1.

Introduction

Magnetic Position and Low-frequency quasi-static magnetic fields have

been used in numerous short-range position and/or
Orientation Tracking System orientation measurement applications. Most of these
techniques are based upon free-space field geometry,
but others utilize the effects of a nearby conducting
medium. A technique developed by Kalmus [1] uses
phase-quadrature excitation of a two-axis magnetic-
dipole source to determine the orientation of a two-
axis sensor. A null-finding technique [2, 3] has been
FREDERICK H. RAAB, Member, IEEE
used to locate a buried vertical magnetic dipole for
ERNEST B. BLOOD mine rescue and underground survey. The position
TERRY 0. STEINER and orientation of a buried magnetic dipole source
HERBERT R. JONES can also be determined from the signals received by
Polhemus Navigation Sciences, Inc. four or five three-axis sensors located at different
positions on the surface [4, 5]. An electromagnetic
channel guidance concept [61 uses a two-axis magnetic
dipole source located on the Earth's surface. Several
other techniques for the guidance of surface and sub-
surface vehicles using the quasi-static fields surround-
Abstract
ing long wires have also been devised [7, 8].
Three-axis generation and sensing of quasi-static magnetic-
dipole fields provide information sufficient to determine both the field coupling
3-AXIS 3AI
position and orientation of the sensor relative to the source. 1<,_MAGNETIC MAGNETIC=
lvSOUJRCE ' SENSOR
Linear rotation transformations based upon the previous
measurements are applied to both the source excitation and
sensor output vectors, yielding quantities that are linearly pro-
portional to small changes in the position and orientation.
CICIS CUT

Changes are separated using linear combinations of sensor out- POSITION AND ORIENTATION
put vectors, transformed to the desired coordinate frame, and MEASUREMENTS
Fig. 1. System block diagram.
used to update the previous measurements. Practical considera-
tions for a head-tracking application are discussed.
The system discussed here [9-12] used both a three-
axis magnetic-dipole source and a three-axis magnetic
sensor, along with the related electronic circuitry
shown in Fig. 1. The excitation of the source and the
resultant sensor output are represented as vectors; all
source axes are to be excited simultaneously with
signals of the same frequency and phase so that an
equivalent single-axis source dipole of arbitrary orien-
tation is produced. The source excitation pattern is
composed of three' sequential excitation states, each
of which produces an excitation vector that is linearly
independent from the other two. The resultant set of
three sensor output vectors contains information suffi-
cient to determine both the position and orientation
of the sensor (6 degrees of freedom) relative to the
source. Operation of the system is synchronous and
under computer control.
This system tracks the position and orientation of
Manuscript received December 10, 1978; revised January 15, 1979, the sensor by determining small changes in the coordi-
March 22, 1979, and April 10, 1979. nates and then updating the previous measurements.
Authors' address: Polhemus Navigation Sciences, Inc., A Sub- To accomplish this, the previous measurement is used
sidiary of The Austin Company, P.O. Box 298, Essex Junction, Vt to compute linear transformations that are approxi-
05452.
0018-9251/79/0900-0709 $00.75 1979 IEEE 'Variations are discussed in Section Vof this paper.

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. AES-15, NO. 5 SEPTEMBER 1979 709
 TABLE
Coordinate Frame Definitions

SuLbscript Nanmc Det'initioni

0 Tracking fo = T- T- f

I Source Reference for other definitions.

2 Position .f =T T'aIf.
3 Sensor-position f = eC SSf2
4 Zero-orientation f4 = T-a T f3
sensor
fly
fiz 5 Sensor-output f, = TI To Tw f

DYt Zt
6

7
Derotated sensor

Restored tracking
f

f
=

=
T , T a T f

T^ T- f

Fig. 2. Position and orientation coordinates. Notes:

