Strap-Down Inertial Navigation Systems (SDINS)

Error Propagation Analysis

S.Senthil Kumar
Data Patterns (India) Pvt td
!hennai" INDIA #$%&$%
& Introdu'tion
The purpose of the inertial navigation systems
(INS) is to provide the position, velocity and attitude of the
vehicle through self-contained sensors in a convenient
coordinate system. The position, velocity and attitude
information in INS, however, is not exact due to many
error sources. The error growth rate is one of the most
important factors that affects the performance of INS. or
this reason, a considera!le amount of the wor" has to !e
performed in the field of INS concentrates on the error
analysis. The sources of errors are categori#ed as follows$
inertial sensor errors (inaccuracies) % including
due to accelerations and vi!rations&
initial alignment errors&
computational errors&
'ny lac" of precision in a measurement used in
INS is passed from one estimate to the next with the
overall uncertainty in the precision of the calculated
(uantity varying or drifting with time. In general, INS
performance is characteri#ed !y a growth in the navigation
error from the position co-ordinate values which are
initially assigned to it. It is common practice to refer to an
inertial navigation system in terms of its mean drift
performance& a one nautical mile per hour system is a
typical performance class of a system. This would !e
typical of an inertial navigation system used in a
commercial aircraft.
' com!ination of analysis and simulation
techni(ues are commonly used to predict system
performance. )ere, e(uations are given relating system
performance to sources of error and illustrates the
propagation of the various types of error with time.
( SDINS Error Sour'es
(.& Inertial Sensor Errors
The performance characteristics of inertial sensors
(either gyros or accelerometers) are affected !y a variety of
errors. *ost errors can !e categori#ed into sensor !ias,
scale factor, acceleration dependent drifts, axes
misalignment, and noise.
The errors in the measurements of angular rate
provided !y a set of gyroscopes whose
sensitive axes are orthogonal, may !e expressed
mathematically as shown !elow$
(+.,)
where
are the accelerations acting along the
principle axes of the host vehicle,
are the applied angular rates acting a!out
these same axes,
is a three element vector representing the residual
fixed !iases which are present,
is a - x - matrix representing the g-dependent
!ias coefficients,
is a - x - matrix representing the anisoelastic
coefficients,
is a - x - matrix representing the anisoinertia
coefficients,
is a diagonal matrix representing the scale-factor
errors,
is a - x - s"ew symmetric matrix representing
the mounting misalignments and cross-coupling
terms,
is a three element vector representing the in-run
random !ias errors.
The gyro errors are summari#ed in Ta!le +., for different
( )
x y z

x x y z y z
y G g y ae z x ai z x
z z x y x y
x x
G y G y G
z z
a a a
B B a B a a B
a a a
S M w






