Precise 3D Positioning Using Magnetic Field Bingh inghao ao Li Li - - PowerPoint PPT Presentation

Bingh inghao ao Li Li1, Kai ai Zhao hao2, S Ser erkan S n Say ayda dam2, Shuai X huai Xu3, A Andr ndrew D Dem emps pster1

1ACSER, S

Schoo hool of l of EE&T, UNSW, A Aus ustral alia ia

2Schoo

• ol of

l of Mining E ining Engine ngineering, g, U UNSW, Aus ustrali lia

3Key

ey Labor Laborat atory of

• f Minis

inistry of

• f Educ

ducat ation

• n for
• r Saf

afe e Mining, ining, Nor

• rthe

heastern Univ niver ersit ity, China hina

Precise 3D Positioning Using Magnetic Field

OUTLIN

INE

• Introd
• duc

ucti tion

• n
• Using G

ng Geom eomagn agnet etic F Fiel eld d for for 3D 3D Pos

• sitioni

ning ng

• Addi

dding ng Arti tificial al Magnet agnetic Fiel eld

• Conc
• ncludi

ding ng Rem emar arks

Usi sing ng a a Magn agnetic F Fiel eld d for

• r P

Posi

• sitioning
• Based on pre-deployed infrastructure

(J. Blankenbach and A. Norrdine, 2010)

• Based on signals-of-opportunity

(J. Chung et al. 2011) Indooratlas

Previ evious us S Study ( dy (1)

0.5 1 1.5 2 2.5 x 106

• 0.4
• 0.2

0.2 0.4 0.6 0.8 1 1.2 1.4 x y z norm 0.5 1 1.5 2 2.5 x 106

• 0.4
• 0.2

0.2 0.4 0.6 0.8 1 1.2 1.4 x y z norm

24 hours stability

Previous Study (2)

200 400 600 800 1000 1200 1400 0.2 0.4 0.6 0.8 X intensity L4 L3 L2 L1 LG 200 400 600 800 1000 1200 1400

• 0.8
• 0.6
• 0.4
• 0.2

Y intensity 200 400 600 800 1000 1200 1400 0.5 1 1.5 Z intensity Location index

Comparison of the measured magnetic intensities at overlapping corridors

Testing S ng Setup up

• A 45*39*76cm wooden cabinet
• A 30*40*60cm cuboid
• 12 layers (Z direction) and each layer has 48 grids, in

total 576 grids

The U he Uni niqu queness of

• f the M

he Magn agnetic F c Fiel eld d

• The average intensities of X, Y and Z directions of one

layer (Z=2)

• The change of the intensity is noticeable in all

directions

The V he Var ariat ation of

• f Magn

agnetic I c Int ntensities ( (1) 1)

Stat atic T Test est

• 20 test points were

randomly selected from 4 layers

• NN applied
• About 70%, the

positioning errors are lease than 2 units (10cm)

Dyn ynamic T c Test est

• The sensor was moved

alone X=3 and Y=3 in layer Z=6.

• Data were grouped every

10 samples as a test point

• Error is quite large
• Systematic offset

Addin ing A Artif ific icia ial M l Magnetic ic F Fie ield ld

• 10 magnetic blocks
• The magnetic force

line against each other to achieve a significant change of the magnetic field

The V he Var ariat ation of

• f Magn

agnetic I c Int ntensities ( (2) 2)

Stat atic T Test est ( (2) 2)

• Test points (in total

240) extracted from data base

• Applied NN and

4NN

• A higher possibility

position to the true location

• Error <5cm

NN:50% vs 86% 4NN:72% and 88% (85% <2.5cm).

Dynami mic c Test

• Results are better
• Systematic error in

X=3 testing test disappear

• Systematic error in

Y=3 testing is smaller

FUT

UTUR URE WORK RK

• The magnetic field positioning may find some specific

applications

• The long term stability of the magnetic field
• Integration of IMU with magnetic field positioning
• Increasing the fingerprint elements
• Best way to deploy the magnetic blocks

THA

HANK NKS

Questions?