1) a,, (3, Q, tJ,, 8,, 4, represent true values.
a,, /1, Q, tp,, 0,, +, represent previous estimates.
mately the inverses of those describing the true 2) Symbols with a wavy underline corresponid to boldface symbols
source-to-sensor coupling. One set of such transfor-
in the text.
mations is applied immediately prior to actual source
excitation, while the other is applied to the sensor out-
put vectors. The resultant transfer function is approxi-
mately that of a similarly aligned source and sensor. measured in the source coordinate frame can be
The set of processed output vectors differs from the converted to any desired coordinate frame mathe-
specified source excitation vectors (prior to transfor- matically.
mation) by quantities that are linearly proportional to The sensor position is specified in rectangular
the changes in position and orientation. Individual (x1, y,, z1) or spherical (a,, j1, Q) coordinates defined
coordinate changes are extracted from linear combina- relative to the source coordinate frame. Sensor orien-
tions of the elements of the processed output vectors tation is specified by a sequence of three rotations.
and subsequently transformed to the desired measure- Azimuthal rotation by tp, first turns the sensor about
ment coordinate frame. its Z axis from + X + Y. The elevation rotation by 01
The required transformations, detection equations, then turns the sensor about its Y1 axis from + X to
and coordinate converstions will be derived subse- - Z. Finally, a roll rotation by +, turns the sensor
quently. However, it is first necessary to formulate about its X axis from + Y to + Z. Note that in the
the field coupling in matrix form and to establish a zero-orientation condition, the three sensor axes are
set of coordinate frames to facilitate the analysis. parallel to the corresponding source axes, and that the
order of the rotations cannot be interchanged without
II. Magnetic-Field Coupling changing the values of tp, &0, and C1. The subscript
"I" denotes a coordinate referred to the source
The first step in synthesizing a position-and- coordinate frame. Other numerical subscripts denote
orientation-finding algorithm is the definition of co- references to other coordinate frames (Table I) that
ordinate frames and a vector-matrix formulation are used in subsequent analyses.
relating sensor output to source excitation. Applications of this system use source and sensor
loops whose diameters are small in comparison to the
Geometry distance separating them. They can therefore be
regarded as point sources and point sensors. The
The geometric relationship between the three-axis wavelength of the excitation signal is then long
source and the three-axis sensor is shown in Fig. 2. enough (Q<<A/2n) that only the near-field or quasi-
The source coordinate frame X1 - Y1 - Z1 defined by static field component (i.e., the term varying as the
the axes of source is used here as the reference for the inverse-cube of distance) is significant.
final measurements. However, the source axes can be Suppose now that the loop shown in Fig. 3 is ex-
effectively aligned with any desired coordinate frame cited with a current i(t) = I cos cot. The magnetic
by altering the excitation. Similarly, coordinates field produced at a distance Q and off-axis angle t is
710 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. AES-15, NO. 5 SEPTEMBER 1979
 Fig. 4. Coupling between aligned source and sensor. Source,
tracking, and sensor frames aligned.

TABLE 11
Orthogonal Rotation Matrices
Position Orientation
cos a sin a O cos + sin q) O
Azimuth
rotates Ta = -sin a cos a O T -sin 4 cos * O
X into Y
O 0 1 O 0 1

cos a 0 -sin cos e O -sin

Fig. 3. Magnetic-dipole field. Elevation
rotates TSa = o 1 0 ie = o 1 0
X into -z
Lsin 0O cos aj sin e 0 cos
described completely [13] by radial and tangential
components
1 0 0 1 0 0
HQ = (M/2nQ3) cos ( ( Roll
rotates
Y into Z
Ty = O cos y sin y
_T, = 0 cos sin
[O -sin y cos y -sin 0 cos
and
Notes:
l)T T T= T_ , T T:0 T0T , To, T0 + etc.
H, = [M/4nQ3J sin ( (2)
2) Position roll is used only as a dummy variable and does not
where M = NIA is called the magnetic moment of the define sensor position or orientation.
loop and A and N represent the area and number of 3) Symbols with a wavy underline correspond to boldface
turns of the loop. A temporal variation of exp(jct) is symbols in the text.
implied; the spatial phase variation of exp(JQ/A) can amplitudes differ by a factor of - from the X2, -X3
be neglected in this application. This characterization coupling which is produced by a radial field compo-
is valid for a loop of any shape as long as the distance nent. The coupling can be described completely in
is more than about four times the radius of the loop. vector-matrix form by
Note that there is no frequency dependence in the
amplitudes of the fields.
Source-Sensor Coupling 1 0 0]