_ _
_ _

+ + +




, ,
, ,
_ _

+ + +


, ,
, ,
x y z
a a a
, ,
x y z

G
B
g
B
ae
B
ai
B
G
M
G
w
G
S
gyro types.
The errors in the measurements of specific force
provided !y an accelerometer triad may !e expressed as
shown !elow. It is assumed that the sensors are mounted
with their sensitive axes nominally aligned with the
principal axes of the host vehicle$
(+.+)
where
is a three-element vector representing the fixed
!iases,
is a - x - matrix representing the vi!ro-
pendulous error coefficients,
is a diagonal matrix representing the scale-factor
errors,
is a - x - s"ew symmetric matrix representing
the mounting misalignments and cross-coupling
terms,
is a three element vector representing the in-run
random !ias errors.
The accelerometer errors descri!ed are particularly relevant
for the pendulous force-feed!ac" accelerometer which is
most commonly used at the present time for many different
strap-down system applications. In the following section,
these errors will !e discussed.
(.&.& )ias
The !ias for gyro or accelerometer is the average
over a specified time of accelerometer or gyro output
measured at specified operating conditions that have no
correlation with input acceleration or rotation. The gyro !ias
is typically expressed in degree per hour (./hr) or radian per
second (rad/s) and the accelerometer !ias is expressed in
meter per second s(uare 0m/s1 or g2.
3ias generally consists of two parts - a
deterministic part called bias offset and a random part.
The !ias offset, which refers to the offset in the
measurement provided !y the inertial sensor, is
deterministic in nature and can !e determined !y
cali!ration. The random part is called bias drift, which
refers to the rate at which the error in an inertial sensor
accumulates with time. The !ias drift and the sensor
output uncertainty are random in nature and they should
!e modeled as a stochastic process. 3ias errors can !e
reduced from the reference values, !ut the specific amount
is range and type dependent.
*a+le (.& ,yro Error Sour'es
Errors -e'hani'al
gyros
(Spinning mass
gyros, 4T5s)
-E-S gyros
(!ased on
6oriolis
effect)
.pti'al
gyros
(75s,
895s)
3ias Sta!ility
(./hr)
/ / /
Scale actor :rror
(;;*)
/ / /
Scale actor
Nonlinearity (<)
/ / /
g-dependent drift
(./hr/g)
/ / 0
g1-dependent drift
or 'nisoelastic
!ias (./hr/g1)
/ / 0
'nisoinertia
errors (./hr/
(rad/sec)1)
/ 0 0
Temperature
effects
/ / /
'xis
*isalignments
(deg or rad)
/ / /
'ngle 8andom
=al" (./s(rt(hr))
0 / /
(Note1 &. g, g1-dependent drifts sometimes occurs in FOGs but never
in RGs&
(. Sca!e factor errors and non-!inearity, cross-coup!ing
errors, g, g1-dependent drifts, anisoinertia errors are
motion-dependent errors " i#e# occurs due to app!ied
acce!erations, angu!ar rates or due to vibrations)
In addition to the a!ove, there are another two
characteristics used to descri!e the sensor !ias. The first is
the bias asymmetry (for gyro or accelerometer), which is
the difference !etween the !ias for positive and negative
inputs, typically expressed in degree per hour (./hr) or
x y z x x
y $ v z x $ y $ y $
z x y z z
f a a a a
f B B a a S a M a w
f a a a a

