projekt

User location estimation by means of WLAN

Carl-Friedrich-Gauss-Str. 247475 Kamp-LintfortGermany

Dennis Vredeveld

IMST GmbH

IMSTipos

2

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektContents

• Location-aware applications

• Indoor positioning: IMST ipos• Experimental testbed

• Deterministic and probabilistic approach

• Software design

• Summary

• Outlook

3

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektLocation-aware Applications

• Locate people and nearby resources

• Product location information

• Example: Metro Future Store, Rheinberg

• TrackingWho/what is close?

• NavigationHow do I get there?

4

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projekt

= AP

Experimental testbed

IMST 1st floor

Standard IEEE 802.11b

# of AP‘s 7

Dimensions 69 x 22 m

OS Linux/Windows

Hardware hp iPAQ

Lucent ORiNOCO

IMSTipos

5

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektExperimental testbed (2)

• Example for 1 AP

• Colour indicates measured average signal strength

Calibrated radio map

• Orientation matters

6

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektipos WLAN overview

• Determine signal strength from beacons received from „visible“ access points

• Average of k Nearest Neighbours in Signal Space (k-NNSS)

• Deterministic

• Data base contains averaged signal strength values

• Determine smallest Euclidean distance

• Probabilistic

• Data base contains histograms of signal strengths

• Determine largest joint probability

• Postprocessing (averaging, etc)

7

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektSimplified example

• 3 visible AP‘s

• Only 1 NNSS used

AP1 AP2 AP3 Room-40 -45 -100 101-43 -68 -87 102-55 -77 -85 103... ... ... ...-70 -33 -52 116-75 -53 -45 117-87 -56 -39 118-98 -67 -33 119

Table

Room 117!

AP1 AP2 AP3-75 -53 -39

Signal Strength

Measurements

WLAN IEEE 802.11bbeacon

8

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektDeterministic approach

• Data base with measured average values for each calibrated location: ss1, ss2,… ssn

2'

1

n

i ii

d ss ss

• Average k NN to obtain location estimation

• Mobile station with real time signal strength measurements from n Access Points: ss1’,… ssn’• Nearest neighbours are calibrated values with minimal Euclidean distance in signal space:

9

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektProbabilistic approach

• Data base with histograms instead of averages

1' '1 1

, ,, , n

n n

Count ss ssP ss ss ss ss

TotalCount

' ' '1 1

1

, ,n

n n i ii

P ss ss ss ss P ss ss

with

' 'i i iP ss ss Hist ss

-74 -73 -72 -71 -70 -69 -66 -65 -64-67-68

signal strength (dBm)

count

6868

Count ssHist

TotalCount

• Estimated joint probability:

• Assume AP‘s are independent:

10

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektipos system design

wireless techn.dependent

independent ofwireless techn.

Locationestimation

Display locationPowerpoint

GUI

Buildingmap HTML

Control

InterfacePostprocessing/

filtering

Floorplanchecker

Sliding windowaverage

Locationprofile

Kalmanfiltering

Trajectoryprediction

...

Data Distribution

WLANmodule

Preprocessing

Raw WLANdata

CE1CE1LCF1

LCF = Location CalculationFilter

Interface

New technologymodule

Raw ... data

Preprocessing

CE1CE1LCF1

Raw UWB data

Preprocessing

CE1CE1LCF1

UWBmodule

11

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projekt

Technology dependent

Technology independent

Location calculation filter

Location Estimation

Raw dataData Collection Raw data

Store3

Store1Store2

Preprocessing/Filtering

Raw data

A1 A3

A2‘A1‘

Postprocessing

Location Estimation

P1 P2 P3

P2‘P1‘

P4

Data Distribution

LocationEstimation

Dist3

Dist2Dist1

ipos data processing

12

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektExample: user tracking

13

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektSummary

• Use existing WLAN infrastructure

• Find best match of AP beacons‘ signal strength measurements to a calibrated radio map

• Realtime positioning with average error of 1-3 meter in testbed

14

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projektOutlook

• Identify and resolve error sources• Influence of environmental changes

(number of persons, etc)

• Differences between WLAN hardware

• Design and implement ipos-software

• Research into new data sources and calculation algorithms

15

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projekt

Questions?

16

Use

r lo

cati

on

est

imati

on

by m

ean

s of

WLA

N /

DV

projekt

Thank you for your

attention!

Vredeveld@imst.de

✉