Title: Wireless Sensor Networks for Habitat Monitoring
1Wireless Sensor Networks for Habitat Monitoring
- Presented by Jas Ahluwalia
- 4/15/03 (Doh! Tax Day)
- CSE291Programming Sensor Networks
- Prof. Andrew Chien
2Why Sensor Networks for Habitat Monitoring?
- Human presence messes things up!
- Can change the behavioral patterns that we are
trying to monitor - Can even destroy sensitive populations by
reducing breeding success, causing shift to
unsuitable habitats, introducing exotic elements,
etc. - Providing food and lodging for human researchers
in such remote locations is costly.
3Why Sensor Networks for Habitat Monitoring? (Cont)
- Small sensors can be deployed during periods
when sensitivity to human presence is minimal. - Non-Breeding times
- When Plants are Dormant
- Implement a Deployem and Leaveem strategy.
- Broaden the scope of study sites which were
previously limited by concerns of human presence
or danger to humans.
4Goals Motivation
- Develop a specific habitat monitoring application
that will represent all problems within this
domain. - Taking an application-driven approach quickly
separates actual problems from potential ones and
relevant issues from irrelevant ones. - True or False?
5Great Duck Island
- 237 acre island 15 km south of Mount Desert
Island. - Wish to monitor the Leachs Storm Petrel
- Usage pattern of nesting burrows
- Environmental differences in burrow and on
surface. - Differences in micro-environments with and
without large numbers of nesting petrels. - Various data needs with various data acquisition
rates.
6Great Duck Island Requirements
- Internet Access
- Support remote interactions with in-situ networks
- Hierarchical Network
- Discussed Later
- Sensor Network Longevity
- 9-12 months
- Operating Off-the-grid
- No wall sockets! Must operate on battery power,
solar power, etc.
7Great Duck Island Requirements (Cont)
- Management at-a-distance
- Goal is zero on-site presence, including
maintenance and administration. - Inconspicuous Operation
- Invisibility as discussed by Prof. Chien.
- System Behavior
- Stable, predictable, and repeatable behavior. But
is this possible in the physical world? - In-Situ Interactions
- Some local interactions may be required via PDAs.
Huh? What about the first bullet?
8Great Duck Island Requirements(Cont)
- Sensors and Sampling
- Light, temp, humidity, barometric pressure, etc.
- Data Archiving
- Archiving of data acquired by sensors for
off-line data mining. Additionally, reliably
offloading this acquired data to databases in the
wired world.
9Planned System Architecture
- Sensor Nodes
- Perform general purpose computing, networking,
and sensing. - Sensor Patch
- Group of Sensor Nodes.
- Gateway
- Responsible for transmitting sensor data from
patch through a local transit network. - Transit Network
- Single hop link or series of networked wireless
nodes in a path from gateway to base station. - Base Station
- Provide Wide area connectivity and a Database
Management System.
10Panned System Architecture (Cont)
11Planned System Architecture (Cont)
- Sensors form a multihop network by forwarding
each others messages. - Each Layer has some form of storage
- Sensor Level Data Logging
- Base Station Level Full Relational DBMS
- Gateway Level Something in between.
12Implementation
- So that was the plan, how did they implement a
Sensor Network at Great Duck Island.
13Sensor Node
- Sensor Node
- MICA as discussed by Johann.
-
14The MICA Weather Board
- Provide the same functionality as a traditional
weather station. - Startup time dominates power consumption, not
sample rate due to low duty cycle. - Through calibration, interchangeability and
accuracy can be reduced to below 1.
15The MICA Weather Board (Cont)
16Energy Budget
- Target Lifetime is Approx. 7-8 months
- Power Budget is 6.9 mAh/Day
Expected Lifetime (months)
Number of Operating Hours per Day
17Energy Budget (Cont)
18Sensor Deployment
- Coat entire sensor package with 10 micron
parylene sealant to provide water protection. - Sensors placed in protective packaging that
minimally obstruct sensing functionality
(ventilated). - Sensors placed in burrows without protective
packaging due to size constraints.
19Sensor Patch Network
- Single Patch Network of 32 Motes
- 9 of which are inside burrows
- Single Hop Model
- Broadcast to gateway during scheduled
communication period. - But didnt we say were going to be using a
multihop network?
20Gateway
- 2 Designs
- CerfCube (solar power)
- Embedded Linux system.
- 1GB Storage Space.
- 802.11b
- Mote-To-Mote (solar power)
- Mote connected to Base Station and Mote in Sensor
Patch. - 14dbi directional 915MHZ Yagi antennae. 1200 ft
range. - Tests yielded equivalent identical packet
reception rates. - Mote-To-Mote solution chosen because less
intrusive hardware and consumes far less power (2
orders of magnitude). - Do we still have Gateway and Transit Network
levels of Hierarchies?
21Base Station
- Wide area connectivity provided through 2 way
satellite. - Laptop running relational database.
22DBMS
- PostGres SQL
- Replicated every 15 min over satellite link to DB
in Berkley
23User Interfaces
- Various Database Interfaces
- Web based Interfaces
- Developed Java applet which provides access to
habitat data. - Gizmos not fully developed
- Experiments iPaq PDA running Linux.
24Results
- Verified that data collected by sensors is
correct. - When satellite goes down, data continues to be
logged on the island and connectivity is brought
up again, the secondary database is brought up to
date. - Power Management
- Calculated 7 months, expect 4
- Petrels are not Mote Neutral
- What does that mean?
- 50 km/hr winds knock equipment down
- 739,846 samples as of 9/23 (3 months after
deployment), network is still running
25Future Work and Discussion
- Data Sampling and Collection
- Sampling rates, aggregate collection, etc.
- Communication
- Ad Hoc Routing
- Network Retaking
- Health and Status Monitoring
- Power Optimization/Conservation Techniques