_
_ _ _


+ + + +




, , ,
,
$
B
$
S
$
w
$
M
v
B
meter per second s(uare 0m/s1 or g2. The second is the
bias instabi!ity (for gyro or accelerometer), which is the
random variation in the !ias as computed over specified
finite sample time and averaging time intervals. This non-
stationary (evolutionary) process is characteri#ed !y a ,/f
power spectral density. It is typically expressed in degree
per hour (./hr) or meter per second s(uare 0m/s1 or g2,
respectively.
(.&.( S'ale 2a'tor
Scale factor is the ratio of a change in the input
intended to !e measured. Scale factor is generally evaluated
as the slope of the straight line that can !e fit !y the method
of least s(uares to input-output data. The scale factor error
is deterministic in nature and can !e determined !y
cali!ration. Scale factor asymmetry (for gyro or
accelerometer) is the difference !etween the scale factor
measured with positive input and that measured with
negative input, specified as a fraction of the scale factor
measured over the input range. Scale factor asymmetry
implies that the slope of the input-output function is
discontinuous at #ero input. It must !e distinguished from
other nonlinearities.
Sca!e factor stabi!ity, which is the capa!ility of the
inertial sensor to accurately sense angular velocity (or
acceleration) at different angular rates (or at different
accelerations), can also !e used to descri!e scale factor.
Scale factor sta!ility is presumed to mean the variation of
scale factor with temperature and its repeatabi!ity, which is
expressed as part per million (;;*). 4eviations from the
theoretical scale are due to system imperfections.
or example a gyro with ->> ppm scale factor
error (?>.>-<) the angular error is in the area of >., degree
after one revolution. =ith a laser gyro or high performance
fi!er optical gyro system with ,>> ppm scale factor error
the angular error is >.>>+ degrees if the rotation angle is +>
degrees.
(.&.% A''eleration dependent dri3ts (3or gyros)
5yros (mechanical or *:*S) produce a drift
(angular rate offset) dependent on linear acceleration and
environmental vi!ration influence, a so-called g-
dependent drift expressed in ./hr/g.
$nisoinertia errors arise in spinning mass
gyroscopes and introduce !iases owing to ine(ualities in
gyroscope moments of inertia a!out different axes.
'nisoinertia is fre(uency sensitive if the rotor is driven !y
a hysterisis motor. This is a conse(uence of the elastic
coupling !etween the magnetic ring on the rotor and the
rotating magnetic field. The resulting !iases are
proportional to the product of angular rates applied a!out
pairs of orthogonal axes. The anisoinertia coefficients may
!e expressed in units of ./hr/(rad/sec)1. To eliminate such
an error source re(uires complicated compensation, so the
system designer must understand the interplay !etween
vehicle environment, sensor design, and the resulting
navigation system performance when selecting a
particular type of gyroscope for a strap-down system.
(.&.4 A5is (Sensor -ounting) -isalignments
'xes misalignment is the error resulting from the
imperfection of mounting the sensors. It usually results in
a non-orthogonality of the axes defining the INS !ody
frame. 's a result, each axis is affected !y the
measurements of the other two axes in the !ody frame
(sometimes called cross-coup!ing errors). 'xes
misalignment can, in general, !e compensated or modeled
in INS error e(uation.
' gyro misa!ignment, allows !oth roll and pitch
gyros to sense a fraction of the yaw rate, causing an
attitude error a!out level axes. The attitude error, in turn,
causes an acceleration error along the level axes. The
sensor misalignment factor depends not on the (uality of
the sensors !ut on how well a factory cali!ration is
performed with inertial sensors rigidly mounted together.
or example, a misalignment !etween the gyro
axes (or accelerometer axes) causes a cross- coupling
!etween the measurement axes. ' misalignment of >.,
mrad inside of the system (e.g. residual cali!ration
mismatch) leads to a roll error of >.>-@ degree during one
revolution around the yaw axis (if t%e system is unaided).
The smaller the re(uired misalignment the higher the
re(uirements to sensor performance and cali!ration
e(uipment such as three-axes turn-tab!es.
(.&.6 7andom 8al9
This value, given in deg/s(rt(hr) for gyro or
g/s(rt()#) for accelerometer, shows the noise of the used
sensors. 8andom wal" arises due to the integration of white
noise present in the inertial sensors ('ngle random wal"
('8=) for gyros or Aelocity random wal" for
accelerometers). Noise is an additional signal resulting from
the sensor itself or other electronic e(uipment that interfere
with the output signals trying to measure. Noise is in
general non-systematic and therefore cannot !e removed
from the data using deterministic models. It can only !e
modeled !y stoc%astic process.
or gyros, the higher the noise the more noise is
measured on the angular rates and on the angles. Some
manufacturers also specify it as the noise density in
deg/hr/s(rt()#). 3oth values are e(uivalent - if the second
value is divided !y @>, you get it in deg/s(rt(hr). 'n angular
random wal" of >.>>- deg/s(rt(hr) indicates, that the
angular error (uncertainty) due to random wal" is e.g. >.>>-
deg after , hour (unaided) or >.>>+ deg after -> minutes (all
values one sigma). The angular random wal" is very
important for the accuracy of north see"ing (gyro
compassing), !ecause if the random wal" decreases times +
then the needed duration for north see"ing decreases !y
times four (if the resolution of the gyro is high enough) as
per the e(uation (+.-),
(+.-)
where
is the alignment time,
is the earth rotation rate (? ,B.>C deg/hr),
is the latitude at which the alignment ta"es
place,
A is the a#imuth error during alignment.
(.&.# :i+ration dependent errors
There are some inertial sensor errors arises !ecause
of the effects of vi!ratory and oscillatory motion and
includes
'nisoelastic !iases
Scale factor Non-linearity
Ai!ro-pendulous errors
6oning errors
Sculling errors
Si#e effect errors
;seudo motion errors
(.&.#.& Anisoelasti' +iases (3or gyros)
$nisoe!astic biases, so-called g1-dependent drift
expressed in ./hr/g1 are proportional to the product of
acceleration along orthogonal pairs of axes. Such !iases
arise in spinning mass gyroscopes !ecause the gyroscope
rotor suspension structure, particularly the !earings, has
finite compliances which are une(ual in different directions.
(.&.#.( :i+ro-pendulous errors (3or a''elerometers)
4ynamic cross-coupling in pendulous
accelerometers arises owing to angular displacement of the
pendulum which gives rise to a rectified output when
su!Dected to vi!ratory motion. This type of error can arise in
any pendulous accelerometer depending on the phasing
!etween the vi!ration and the pendulum displacement. The
magnitude of the resulting error is maximi#ed when the
vi!ration acts in a plane normal to the pivot axis at CB. to
the sensitive axis and when the pendulum displacement is in
phase with the vi!ration. This error may !e expressed in
units of g/g1.
(.&.#.% !oning errors
6oning is the conical (or near conical) motion in
inertial space of one of the gyroscope input axes, as
+
cos
a
e
$R&
'
$
_

