Presenter: Kevin Wine WINLAB, Wireless Information Networks Laboratory, Rutgers University - PowerPoint PPT Presentation

1 / 15
About This Presentation
Title:

Presenter: Kevin Wine WINLAB, Wireless Information Networks Laboratory, Rutgers University

Description:

Large packets that cover one or more dips probably won't get through! ... This packet will probably NOT get through unless you're very close. ... – PowerPoint PPT presentation

Number of Views:81
Avg rating:3.0/5.0
Slides: 16
Provided by: rma116
Category:

less

Transcript and Presenter's Notes

Title: Presenter: Kevin Wine WINLAB, Wireless Information Networks Laboratory, Rutgers University


1
  • Presenter Kevin Wine WINLAB, Wireless
    Information Networks Laboratory, Rutgers
    University
  • PI Badri Nath
  • Site Visit Briefing
  • August 2001
  • http//www.cs.rutgers.edu/dataman/webdust
  • badri_at_cs.rutgers.edu
  • Co-PIs Tomasz Imielinski, Rich Martin

2
Mobility in Sensor Networks - Motivation
  • We are interested in mobility in sensor networks
    because
  • Users interested in sensor data will not always
    be stationary
  • Mobile users will need to inject a sensor query
    into the sensor network while they are mobile.
  • Mobile users will need to retrieve a sensor
    response from the sensor network while they are
    mobile.

Vehicle travel path
Query submitted to sensor network
Response retrieved from sensor network
Sensor Field
3
Mobility in Sensor Networks Mobile Channel
  • Mobility does bad things to the channel

Path Loss
As much as 50db!!
On the order of milliseconds
Time
4
Mobility in Sensor Networks Mobile Channel
  • Time between dips is a function of velocity and
    carrier frequency

(60 MPH 5280 ft/mile)
88 feet/sec
Mobile Velocity
(60 min 60 seconds)
(186,000 MPH 5280 ft/mile)
1.09 feet
Wavelength
(900MHz carrier)
Rule of Thumb One dip for every one half
wavelength of travel. So that would be a dip for
every 6 inches of travel.
1
5.7ms
Time between dips
(88 ft/sec 2 dips/ft)
5
Mobility in Sensor Networks Packet Sizing
  • Packets should fit between the dips
  • Large packets that cover one or more dips
    probably wont get through!
  • General idea the smaller the packet the better.

These packets will probably get through.
5.7 ms (900MHz, 60MPH)
This packet will probably NOT get through
unless youre very close.
6
Mobility in Sensor Networks Mobile Channel
  • The previous mobile channel plot I showed you is
    a best case example controlled environment
    line of site.
  • Add stationary obstacles (trees, buildings) and
    obstacles in motion (cars, trucks, people) and
    the situation gets much worse!!
  • Result widely varying channel conditions that
    change quickly from good to bad.

7
Mobility in Sensor Networks Mobile Channel
Truck in the way
Behind a building
Path Loss
Time
  • Further measures are usually required in a mobile
    environment to achieve same throughput as in a
    fixed environment.
  • We want to get as much throughput as possible
    given this channel and power-constrained radios.

8
Mobility in Sensor Networks Adaptive Redundancy
  • Transmit the same packet multiple times to
    increase chances of reception.
  • Separate the multiple packets in time to overcome
    extended channel outages (time diversity).
  • Adapt the number of multiple packets and their
    time separation to the current channel conditions
  • Good channel Fewer duplicates, closer in time
  • Bad channel More duplicates, further in time

9
Mobility in Sensor Networks Adaptive
Retransmission
  • Simple re-transmission use acknowledgements to
    cause lost packets to be retransmitted.
  • Adaptive re-transmission - Measure the channel
    and retransmit lost packets depending on channel
    conditions
  • Good channel retransmit packet once
  • Bad channel retransmit packet several times
  • Should be implemented at the link layer for best
    possible channel condition tracking.

10
Mobility in Sensor Networks Adaptive Data Rate
Switching
  • Lower data rates are more reliable, but less
    throughput
  • Higher data rates give more throughput, less
    reliability
  • Adjust the link data rate depending on channel
    conditions
  • Good channel use a higher data rate
  • Bad channel use a lower data rate
  • Again, should be implemented at the link layer
    for best possible channel condition tracking.

11
Mobility in Sensor Networks Summary
  • The trick is to find the optimal combination of
  • Packet size
  • small packets cause more overhead but more get
    through
  • Packet redundency and retransmission
  • How many copies to send
  • When to give up
  • How much coding to apply
  • Data rates and switching thresholds
  • At what link quality do you change rates
  • What data rates do you use

Retransmission
Solution
Packet Size
Parameter Space
Coding and Redundency
12
Mobility in Sensor Networks Adding Mobility
Using a Sensor Network Gateway
Gateway acts as bridge between two networks
Other wireless network ie. Wireless LAN
Gateway and Sensor Network Bridge
Sensor Network
Mobile network System
Video Camera
13
Mobility in Sensor Networks Magnetic Field
Acceleration Sensors
  • We are in the process of building magnetic field
    and acceleration sensors for the mote platform.

14
Status of Work / Plan
  • Task1 Evaluated impact of mobility on sensor
    nodes (motes) (Completed July 01)
  • Task2 Design, evaluate, implement schemes for
  • Supporting mobility in sensor networks
  • Size, coding, rate adaptive schemes (Year 2)
  • Task3 Implement schemes to support
  • mobility on sensoria nodes. Experiment, evaluate
    choices
  • For mobility support in sensor networks.
    Scalability issues (Year 3)

15
Information
  • http//www.cs.rutgers.edu/dataman/webdust
  • E-mail kevinw_at_winlab.rutgers.edu
Write a Comment
User Comments (0)
About PowerShow.com