The excitation of a three-axis magnetic-dipole f3 = (C/I3)Sf2 = (C/Q3) 0 - ½/2 0 f2 (3)

source and the resultant three-axis sensor output are
most conveniently described in vector notation. The 0 0 -½2
excitation of the source is therefore represented by
fi = [fix, fiy, fil]T. The number of turns and area of
the three source loops are assumed to be identical; where C = NAG/2n incorporates sensor gain G and
hence fi, fly, and f1. represent the amplitudes of the the common source factor NA/2n.
currents exciting the loops of Xi axis, Y1 axis, and Z1
axis orientation, respectively. Effects of Position and Orientation
Now let the output of a three-axis sensor be
similarly represented by f3 [f3t, f3y, f3jTP and con-
= The coupling between a source and sensor of arbi-
sider the coupling between that sensor and a similarly trary position and orientation can be determined by
aligned sourcef2. As depicted in Fig. 4, each source inserting orthogonal rotation matrices [14] into (3).
axis is coupled only to the corresponding sensor axis. These matrices are based upon position azimuth and
Furthermore, since the Y2 Y3 and Z2 Z3 couplings
- - elevation (a1 and P() and orientation azimuth, evalua-
are produced by tangential field components, their tion, and roll (lpo, +,, and 0), as shown in Table II.
RAAB ET. AL.: MAGNETIC POSITION AND ORIENTATION TRACKING SYSTEM 711
 TRUE
/
POSITION

PREVIOUSLY
.11, - ESTIMATED
PREVIOUSLY ORIENTATION
ESTIMATED
POSITION

.o(s2)

YOvo / Z o; -" o- TO
SOURCE

Fig. 5. True and previously measured position. Fig. 6. True and previously measured orientation.

Note that the subscript defines both the type of of the sensor. This is accomplished by using the
transformation and its independent variable. previous measurements of position and orientation to
Consider first the coupling between the source and compute rotation matrices that approximately undo
a zero-orientation sensor (whose output is f.) located the effects of the (large-angle) sensor position and
at (a1, (1, Q), as shown in Fig. 2. The excitationf2 of orientation.
an equivalent source whose X-axis is aligned with the
line connecting the source and sensor can be deter-
mined by rotating the excitation vector of the real Use of Previous Measurements
source by position azimuth and elevation, thus
The relationship of the true sensor position
f2 = T51 Ta,fl. (4) (a1, Q3,e) to the previously estimated sensor position
(&,9) (31, is illustrated in Fig. 5; note that both posi-
The coupling to a similarly aligned equivalent sensor tions are defined in source-frame coordinates. The
f, then has the same form as (3), i.e., f3 = (C/Q3) S differences in the source-frame position angles can be
f2 The output of the zero-orientation sensor is then represented by Aa1 and AP1; hence a1 = a, + Aa1 and
found by applying inverse position rotations, thus 31 = (31 + AP1. The difference in the position angles
can also be defined by angles Aa0 and AP,0, which are
f4 = (C/23)T al T_0iSTI Ta fi = (CIQ3)Qfl. (5) referenced to the tracking coordinate frame XO, - Y,
- ZO. The tracking frame is defined by rotations
The equivalent sources and sensors used above are from the source coordinate frame X1 - Y, - Z1 by
listed in Table 1. previously estimated position azimuth a1 and elevation
The output of the three-axis sensor of arbitrary (31 Differences between estimated and true orientation
orientation (wP1, 01, +1) is determined by applying angles can similarly be defined relative to either the
orientation azimuth, elevation, and roll rotations to source-frame or tracking frame, as shown in Fig. 6.
the output of the equivalent zero-orientation sensor, Now consider an equivalent source excitation
thus vector f, specified in the tracking frame. This vector is
converted to the true source excitation vector f1 by in-
verse rotations by estimated elevation and azimuth;
f5 T+, To, T,f4
= = (C/Q3)A f4 = (C/Q3)A Qfi. (6) thus

These expressions will be used subsequently to derive fi = T , T fo. (7)

the position-and-orientation-finding algorithm.
Combining this with (6) and (5) yields
111. Determination of Position and Orientation
Changes f5 = (C/Q3) A T T -,S(T,, T., T -, TT _)fl