,
a
'
e

illustrated in igure +.,. Such motion results from the

simultaneous application of angular oscillations a!out two
orthogonal axes of the system, where the oscillations are
out of phase. It depends on the integration order in which
the :uler angles are generated, for example, using Mi!!er(s
algorithm.
2igure (.&1 !oning motion
(.&.#.4 S'ulling errors
Sculling is made up of a com!ination of linear and
angular oscillatory motions of e(ual fre(uency in
orthogonal axes. In the presence of such motion, errors can
arise in the strap-down computing tas" which is concerned
with the resolution of the measured specific force vector
into the chosen navigation reference frame. 'n acceleration
!ias can arise through failure to ta"e account of the rapid
changes of attitude occurring !etween successive specific
force vector resolutions.
(.&.#.6 Si;e e33e't errors
The specific force acting on a vehicle is detected
!y sensing motion along three orthogonal axes, often using
a triad of linear accelerometers. In order to navigate, it is
re(uired to sense the linear accelerations acting at a
particular point in the vehicle, at its center of gravity for
example. =hether the inertial navigation system is mounted
precisely at the vehicle center of gravity or, as is more
usual, at some off-set location, it provides a measure of the
motion of that point within the vehicle. This assumes that
the inertial system is a!le to sense all motion accurately,
including any centripetal and tangential forces induced !y
vehicle rotation, and to process accurately the inertial
measurements which are generated.
Ideally, it is re(uired that all three accelerometers
should !e mounted precisely at the same location in the
vehicle. This is clearly impossi!le !ecause the sensors are
of finite si#e and !ecause of design constraints on the
positioning of hardware. The centripetal and tangential
forces sensed !y the accelerometers !ecause of their
physical displacements with respect to the desired position
are referred to as the EsizeE effect.
(.&.#.# Pseudo-motion errors
The inertial sensors themselves may produce false
signals which are correlated with each other. The navigation
system will then interpret these signals as indicating the
presence of coning or sculling motion for instance, where
no such motion is actually present. The apparent motions
are sometimes referred to as pseudo-coning or pseudo-
sculling. If these motions are treated as true !y the
navigation system computer, the performance of the system
will !e degraded.
6oning, Sculling, pseudo-motion and Si#e effect errors are
discussed in detail in reference )*+.
(.( Initial Alignment Errors
Initial alignment consists of acce!erometer !eve!ing
and gyro compassing. 4uring the leveling of
accelerometers with respect to hori#ontal plane, the initial
roll and pitch angles can !e o!tained from the tilt of the
accelerometers. 4uring gyro compassing the heading (or
a#imuth) error will occur and will depend on x-axis gyro
accuracy. Thus alignment errors incurred !y initiali#ing the
transformation matrix using the accelerometers for leveling
information result in tilt errors which produce navigation
errors.
(.% !omputational Errors
6omputation-induced errors in strap-down systems
refer to those navigation errors introduced !y the incorrect
transformation of the !ody-measured specific forces to the
computational frame.
The factors causing the incorrect transformation matrix
include$
the algorithm used to generate the transformation
matrix,
the speed at which the transformation occurs
(!andwidth limitations),
the roundoff due to computer word length used
(truncation of the mathematical functions),
(uanti#ation of gyro outputs,
truncation due to integration scheme,
renormali#ation of the (uaternion vector.
They all play a part in contri!uting to computational errors.
Since this type of error also depends on the vehicleEs
dynamic environment, most designers evaluate it !y
computer simulation.
% Error Propagation in an <naided INS
)ere, the error propagation for s%ort duration (less
than +> .. -> minutes) and !ong duration (more than ->
