CMPE 257: Wireless and Mobile Networking

1
CMPE 257 Wireless and Mobile Networking

• Spring 2005
• Location management

2
Announcements

• Homework 2 due tomorrow by midnight.
• Stay tuned for homework 3.
• Class evaluations on Tuesday, 05.31.
• Need campus volunteer.
• Final exam on Thu, June 2.
• In class, closed books/notes.
• Project presentations on June 9th, 4-7pm.

3
Today

• Finish reliable multipoint e2e.
• Location management.

4
Location Management
5
Why is location management needed?

• In wired networks, hosts dont move.
• Constant association between host (id, address)
  and its location.
• In mobile wireless networks, hosts can move.
• Host id/address no longer provides location
  information.
• Need location tracking mechanism to deliver
  information destined to host.

6
Location for the Active Office Ward97

• Indoor sensor system that tracks location of
  people (active badge), equipment (equipment
  tags), etc.
• Requirements accurate (within 15cm), 3
  dimensions, scalable (number of objects
  locatable, area covered), cost.
• RF communication.

7
System components

• Transmitters attached to every locatable object.
• Matrix of receiver elements in all rooms where
  objects are to be tracked.
• Controller which polls one mobile object at a
  time.

8
Operation

• Periodically, mobile node is polled.
• Polled mobile broadcasts signal.
• Controller synchronizes receivers, who listen for
  some time to detect the peak of mobiles
  transmission.
• Controller polls receivers for the measured time
  interval between the sync signal and the signal
  peak (if any).

9
Distance computation

• Time measured by receiver composed of time to
  transmit the polling signal (from controller to
  mobile)time to transmit pulse (function of
  distance being calculated)processing time.
• Distance between mobile and receiver calculated.
• Empirically computed speed of sound in the room
  and service times.

10
Position calculation

• Triangulation using 4 receivers to determine a
  point in 3 dimensional space as estimate of
  position.
• In this particular set up, since all receivers
  are in the ceiling, only 3 distances required.
• Extra reported distances can be used for higher
  accuracy.

11
Evaluation

• Experiments with prototype show 95 of readings
  within 14cm accuracy. Even better accuracy for
  averaged readings.
• Addresses limit number of trackable objects.
• Large number of receivers and ultrasound nature
  of transmission from mobile proved to pay off
  regarding accuracy.
• Power savings mode minimizes maintenance.
• Low interference levels from office equipment.

12
Testbed

• Single floor (10500 sq. ft.) with 50 rooms.
• 3 base stations covering entire floor.
• Lucent WaveLAN RF technology.
• 2 Mbps.
• 1-2 ms one-way delay.
• 200m and 25m range (open/close environments).

13
RADAR Bahl et al.

• Similar to the Ward97 paper.
• Provide indoor location service.
• RF.
• Use received signal strength triangulation.
• Low cost.
• Off-the-shelf hardware.

14
Testbed

• Single floor (10500 sq. ft.) with 50 rooms.
• 3 base stations covering entire floor.
• Lucent WaveLAN RF technology.
• 2 Mbps.
• 1-2 ms one-way delay.
• 200m and 25m range (open/close environments).

15
Operation

• Off-line and real-time functions.
• Off-line derive and validate accurate signal
  propagation models.
• Real-time user location.

16
What is being collected?

• Signal strength (in dBm).
• s (Watts) 10log10(s/.001) (dBm)
• Signal-to-noise ratio (SNR) (in dB).
• SNR (dB) 10log 10 (s/n) (dB).
• For each received packet, SS and SNR are
  recorded.

17
Data collection process

• Mobile broadcasts beacons periodically.
• Base stations record SS and SNR.
• Different than the ORL system.
• Scalability?
• Path asymmetry.

18
More on data collection

• All clocks synchronized.
• Mobile broadcasts packets (4 pkt/sec).
• BS records (t, bs, ss).
• Off-line mobile also provides its location by
  using a floor map.
• Orientation is important (LoS, obstruction,
  etc.).
• In off-line phase, collected SS in all 4
  directions at 70 different floor locations.
• For each (x, y, d), 20 ss samples.
• d is direction N, S, E, W.

19
Processing data

• Off-line data used to build signal propagation
  model.
• Validation of assumption that from signal
  strength location can be inferred.
• How is location determined?
• Signal strengths from 3 BSs.
• Compare to floor layout/energy map.
• Pick location that minimizes (Euclidian) distance
  between measured and recorded set of SSs.

20
Results

• Empirical method performs better than random
  and strongest BS.
• Error approx. size of a room
• Taking k nearest neighbors shows some
  improvement.
• Analysis of impact orientation, number of data
  points, and number of samples.
• User tracking.

21
Radio propagation model

• Model of indoor signal propagation.
• No need for empirical data.
• Indoor propagation
• Reflection, diffraction, scattering.
• Multipath effect.
• Receiver gets signal from multiple paths.
• Distorted signal.
• Challenges dependency on layout, material,
  obstacles (number and type), etc.

22
Radio propagation model (contd)

• Adaptation of existing model to single floor.
• Consider effects of walls.
• Signal strength varies with distance AND number
  (and type) of obstacles.
• Empirical characterization of wall attenuation.
• Use (corrected) empirical data and linear
  regression to determine other parameters.
• Similar values for different BSs (location,
  surroundings, etc.)
• Less accurate results than empirical model, but
  more practical.

23
Localization in Sensor Networks Bulusu01
24
What are sensor networks?

• Large number of small, low-power devices
  (wirelessly) connected.
• Applications
• Monitoring, surveillance, tracking, etc.
• Typically ad-hoc deployable, unattended
  operation.
• Data-centric (instead of node-centric).

25
Localization

• Estimation of physical position (coordinates).
• Localization.
• Why is this important?
• Data usually identified by location (temperature
  of a given area, target tracking, signal
  processing applications).
• No a priori knowledge of location.
• GPS?

26
Approaches

• Multilateration nodes measure enough pair-wise
  distance estimates.
• Combination of radio (for time reference) and
  acoustic (time of flight for distance) signals.
• Proximity-based beacon nodes periodically
  broadcast position nearby nodes then estimate
  their position.
• Iterative multilateration beacon information
  propagated multi-hop.
• Beacon density sparse in some areas.

27
Self-configuring localized algorithms

• Adjust to current conditions (load, environment,
  etc).
• Localized algorithms distributed computation
  where communication is restricted to given
  neighborhood.
• Node density.
• Multiple modalities.
• Environmental adaptation.

28
Density

• Trade-off sparse vs. dense networks.
• Controlling density transmit power.
• Higher power makes networks more dense.
• Multiple power levels for tiered structure.
• Problem right balance between number of beacons
  (for coverage) and good localization.
• Power conservation.
• Interference.

29
Sensor modalities

• Use different modalities (acoustic sensors,
  cameras, etc) to overcome environmental
  unpredictability.
• Example acoustic sensors and acoustic/visual
  sensors.
• Acoustic sensing prefers LoS.
• Cameras can help by determining LoS sensors.

30
Adapting to the environment

• Not only to dynamics but also to fixed
  characteristics (e.g., obstructions, terrain,
  etc.).
• Example boundary beacon can extend its lifetime
  by cutting down its duty cycle.
• Example adapting to the dynamics of wireless
  channel using learning algorithms.