The matrix multiplications in (6) imply source-to- = (C/Q3)ATA TTlSAPfO. (8)

sensor coupling equations that contain multiple
trigometric products and sums. These nonlinear If there has been no movement of the sensor since the
coupling equations can be reduced to linearized equa- previous estimate of position was made, a1 = a1 and
tions in small changes in the position and orientation 31 = P3, hence T,, TQ, Ta., T_= I. It is therefore
=

712 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. AES-15, NO. 5 SEPTEMBER 1979
necessary for that sequence of rotations to reduce to a The above transformations produce a differential
differential rotation AP when the changes in position rotation matrix which must reduce to (AP)-' in the
are small. absence of orientation errors; hence
The nature of the differential rotation matrix AP
can be determined in two ways. Direct reduction of (Tp,TX ,A-1)(AT al,T,) - (A-'A)(AP)-' = AA(AP)-l (12)
(8) yields the source-frame formulation
Rearrangement of the above yields
AP =T,T (TT,)
(AT aT -M)[AA(AP-)] =VAT T -p (13)
TpTaT,(T , T^, T 0T ',TA'3)
hence the differential orientation rotation AA is based
=pT'3Ta,T '3TAP, upon tracking-frame orientation angles (Fig. 6) and

Aa, cos P, 1 Alp - A00

= - Aa, cos P, -
Aa, sin P, (9) AA = TA+OTASOT&Wo = - A4o 1 A+o . (14)

A(3, Aa, sin flj 1 A/o -A+0 1

However, it is more convenient to equate the true Substitution of (14) and (8) into (10) now produces
position transformation Tp, Ta, to the estimated
position transformation followed (premultiplied) by a f7 = (C/Q3)"A(AP)- SAPfo = (C/Q3)Rfo (15)
sequence of differential rotations, thus
where
AP=[TpO, T,]TI-., T_p

E(TAyoTvoTad)Tp Ta.] T-.,T-Pi 1 2Aa0 -2^&°

I
R i A . (16)
1 Aa1ro -Ago 1A+o I

Aao I AyO (10)

TAyoTAPoTA.o
-

The preexcitation, true position and orientation,

Ago -A iYo I
field-coupling, and post-sensing transformation se-
quence is shown in Fig. 7. Note that all large-angle
effects have been reduced to small-angle effects and
Note that the differential angles Aa0, AP,0, and Ay. are replaced by small-angle approximations; position roll
referenced to the tracking coordinate frame. The in- (Ayo) has disappeared. The transfer function between a
clusion of a roll rotation (TAY) is necessary to allow tracking-frame vector f0 and a restored tracking-frame
for the misalignment of coordinate axes when the vector f, is therefore linear in small changes in position
same position is obtained by two different sequences and orientation. Furthermore, the changes are readily
of rotations. However, position roll is not an inde- identifiable in tracking-frame coordinates.
pendent variable, and its effects are subsequently
shown to be cancelled and not directly observable.
The application of preexcitation transformations
therefore reduces the large-angle transformations on Excitation States
the right side of S in (8) to small-angle transfor-
mations. The large-angle transformation on the left Determination of position and orientation requires
side of S can similarly be reduced to small-angle sensor outputs corresponding to three linearly in-
transformations by applying inverse transformations dependent source-excitation vectors. The source and
(based upon estimated position and orientation) to the sensor vectors produced during different excitation
sensor output vector f5. To do this, a restored track- states are denoted by an attached parenthesized
ing frame equivalent output vector f7 is defined by excitation-state symbol. The three source excitation
vectors used here correspond to the tracking-frame
f7= 7jTj.A-'f5 = T jT -IT 5T -f. (1 1) axes:

RAAB ET. AL.: MAGNETIC POSITION AND ORIENTATION TRACKING SYSTEM 713
 A

to o
ft~~~~~~~~~f
p

OUTPUT

TRUE POSTI

MEASRED POSTION JAND ORIENTATION a, I,. ;,e. ;

Fig. 7. Processing sequence.