minutes) flights with an unaided INS are given. 3efore
discussing the position error propagation in INS, the
concept of Sc%u!er Osci!!ation has to !e understand. The
errors which propagate in an inertial navigation system over
long periods of time are characteri#ed !y Schuler
oscillation.
2igure %.&1 *he Schuler Pendulum
%.& S'huler .s'illation
=hen a pendulous gyroscopic mass, such as an
INS platform, is accelerated over the earth, it will tend to
oscillate, and as a result the outputs from the sensors will
!ecome inaccurate. 4r. Maximi!ian Sc%u!er solved the
pro!lem !y demonstrating that no oscillation will occur if
the platform exhi!its the characteristics of an E,art%
-endu!umE as shown in igure -.,.
'ssume that the length of the pendulum is e(ual to
the radius of the earth, and its !o! (plum! weight) is
positioned at the earthEs center with its point of suspension
at the earthEs surface. If the suspension point of such a
pendulum were to !e accelerated over the earthEs surface, its
inertia coupled with the force of gravity would com!ine to
hold the !o! stationary at the center of the earth, and the
shaft of the pendulum would remain vertical throughout. If
the !o! of an earth pendulum were distur!ed, as would !e
the case if an aircraft was the suspension point, it would
oscillate with a period of FC.C minutes.
'ny errors experienced, as a direct result of SchulerEs
oscillation will thus average out to #ero over the FC.C
minute cycle. or example if a ground speed error of B
"nots was generated, the error would !uild up to B "nots
during the first +,., minutes, reduce to #ero !y C+.+
minutes, increase to B "nots in the opposite sense !y @-.-
minutes, and return to #ero again !y FC.C minutes. In
practical terms, this means that the output from the INS will
!e correct three times every Schuler period& once when the
period starts and then again at the end. It will also !e correct
in the middle, at C+.+ minutes. 't +,., minutes the error will
!e a maximum high value and at @-.- minutes it will !e a
maximum low value.
Schuler oscillation is given !y
(-.,)
%.( INS Single !hannel Position Error Propagation
)ere the position error propagation in a single
channel in INS considering long term and short term
applications are discussed.
>
s
g
R

%.(.& Single 'hannel error model
or a strap-down INS mounted in a vehicle
traveling at constant speed and at constant height a!ove
the :arth, the error dynamics for the nort% c%anne! may
!e expressed in terms of the following set of coupled
differential e(uations,
(-.+)
where is the alignment error with respect to vertical,

is the heading error,
represent the effective gyro !iases acting a!out
the east axis,
represent the effective gyro !iases acting a!out
the vertical axis,
is the net accelerometer !ias acting in the north
direction.
' !loc" diagram representation showing the sensor errors
and the initia! condition errors is shown in igure -.+.
2igure %.(1 Simpli3ied +lo'9 diagram o3 north 'hannel o3 an
INS 3or long term navigation
rom the a!ove diagram, it is noted that in the
presence of a vehicle acceleration, the a#imuth alignment
error is coupled directly into the hori#ontal accelerometer,
as indicated !y the additional signal path shown dotted in
the !loc" diagram in igure -.+.
or very short term navigation, that is, when the
navigation time is a small fraction of a Schuler period, the
Schuler feed!ac" has relatively little effect on the growth
of errors and the single channel error model can !e
reduced further to the form given in igure -.-.
In the following section, the expressions for the
nort% position errors which are applica!le for long and
short term applications are given.
2igure %.%1 Simpli3ied +lo'9 diagram o3 north 'hannel o3 an
INS 3or short term navigation
%.(.( Single 'hannel error propagation e=uations
' constant velocity error gives rise to a north
position error of $
(-.-)
'n effective acceleration !ias acting in the north
channel gives rise to a north position error
of$
(-.C)
'n effective angular rate !ias acting a!out the east
axis gives rise to a north position error
of$
g,
B
g.
B
a/
B
>
sin
s
/
s
t
x v

