be solved

Nature is ugly, and I prefer the monsters of my
fancy to what is positively trivial Charles
Baudelaire