2
3(/CO[2f7.(S2) - f(S1)] (22)
0 0

I Apo =2 /Q17z(SI) - 2f7.(S3)] (23)

fo(S1) = 0 ,fo(S2) 1I and f0(S3) 0 (17)
Atp0 = 2('3/CC)[f7(S2) - f7y(S1)J
= =

(24)
0 0
A0o = 2(Q3/Cf7z(SI) - f7.(S3)] (25)

A set of three excitation vectors constitutes an excita- A+o= 2( 3/CWf7x(S2) = - 2(Q 3/C)f7,(S3). (26)
tion pattern. The use of other excitation patterns con-
taining more excitation vectors is readily related to the Linear combinations of elements can be used to
use of the above three-state pattern by linear transfor- minimize the effects of noise upon the range and roll
mation. estimates. This operation is represented by the "error
The restored tracking-frame sensor response vectors detection" block in Fig. 7.
corresponding to the tracking-frame excitation vectors of
(17) correspond to the columns of matrix R in (16); i.e., Error-Angle Coordinate-Frame Conversion
'

f7(Sl) = (C/Q3) l 1, ~ A02 3 + While the range estimate is universal to all coor-
dinate frames, the error angles (Aao, Ap0, Atpo, A00,
(18) A+0) are referenced to the tracking frame. Before they
can be used to update the previous source-frame
[3,aCo 11 IT
f7(S2) = (C/Q3) _
p,& -2' 2+A° measurements of position and orientation, they must
(19) be converted to source-frame error angles (Aa,, AP1,
Atp, A0,, A+,). This operation follows the error-
f7(S3) = (C/e3) [-_ AP +20°, 2^+o _1 ]T- detection block in Fig. 7.
The method of conversion of the position error
(20) angles Aa0 and Ap0 to Aa1 and AP, is most readily
derived by equating AP as given by (9) and (10). If
The first unknown to be estima Lted is distance, since it
is given explicitly by the amplitudes of the constant terms
lp1 n/2, this produces
in (18), (19), and (20): Aa, = Aa./cos p1 (27)
3 3
Q = \ C/f7(Sl) = C/2f,y(S2j and
3
=-\/ - C/2f71(S3) (21)) AP1 AP,-
= (28)
The factor C/l3 is then computed,2, and the five error Modified conversion formulas must be used when
angles are determined from linear combinations of the PI-± i/2.
nine quantities in (18), (19), and (20): The equations that convert tracking-frame orienta-
tion errors to source-frame orientation errors can
similarly be derived by substituting tp = Aj1 + Awl, 01
AA
= 01 + A0,, and 1 = +1 + A+, into (12) and (13),
21If the constant C is unknown, an invalid range estimate will be
determined; however, determination of the error angles is unaffected. producing
714 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. AES-15, NO. 5 SEPTEMBER 1979
T$+A+1 T- +AO T,,1±A,, T al T A
+40
ft .- ---

= TT T,el T 1 T)(TAlo)TAo0TA4°).
T (29) AL
--

F- oee k~ ~ ~ ~ ~ Le.-

Note that in (29) position angle estimates have been aB.

L00
set equal to true position angles, since the differential s30' .0

sF50*~ 0-
position angles do not affect the differential orienta- 0

tion angles (A+1 Aa- 0, etc.). However, the tedious ti 0

_
0

4 0-00640
reduction required can be avoided by observing [14] ------
00 .06.89.888- -swelles-9 , -L
- 5 -- 8 I mmsmmmvv)

that a differential angle (angle rate) is transformed

from one coordinate frame to another as is a co- 0. 5~~~~~~~~~~~
ordinate of the axis about which the differential angle i0o49.0'0.8--Ssms)k
produces rotation. To visualize this, consider that
both Aa and z are invariant under azimuthal rotation 001-1--o-- ----
----I I t'
o
L
k
T. (excitation cyches)
r,
Premultiplying (29) by T4-, T ;, and some re- _50*
J

arrangement yields
T aITA+,IoTaI,TAO)TAJIIT al F, Fig. 8. Responses to large-angle errors. Solid lines indicate
true values (inputs) and dots indicate measured values
= TV, -T ,, T.OTA0OTAO4.0 (30) (outputs) .

Vectors of differential angles are now formed by plac-

ing the angle errors in the elements corresponding to tions, the magnitudes of the measured errors are in-
their axes of rotation (e.g., [A+0, A00, AWIo]T). Since correct if the true angular errors are large. However,
differential angles are subjected to only those the polarities (signs) of the errors are correct, and this
transformations occurring to their left in (30) and allows the system to achieve proper tracking within a
since A+1 is subjected to a rotation by -01, while AO, few iterations after the occurrence of a large angular
and Ap1 are subjected to rotations by both -_, and error (Fig. 8). The system contains one inherent 1800
+ 01, the differential-angle conversion equation has the ambiguity3 that is in general resolved through "com-
form mon sense" about realistic positions and orientations.

