ZebraNet

1
ZebraNet

• Rolf Kristensen Torben Jensen
• s022361 s022359

2
Introduction to ZebraNet

• Tracking of Zebra behaviour at Mpala Research
  Center, Kenya
• Wireless sensor nodes
• Sensor network
• High mobility within the sensor nodes and the
  base station

3
Goals

• GPS position samples taken every three minutes
• Activity log taken for 3 minutes every hour
• 1 year of operation
• Operation over a wide range of open land
• No fixed base stations
• High data homing rate (close to 100 pct.)
• A weight of max 1,4 to 2,3 kg

4
A day in the life of a zebra

• Two kinds of Zebras
• Grevys zebra
• Forms large loosely-bounded herds
• Plains zebra
• More common
• Forms tight-knit uni-male multi-female harems
• Possible to add collar just to the male
• From time to time meet up with other zebras to
  form herds, for an example at water holes

5
A day in the life of a zebra

• Movement patterns
• Grazing
• Graze-walking
• Fast-moving
• Zebras spend most day grazing and head for water
  about once a day
• Zebras tend not to have long periods of
  motionless sleep

6
Collar design

• Generally weight and power consumption are very
  important issues
• Two kinds of radio
• Short range (100 m) for communicating between the
  nodes
• Long range (8 km) for communicating with the base
  station

7
Collar design

• Energy issues
• Battery recharged by solar power
• Battery must be able to operate five days without
  being recharged
• Traditional acid batteries are too heavy, so a
  lithium-ion battery must be used
• Space is conserved using a system, where the
  nodes delete old data as newer arrives

8
Collar design

• 30 position samples per hour, 24 hours a day
• 6 hours for searching for peer nodes and
  transferring data over the short-range radio
• 3 hours for searching for the base station using
  long-range radio (overlapping with short range to
  minimize CPU usage)

9
Possible network protocols

• Flooding network protocol
• Simple
• Given that the Zebras move extensively there is a
  high data homing rate
• Large amount of data (requires large bandwidth,
  large storage and much energy)

10
Possible network protocols

• History based network protocol
• Works with a system ranking nodes depending on
  whether they have been close to the base station
  or not
• Data is transfered to the neighbour that has the
  highest hierachy ranking
• If the network changes too dynamically this does
  not work well
• If sucessfull, the requirements for bandwidth,
  storage and energy are lower

11
Experimental results

• Tests performed in software
• Simulating 400 square km map
• Zebra movement based on actual moving patterns
• Each zebra independently selects speed and
  direction
• Once per day the zebra goes to get water

12
Experimental results

• Direct or indirect connectivity
• Direct connectivity requires a radio range of 12
  km for 100 pct. network coverage
• Indirect connectivity requires a radio range of 2
  km for 100 pct. network coverage
• Thus indirect connectivity is necessary, since
  energy consumption is heavier the longer the
  radio range is

13
Experimental results

• Protocol evaluations, given that there are no
  constraints
• Direct transmission is considered, but requires
  long data range (11 km)
• Flooding and history both require only 6 km radio
  range to achieve 100 pct. sucess rate

14
Experimental results

• Protocol evaluations given constraints
• Storage constraints
• Flood and history perform better than direct
  transmission.
• History performs the best, because of storage
  limitations for flooding
• Bandwidth constraints
• With low radio range, and thus few neighbour
  nodes, flooding performs best, but as radio range
  increases history performs better due to less
  redundant data

15
Experimental results

• Energy tradeoffs
• At long ranges flooding is very energy expensive,
  as it sends vast amounts of redundant data
• History fares a lot better, being only 1.04 times
  as expensive as direct transmission at a radio
  range of 15 km
• Flooding only makes sense with low radio ranges
  and low connectivity

16
Experimental results

• Final design choices
• Short-range radio
• 100 m
• 19,2 Kbps
• Long-range radio
• 8 km
• 2,4 Kbps

17
Experimental results

• Final design choices
• 640 KB flash memory
• There can be stored 300 collar-days
• 66 ampere-hours energy consumption per month
• Simulation results in a 83 pct. success rate