1
SDJS Efficient Statistics in Wireless Networks

• Albert Krohn, Michael Beigl, Sabin Wendhack
• TecO (Telecooperation Office)
• Institut für Telematik
• Universität Karlsruhe
• www.teco.edu

2
SDJS research and application area

• WSN (wireless sensor network)
• Battery powered
• Low computation capabilities
• MANET (mobile ad hoc networks)
• Fast changing environment
• Devices frequently join and leave a group
• BAN (body area network),
• PAN (personal area networks)
• Sensors attached to people
• Many small devices
• Ubiquitous and Pervasive Computing
• Settings with many devices (typically gt100)
• Battery powered
• Mid computation capabilities

3
SDJS Synchronous Distributed Jam Signalling

• What is SDJS?
• Method for ultra fast estimation of a parameter
  of a group of devices
• Novel transmission scheme
• Extension of standard wireless ad hoc protocol
• Synchronous, parallel, superimposing jam signals
• Works infrastructure less
• For highly mobile settings with high number of
  networked devices

Example for this talk How many devices are
present in the cell?
4
Related Work

• ExampleHow many devices are present in the
  cell?
• Budianu et al. 2003
• Collect IDs from the Devices and do a Good-Turing
  estimation, can be done iteratively
• Targeted on large scale networks, not on speed
• Also probabilistic
• Vogt 2002
• For passive RFID
• Using a slotted aloha protocol, where tags
  randomly select a slot
• Adaptive frame size
• Time to estimate 200 nodes with 99 reliability gt
  3 sec. (assuming ISO 18000 RFID standard)
• Normal ping on 802.11b
• Around 5 seconds (best case) for 100 stations

5
Motivation Idea of SDJS

• ExampleHow many devices are present in the
  cell?

• Novel
• Specific solution for collecting data of the same
  context
• Reduce redundant overhead
• Reduce transported information to necessary
  minimum
• SDJS include the physical layer
• Ultra Fast and efficient typ. 1000x faster
• Probabilistic, but adjustable accuracy/reliability
  (trade-off)

• Traditional
• Ping HELO, OLEH
• Slow, each node answers
• Packet implosion, collisions
• High bandwidth necessary
• deterministic
• Generic functionality of data transport in the
  network
• Same mechanisms for all information flow

6
SDJS Activity Flow

1. Station B starts SDJS
2. Each node prepares its transmission vector
3. SDJS scheme is processed
4. Each node has a reception vector

• Slotted (framed) Aloha
• Reduce Information to a single jam signal
• Full distributed operation
• Hardware Requirements?
• Network Requirements?
• Collisions?

7
SDJS The duck hunter problem

• ExampleHow many devices are present in the
  cell?

• Estimation of the real number from a given number
  of signals (the reception vector)
• Classical Duck Hunter Problem
• Solution surjective mapping, partition theory

How many hunterswere there?
Group of hunters
8
SDJS The Estimation 1

• Duck hunter problem analogon in SDJSs Slotsk
  Devices sending one jam signal eacha
  received jam signals
• gt P(ak) Distribution
• No a-priori informationMaximum Likelihood
• kMLEarg maxk P(ak)
• With a-priori informationMaximum a-posteriori
• kMAParg maxk P(ak) P(k)

9
SDJS The Estimation 2

• How is estimation done in practice?

Start count the number of received jam signals a

1. ML-Point estimationGive an estimationFor k
   (MLE)

2. MAP-Confidence intervalGive an interval,
kmin,kmax that contains the actual k with a
given confidence (e.g. 90)
In both cases look-up table that can be prepared
(no computation on nodes necessary)
10
SDJS Accuracy and Noise

• Noisefalse positives and detection errors
  duringcarrier sense affect theestimation

• Accuracy vs. Speed trade-offaccuracy depends on
  number of slots s

11
SDJS The Implementation

• TecOs particle computer
• Wireless sensor platform with 8Bit 20 Mhz
  processor
• 4kRAM, 4MBit Flash
• 125kbit/s wireless communication
• Customized ad hoc protocol
• Find a partner lt20ms
• Low power
• Low collisions
• Development tools
• Over 1000 produced, largedeveloper community all
  over the world

12
SDJS The Experiment

• Setting in an office with up to 50 particle
  computer

• Impressive prove of concepttheory and real
  world setting are nearly identical

13
SDJS Conclusion

• SDJS is
• An extension to wireless radio protocols
• Efficient group communication for very specific
  tasks
• Probabilistic by nature
• SDJS can
• Efficiently and fast estimate parameters (1000x
  faster)
• Achieve adjustable accuracy (speed accuracy
  trade off)
• Overall performance of SDJS depends severely on
  the underlying technology

14
SDJS Efficient Statistics in Wireless Networks
Thank you for your attention!

• Albert Krohn, Michael Beigl, Sabin Wendhack
• TecO (Telecooperation Office)
• Institut für Telematik
• Universität Karlsruhe
• www.teco.edu