A+1 COS e0 IV. Application Considerations

A01 Magnetic tracking systems have been used in a

number of helmet-mounted sight (HMS) applications
Aw1 -A+1 sin Gi to measure the orientation of the helmet and conse-
quently its wearer's line-of-sight. An HMS system
can, for example, enable the pilot of a fighter aircraft
to transfer a visually acquired target to an electronic
tracking system. The line-of-sight to a target acquired
by an electronic system or one crew member can be
=T_ 0° + AO, T=T ,,,T1 0. (31) transferred to another crew member's vision by a cue-
ing display that indicates the direction in which to
look. By slaving a range-measuring radar to the line-
of-sight specified by an HMS, the relative position of
a visually acquired target can be determined. This in-
formation can also be used to facilitate search-and-
Updating and Acquisition rescue operations by combining it with true position
(measured by a navigation system) to map
The source-frame position and orientation error automatically the area that has been searched. The
angles are now used to update the previous measure- HMS application illustrates several practical con-
ments by simple addition; i.e., 'a + Aa1--a, etc. In- siderations relating to the magnetic tracking system.
itially (or after a loss of power or signal), the tracking
system must acquire approximately correct measure- 3For example, the signals for a, (3P = tp, , +, = O are
ments in order to track properly. Since the transfor- indistinguishable from those for P3 0, = +, = O and a, -p,
= =

mations used are based upon small-angle approxima- 1800.

