1. Wireless Sensor Networks for Habitat Monitoring
Alan Mainwaring1
Joseph Polastre2
Robert Szewczyk2
David Culler1,2
John Anderson3
1: Intel Research Laboratory at Berkeley
2: University of California, Berkeley
3: College of the Atlantic

2. Introduction
Application Driven System Design, Research, and Implementation
Parameterizes Systems Research:
Localization
Calibration
Routing and Low-Power Communications
Data Consistency, Storage, and Replication
How Can All of these Services and Systems Be Integrated into a Complete Application?

3. Great Duck Island
Breeding area for Leach’s Storm Petrel (pelagic seabird)
Ecological models may use multiple parameters such as:
Burrow (nest) occupancy during incubation
Differences in the micro-climates of active vs. inactive burrows
Environmental conditions during 7 month breeding season
Clearly, such a model would consider multiple parameters. We’re focusing on ones in-and-around the underground nests (burrows) where eggs are laid.
p.s. If anyone asks why they are called ‘petrels’, here’s the story: The birds are planktonic feeders and spend hours during the day hovering above the ocean’s surface picking bits of planton out with their feet. Sailors being a superstitious bunch likened this to walking on water, something for which St. Peter was famous.

4. Application
> 1000 ft

5. Sensor Network Solution

6. Outline Application Requirements Habitat Monitoring Architecture
Sensor Node
Power Management
Sensor Patch
Transit Network
Wide Area Network and Disconnected Operation
Sensor Data
System Analysis
Real World Challenges

7. Application Requirements
Sensor Network
Longevity: 7-9 months
Space: Must fit inside Small Burrow
Quantity: Approximately 50 per patch
Environmental Conditions
Varying Geographic Distances
Inconspicuous Operation
Reduce the “observer effect”
Data
As Much as Possible in the Power Budget
Iterative Process

8. Application Requirements
Predictable System Behavior
Reliable
Meaningful Sensor Readings
Multiple Levels of Connectivity
Management at a Distance
Intermittent Connectivity
Operating Off the Grid
Hierarchy of Networks / Data Archiving

9. Habitat Monitoring Architecture
Transit Network
Basestation
Gateway
Sensor Patch
Patch
Network
Base-Remote Link
Data Service
Internet
Client Data Browsing
and Processing
Sensor Node
Pictorial outline

10. Sensor Node: Mica Hardware Software Atmel AVR w/ 512kB Flash
916MHz 40kbps Radio
Range: max 100 ft
Affected by obstacles, RF propogation
2 AA Batteries
Operating: 15mA
Sleep: 50mA
Software
TinyOS / C Applications
Power Management
Digital Sensor Drivers
Remote Management & Diagnositcs

11. Sensor Node: Power Management
AA Batteries have ~2500 mAh capacity
Mica consumes 50mA in sleep = 1.2 mAh/day
Mica Expected Lifetime
Node Activity
Days
Years
Mica Always On 7
0.1
Mica Always Sleeping
2081
5.7
Expected Lifetime (days)
Number of Operating Hours per Day

12. Sensor Node: Power Management
Operation
nAh
Transmitting a packet
20.000
Receiving a packet
8.000
Radio Listening for 1ms
1.250
Operating Sensor for 1s (analog)
1.080
Operating Sensor for 1s (digital)
0.347
Reading a Sample from the ADC
0.011
Flash Read Data
1.111
Flash Program/Erase Data
83.333
Target Lifetime: 7-8 months
Power Budget: 6.9mAh/day
Questions:
What can be done?
How often?
What is the resulting sample rate?
Operation
Operating Time per Day
Duty Cycle
Sample Rate
Always Sleep
24 hours 0%
0 samples/day
+ mCPU on
52 minutes
3.61%
+ Radio On (Listen)
28 minutes
1.94%
+ Sample All Sensors
21 minutes
1.45%
630 samples/day
+ Transmit Samples
20 minutes
1.38%
600 samples/day

13. Sensor Node: Mica Weather Board
Digital Sensor Interface to Mica
Onboard ADC
Designed for Low Power Operation
Individual digital switch for each sensor
Designed to Coexist with Other Sensor Boards
Hardware “Enable” Protocol to obtain exclusive access to connector resources
Meeting the guidelines for env monitoring system