+
, cos
s
/ a/
s
t
x B

,
( )
g,
B
( )
>
v
( )
a/
B
(-.B)
Similarly, a heading error gives rise to a
north position error of$
(-.@)
'n effective angular rate !ias acting a!out the
vertical axis gives rise to a north position
error of$

(-.G)
7ver very short periods of navigation, that is,
navigation over a small fraction of a Schuler period, further
simplifications may !e made to these expressions. The
position error contri!utions in the long and short term are
summari#ed in Ta!le -.,. These e(uations can !e used to
assess system performance or to specify an inertial
navigation system to fulfill a particular application. The
propagation of each error type with time in an unaided 0/S
is illustrated in igures -.C % -.H.
In igure -.C % -.H, plots are given illustrating the
propagation of position errors with time over a --hour
period. The system is assumed to !e stationary and located
on the surface of the :arth at a latitude o3 &>?. The
following errors sources are included.
Initial alignment errors - roll/pitch
? >., mrad
Initial heading error ? ,.> mrad
Initial velocity error ? , m/sec
5yro !ias ? >.>, deg/hr
'ccelerometer !ias ? +>> micro-g
*a+le %.&1 ,rowth o3 INS position errors in the long term
and short term
igure -.C shows the north position error resulting
from the hori#ontal component of gyro !ias. It is seen that
given a residual gyro !ias of >.>, deg/hr, the position error
will grow at the rate of >.G nm/hr. It should !e noted that
this represents the error !uild up along the North and :ast
axes, so that the radial error will !e multiplied !y s=rt(() or
s=rt(north@( A east@() and hence will !e e(ual to & nmBhr.
Thus in order to achieve * nm1%r position accuracy, one
should go for >.>, deg/hr class gyros.
2igure %.41 North position error verses time 'aused +y hori;ontal
gyro +ias
>
sin
s
/ g,
s
t
x B R t