RAAB ET. AL.: MAGNETIC POSITION AND ORIENTATION TRACKING SYSTEM 715
Configuration (due to cross coupling between axes). The elements of
the sensor sensitivity matrix can be determined by
In an HMS application, the three-axis sensor is measuring its responses (on all three axes) to three
mounted on the helmet and the three-axis source is orthogonal fields produced by a Helmholtz coil. The
mounted above and behind the nominal head position elements of the source matrix can similarly be
on a nonmetallic structure such as the aircraft canopy. measured by exciting its three axes in each of three
The helmet is equipped with a collimated (focused-at- orientations and measuring the responses in the
infinity) reticle that is used to sight the target and Helmholtz coil. Inclusion of the inverses of these
optionally equipped with a cueing display. Sensor matrices in the computations then removes the
mounting is noncritical because the system is initial- measured imperfections. This can be done either in
ized by "bore sighting" in which the system derives a the hardware by cross-coupling in the source-driving
rotation sequence that aligns the sensor with the air- circuitry or in the software by insertion of the inverse
craft reference frame. The source and the sensor are sensitivity matrix before A` in Fig. 7.
typically separated by a range of a few centimeters to
one meter. Nearby Metallic Structure
The sensor is connected to the system control
panel through a cable roughly three meters in length. A magnetic system is often required to operate in
There the three sensor outputs are preamplified, an environment containing metallic structures that
multiplexed, and transmitted to the system electronics produce scattering (distortion) of its magnetic-dipole
unit. The multiplexed signals are then synchronously fields, hence significant measurement errors. Ferro-
demodulated and digitized by a 12-bit analog-to- magnetic material can alter the shape of any magnetic
digital converter. The source of driving signals are field. The ac magnetic fields used in the HMS
generated in the system electronics unit; their instan- system produce circulating (eddy) currents in nearby
taneous amplitudes are produced by 12-bit multiplying conductive material thereby generating secondary ac
digital-to-analog converters controlled by the com- magnetic fields and distorting the field pattern.
puter. Stepped automatic gain control is included in The complexity of the metallic structure in general
both the source and sensor circuits to reduce dynamic precludes analytical prediction of the scattered fields
range requirements. (and makes numerical prediction difficult). A useful
Carrier frequencies are typically in the 7- to limit can be obtained by assuming that an infinite
14-kHz range, and the choice of a carrier frequency conducting plane produces as much or more field
represents a compromise between equipment com- distortion as would any object at the same distances
plexity, sensitivity, noise, and source inductance. The from the source and sensor. By using simple image
excitation pattern and associated processing leading to theory, the conducting plane can be replaced with an
updated measurements are repeated typically at 30- to image source; the ratio of the free-space and scattered
120-Hz rates. These rates are limited by computation fields is then estimated to be the ratio of the inverse
time, source inductance (response time), noise, and cubes of the distances from the sensor to the source
allowable source moment. and to its image. This produces a rule of thumb
Since the sensor must be both small and stating that an object whose distance from the source
lightweight, yet must deliver a signal sufficient to is at least twice the distance separating the source and
allow (in this application) 12-bit accuracy (1 part in sensor produces a scattered field whose magnitude is 1
4096), it generally imposes the most severe design con- percent or less of the magnitude of the desired field.
straints upon the system. Both the sensor and the Metallic structure that is fixed in position and
source are wound on ferrite cores to increase their orientation with respect to the source produces field
effective areas (sensitivity) by factors of 2 to 3.5, de- distortion that is constant at any given position. Com-
pending on the particular core shape. In the example pensation for the field distortion is therefore possible
HMS application, the sensor is roughly 1.3 cm x 1.9 as long as the fields at each allowed position are
cm x 2.6 cm, while the source is roughly 2.5 cm on unique and span three-dimensional vector space (i.e.,
each side. The source magnetic moment is approx- the field vectors produced by the three source excita-
imately 0.086 At . m2 when driven by a peak current tion states are linearly independent). Corrections can
of 200 mA. The sensor produces an output of approx- take the form of additive vectors or a sequence of
imately 0.12 V/(At/m). rotations, and can be stored in either a look-up table
or as polynomials in the position parameters.
Source and Sensor Imperfections
The imperfections in the three-axis source and Variations
sensor can be described by 3 x 3 fixed-valued
matrices that include both rotation (due to physical Three other multiaxis magnetic position and orien-
mounting of the source or sensor), gain, and skewing tation measurement concepts provide flexibility for
716 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. AES-15, NO. 5 SEPTEMBER 1979
varied applications. The first uses a fixed source- Trans. Geosci. Electron., vol. GE-Il, pp. 178-185, Oct.
frame excitation pattern. The preexcitation and post- 1973.
sensing rotations described here are then applied to a [4] T.W.H. Caffey, "Locating a buried Earth penetrator," San-
dia Laboratories, Albuquerque, N. Mex., Rep.
3 x 3 matrix formed by assembling the true sensor SAND77-0646, Nov. 1977.
output vectors, producing a 3 x 3 matrix of the sen- [5] T.W.H. Caffey, "Locating a buried magnetic dipole," Proc.
sor output vectors of the tracking system. The second Electromagnetic Guided Waves Workshop, Boulder, Colo.,
concept [15] uses squared vector magnitudes and dot Mar. 28-30, 1978.
products to determine position and orientation [6] G.S. Barta and R.G. Olsen, "A short-range VLF navigation
system for rivers and harbors," IEEE Antenna and Propaga-
through direct, closed-form computations. The third tion Soc. Internatl. Symp. 1977. Stanford, Calif., pp.
concept [161 uses two three-axis ferrite antennas alter- 124-127, June 20-22, 1977.
nately as sources and sensors, thereby enabling two [7] J.C. Coyne, F.J. Elia, and H. Southworth, Jr., "Location
separate units to determine their relative positions and detection and guidance systems for burrowing device," U.S.
orientations. Patent 3 529 682, Sept. 22, 1970.
[8] T. Gullich, "Indicator," U.S. Patent 3 777 304, Dec. 4,
1973.
Acknowledgment [9] A.R. DeRuyck and J. Kuipers, "An electromagnetic position
sensor," Polhemus Navigations Sciences, Inc., Burlington,
This paper describes a position and orientation Vt., Tech. Rep. AFAL-TR-73-281, Nov. 30, 1973.
measurement concept invented by J. Kuipers [10-12] [10] J. Kuipers, "Object tracking and orientation determination
means, system, and process," U.S. Patent 3 868 565, Feb.
and provides a detailed analysis of the concept using 25, 1975.
the inventor's system model illustrated in Fig. 7. A.R. [II] J. Kuipers, "Tracking and determining orientation of object
DeRuyck and A.G. Rodgers played significant roles in using coordinate transformation means, system and
the development and testing of related hardware. process," U.S. Patent 3 983 474, Sept. 26, 1976.
[12] J. Kuipers, "Apparatus for generating a nutating elec-
tromagnetic field," U.S. Patent 4 107 858, Apr. 12, 1977.
[13] M.R. Kraichman, Handbook of Electromagnetic Propaga-
References tion in Conducting Media, 2nd ed. Washington, D.C.; Head-
quarters Naval Material Command, 1976. (USGPO stock no.
[I] H.P. Kalmus, "A new guiding and tracking system," IRE 008-040-00074-5.)
Trans. Aerosp. Navig. Electron., vol. 9, pp. 7-10, Mar. [14] R.L. Pio, "Euler angle transformation," IEEE Trans.
1962. Automat. Contr., vol. AC-Il, pp. 707-715, Oct. 1966.
[21 J.H. Murphy and H.E. Parkinson, "Underground mine [15] F.H. Raab, "Remote object position locator," U.S. Patent
communications," Proc. IEEE, vol. 66, pp. 26-50, Jan. 4 0054 881, Oct. 18, 1977.
1978. [16] J. Kuipers, "TWOWAY, a position and orientation
[3] R.G. Olsen and A.J. Farstad, "Electromagnetic direction measurement system," Proc. MTS IEE OCEANS '77,
finding experiments for location of trapped miners," IEEE pp. 38E-1 to 38E-13, La Jolla, Calif., Oct. 1977.