14. Sensor Node: Mica Weather Board
Accuracy
Interchange
Max Rate
Startup
Current
Photo
N/A
10%
2000 Hz
10 ms
1.235 mA
I2C Temp
1 K
0.2 K
2 Hz
500 ms
0.150 mA
Pressure
1.5 mbar
0.5%
10 Hz
0.010 mA
Press Temp
0.8 K
0.24 K
Humidity 2% 3%
500 Hz
0.775 mA
Thermopile
3 K 5%
200 ms
0.170 mA
Thermistor
5 K
0.126 mA
Work through power budget
Energy per sample vs current x sample
Important to Biologists
Affect Power Budget

15. Sensor Node: Packaging
Parylene: Great but sucks for connectors or exposed sensors (non soldered connections)
Acrylic
Size matters
Ventilation
Parylene Sealant
Acrylic Enclosures

16. Sensor Patch Network Nodes: Transmit Only Network Single Hop Repeaters
Approximately 50
Half in burrows, Half outside
RF unpredictable
Burrows
Obstacles
Drop packets or retry?
Transmit Only Network
Single Hop
Repeaters
2 hop initially
Most Energy Challenged
Adheres to Power Budget

17. Transit Network Two implementations Antennae
Linux (CerfCube)
Relay Mote
Antennae
No gain antenna (small)
Omnidirectional
Yagi (Directional)
Implementation of transit network depends on:
Distance
Obstacles
Power Budget
Duty cycle of sensor nodes dictates transit network duty cycle
Same power budget could apply here using sophisticated scheme using traffic for entire patch, or you can engineer around it… Provision node adequately

18. Transit Network Renewable Energy Sources CerfCube needs 60Wh/day
Assuming an average peak of 1 direct sunlight hour per day:
Panel must be 924 in2 or 30” x 30” for a 5” x 5” device!
A mote only needs 2Wh per day, or a panel 6” x 6”

19. Base Station / Wide Area Network
Disconnected Operation and Multiple Levels of State
Laptop
DirecWay Satellite WAN
PostgreSQL
47% uptime
Redundancy and Replication
Increase number of points of failure
Remote Access
Physical Access Limited
Keep state all areas of network
Resiliency to
Disconnection
Network Failures
Packet Loss
Potential Solution: Keep Local Caches Synchronization
Challenges and Accomplishments: CoLo analogy etc

20. Sensor Data Analysis
Tell story about john anderson

21. Sensor Data Analysis
Outside Burrow
Inside Burrow

22. System Analysis Power Management Goals
Calculated 7 months, expect 4 months
Battery half-life at 1.2V
Predictable Operation
Observed per node constant throughput, % loss
739,846 samples as of 9/23, network is still running
Battery Consumption at Node 57
Packet Throughput and Active Nodes

23. Real World Experiences
System and Sensor Network Challenges
Low Power Operation (low duty cycle)
Affects hardware and software implementation
Multihop Routing
Allows bigger patches
Route around physical obstacles
Must have ~1% operating duty cycle
In Situ Retasking/Reconfiguration
Let biologists interactively change data collection patterns
Not Implemented due to conservative energy implementation
Lack of Physical Access
Remote management
Disconnected operation
Fault tolerance
Reliance on other people and their networks
Physical Size of Device
Affects microcontroller selection, radio, practical choice of power sources

24. Real World Experiences
Failures
Extended Loss of Wide Area Connectivity
Unreliable Reboot Sequence in Windows
Solderless Connections Fail (expansion/contraction cycles)
Node Attrition (Petrels are not mote neutral)
Environmental Conditions (50km/hr gale winds knock over equipment)
Lack of post-mortem diagnositics

25. Conclusions
First long term outdoor wireless sensor network application
Application driven sensor network design
Defines requirements and constraints on core system components (routing, retasking, fault tolerance, power management)

27. Backup Slides

28. Mote 18: Outside

29. Mote 26: Burrow 115a

30. Mote 53: Burrow 115b

31. Mote 47: Burrow 88a

32. Mote 40: Burrow 88b

33. Mote 39: Burrow 84