Department of Computer Science University of Rome Sapienza Italy

1
Department of Computer Science University of
Rome Sapienza Italy
CEOS-WGISS meeting Monaco, 16/10/07
Precision agriculture via Sensor Networks Un. Of
Rome La Sapienza
Chiara Petrioli
Security Lab, Sapienza Innovazione, Sapienza Un.
of Rome
2
Wireless Sensor Networks

• WSNs are wireless ad-hoc multi-hop
  self-organizing networks made of tiny sensor
  nodes cooperating to monitor the environment

3
Application of Interest Frascati Living Lab
Serial
SCU (WLAB)

• Objective
• Extending the SCU functionalities by a
  distributed
• wireless system for monitoring environmental
• parameters (e.g. temperature, humidity, light
  ...)
• on the vineyard
• Tmote SKY devices by Moteiv corporation
• TinyOS 2.x
• Distributed vs centralized monitoring
• Minimal infrastructure
• Wireless communication
• Ad-hoc network
• Mesh topology
• Devices are battery powered
• energy constraints
• energy efficient protocols and algorithms
• Integration with the Special Communication Unit
  (SCU)
• acting as the sink of the WSN

(Un. of Rome)
GPRS
GIS (ESA)
4
Sensor nodes platforms

• ALBA, IRIS protocol stacks
• Implementations on TmoteSky, EYES v2.0 platforms

Texas Instruments Mps430 micro-controller, 16-bit
RISC CPU, 8 Mhz, 10Kb RAM, 48Kb ROM, fast wakeup
(lt 6us), integrated 12-bit ADC/DCA converter,
expansion SPI bus. Light, temperature on board
sensors.
EyesIFXv2 radio chip TDA5250, 868Mhz, FSK
modulation, datarate 64Kbps, on board 512Kb
serial EEPROM
5
TmoteSky Energy model
Transceiver states
CC2420 Modules
Energy model
6
Nodes awake-asleep schedule

• At the end of every sleeping cycle it randomly
  picks a real number ta in 0,T(1-d) (d is the
  duty cycle).
• In the following sleeping cycle the node will
  sleep for the first ta seconds, will then wake up
  for Td seconds, and go to sleep again till the
  end of the sleeping cycle.

7
EYES IFXv2 Energy model
Transceiver states
Energy model
8
Sensor Netwoks prototypes solutions available

• What is currently available
• Development, optimization and implementation of
  two general purpose protocol stacks for sensor
  networks

• Features
• Awake-asleep schedule, MAC,
• Hop count based routing, cost-based
• relay selection
• Integrates interest dissemination and
• convergecasting

• Features
• Awake-asleep schedule, MAC,
• geographic routing, load balancing
• Copes with dead-ends
• Resilient to localization errors,
• adapts to network dynamics

9
IRIS-Interest Dissemination
Neighborhood size estimation

• Say that an estimation procedure lasts r rounds
• Let ki be the number of active neighbors at the
  i-th round which have not been counted before
• After r rounds the probability that the number
  of sampled active neighbors k1, k2, ,kr if the
  number of neighbors is n and the duty cycle is d
  is given by

• n is the number of neighbors and

Interest Dissemination

• Each node receiving an interest tosses a coin
• with probability p it rebroadcast the interest
  to all its neighbors
• with probability (1-p) it picks c of its
  neighbors randomly and send the interest to them
• p0.2 and c4 is a proper parameter setting (all
  intended destinations reached)

10
IRIS-Convergecasting
Relay selection criteria

• During the interest dissemination nodes discover
  their distance in hops from the sink (Hop Count
  or HC)
• If a node h hops from the sink has a packet to
  transmit it selects a relay among awake neighbors
  which are h or h-1 hops from the sink
• Each node has associated a cost
• The cost can reflect residual energy, congestion
  level of the traversed nodes, links reliability,
  nodes capability to aggregate packets
• The way the relay is selected aims at minimizing
  the cost to advance of one level

MAC operations

• CSMA like, simple, implementable on real
  prototypes
• Mechanisms to allow nodes to exploit all
  information available to go to sleep as much as
  possible

11
ALBA-Adaptive Load Balancing Algorithm

• The relay selection works in phases
• Selection of the best QPI
• Awaking nodes can participate
• Selection of the best GPI
• Performed if more than one node with the
• same QPI was found
• Nodes awaking in the middle of a GPI
• contention cannot participate

12
Rainbow Coping with Dead-Ends
13
Size, cost, network lifetime

• Sensors for outdoor environments available
• Miniaturized sensors available
• Easy and inexpensive housing in plastic boxes of
  standard sensors possible
• The technology is there, costs already limited
  (70 euros each), can decrease to a few euros in
  case of massive deployment
• Challenge is network lifetime if d0.1 current
  prototypes last around 3-4 months with a 4
  battery packs
• Prototypal transceivers are being deployed which
  reduces energy consumption to 1/5 wrt ZigBee
  transceivers
• The network could normally operate at a lower
  duty cycle switching on to a higher duty cycle
  only when there is traffic to transmit
• Protocols can exploit application features,
  saving energy (switching the transceiver ON only
  when needed)

TmoteINVENT
TmoteMINI