Frederick H. Raab (S'66-M'72) received the B.S., M.S., and Ph.D. degrees in
electrical engineering from Iowa State University, Ames, in 1968, 1970, and 1972,
respectively.
He then worked for Cincinnati Electronics Corporation for 31/2 years in the
areas of high-efficiency power amplifiers and search-and-rescue navigation. Since
joining Polhemus Navigational Sciences, Inc.,, in 1975, he has been involved with
terrestrial applications of Loran-C and the development of an electromagnetic
system for locating trapped mine workers. His professional interests include com-
munications, navigation, signal processing, and power amplifiers.
Dr. Raab is a member of the Institute of Navigation, the International Omega
and Wild Goose Associations, Eta Kappu Nu, and Sigma Xi.
RAAB ET. AL.: MAGNETIC POSITION AND ORIENTATION TRACKING SYSTEM 717
 Ernest B. Blood studied physics at Lowell Technology Institute, received the
B.S.E. degree in aerospace engineering from the University of Michigan in 1966,
and did graduate work there in aerodynamics.
He was then employed by Lockheed Aircraft where he participated in simula-
tion, modeling, and prediction of aerodynamic and performance characteristics
of various aircraft. He joined Polhemus Navigation Sciences, Inc., in 1970 and
has participated in numerous radio navigation studies of the Loran-C, Decca,
Omega, and Transit Systems. As head of the Computer Sciences Group, he is
responsible for the design of the software and processors used in the SPASYN
systems.

Terry 0. Steiner received the B.S. degree in physics from Case Institute of
Technology, Cleveland, Ohio, in 1970 and is currently doing graduate work in
physics at the University of Vermont.
He served in the U.S. Air Force for 4 years where he was involved in RF en-
vironmental studies and in establishing personnel exposure criteria. Since joining
Polhemus Navigation Sciences, Inc. in 1975, he has been involved in the design
of SPASYN three-axis source and sensor arrays.
Herbert R. Jones received the B.S.E.E. degree and (while a National Science
Foundation scholar) the M.S.E.E. degree from the University of Vermont in 1968
and 1972 respectively. He was a Ph.D. candidate at the University of Vermont
where he tutored courses in electric and magnetic fields, solid-state physics, and
computer logic.
Since joining Polhemus Navigation Sciences, Inc., in October 1977, he has
been charged with the design of 16-bit processor-to-resolver, serial-digital inter-
faces, and high-precision analog signal conditioning circuitry for a number of
SPASYN system variations. In addition to his circuit design tasks, he was project
engineer/program manager on a number of SPASYN system installations and has
been heavily involved in developing SPASYN calibration and alignment pro-
cedures.
718 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. AES-15, NO. 5 SEPTEMBER 1979