_


,
( )
>

> >
sin
cos
s
/
s
t
x R t

_


,
&
( )
g.
B
+
> +
, cos
cos
+
s
/ g.
s
t t
x B R

_

' ;
,

&
( )
> >
,
( )
>

( )
> /
v
( )
g
B
( )
a
B
igure -.B shows a plot on the north position error
resulting from an initial velocity error. This error is !ounded
!y the effect of the Schuler loop and a figure for the
maximum position error caused !y a , m/sec initial velocity
error.
2igure %.61 North position error verses time 'aused +y initial
velo'ity error
igure -.@ shows a plot of the north position error
resulting from the hori#ontal component of accelerometer
!ias. The propagation of this type of error is also !ounded
!y the effect of the Schuler loop. ' residual accelerometer
!ias of +>> micro-g (corresponding to an initial tilt error of
>.+ mrad) will result in pea" position error of ,.-F nautical
miles.
2igure %.#1 North position error verses time 'aused +y
a''elerometer +ias
igure -.G shows a plot of the north position error
resulting from an initial error in a#imuth alignment. The
resulting position error comprises a Schuler oscillation as a
ramp function, the magnitude of the error varying with
system latitude and speed. Some examples of the resulting
error magnitudes are given for an a#imuth misalignment of
, mrad.
2igure %.>1 North position error verses time 'aused +y initial
a;imuth misalignment
igure -.F shows a plot of the north position error
resulting from an initial misalignment error with respect to
vertical (initial roll or pitch error). The propagation of this
type of error is also !ounded !y the effect of the Schuler
loop. 'n initial roll or pitch error of >., mrad (resulting
from ,>> micro-g accelerometer !ias) will result in pea"
position error of around >.G nautical miles.
2igure %.C1 North position error verses time 'aused +y initial
misalignment w.r.t verti'al
igure -.H shows a plot of north position error
resulting from a vertical gyroscope !ias component. The
position error comprises a (uadratic term with a
superimposed Schuler oscillation& the relative magnitude of
the oscillatory component is small and hence is not readily
apparent in the figure with the scaling adopted here. The
propagation of this error also varies with latitude and
example figures are given in the figure for a >.>,./h
gyroscope !ias.
2igure %.D1 North position error verses time 'aused +y verti'al
gyro +ias 'omponent
's with the north channel, the growth of many of
the errors is !ounded !y the effects of the Schuler tuning.
)owever, this is not the case in the vertical channel where
the errors increase rapidly with time. or example, a
vertical accelerometer !ias, B2az, will give rise to a
position error of B2az3t114# It is for this reason that aircraft
navigation systems commonly operate in conDunction with a
barometric or radar altimeter in order to restrict the growth
of vertical channel errors.
2igure %.&$1 Air'ra3t unaided INS single 'hannel error
propagation
The errors are summari#ed in igure -.,> for an
aircraft navigation system over a - hour period. The figure
illustrates the form of the position errors resulting from the
various error sources whilst the upper curve (dotted !lue
line) represents the com!ined effect of the individual errors.
The upper curve has !een o!tained !y summing the
individual error components (uadratically to give the total
navigation error. It will !e seen that the gyroscope !ias and
a#imuthal misalignment contri!utions grow with time
whilst the other terms are !ounded as a result of Schuler
tuning. It should !e noted that these are simplified results
for a single channel. In a full model, the Foucau!t effect is
noticea!le even at the first Schuler period.
The oucault oscillation, given !y e = sin .
This maintains itself as a modulation of the Schuler
oscillation, the modulation in the two hori#ontal channels
!eing H>. apart in phase. The oucault oscillation has a
period of 4 1 sin . The period of this oscillation is
a!out -> hours for moderate latitudes.
%.% Error propagation during air'ra3t maneuvering
The analysis of inertial navigation system
performance for air!orne applications rapidly !ecomes
complex when account is ta"en of realistic vehicle
traDectories and maneuvers, in which case the analyst will
usually turn to simulation to aid the design process.
In addition to the usual effects of alignment errors
and sensor !iases illustrated in the a!ove error plots, a
num!er of error contri!utions arise as a result of the
acceleration experienced !y an aircraft during ta"e-off.
3oth alignment errors and mounting misalignments of the
accelerometers will give rise to cross-trac5 ve!ocity errors
as the aircraft accelerates during ta"e-off. or example, an
a#imuthal misalignment of will give rise to a velocity
error of , where 6 is the cruise speed of the aircraft.
The velocity error which has !uilt up during ta"e-off then
propagates during the su!se(uent cruise phase of flight in
the same way as an initial velocity error. Sensor scale-factor
inaccuracy and acceleration dependent !iases will give rise
to additional navigation errors in the event of an aircraft
maneuver. 3ecause of applied accelerations, angular rates,
induced vi!rations during maneuvering degrades the
performance of a S4INS.
or example, an aircraft at vertically down
position, the accelerometer will measure , g at that time
(3ut in real case, it was not always at ,g). 't that time with
accelerometer scale factor error of +>> ppm will
corresponds to +>> micro-g. 3ecause of this scale factor
error, the position error also propagates and is bounded
within Schuler oscillation. =hen using an optical gyros, a
coning motion of >.,I at a fre(uency of B> )#, for instance,
can result in a drift in the computed attitude of almost ,>>I/
hr.
3ecause of Sculling error, a steady acceleration
component of $.6A.'os occurs in the x-direction. It is
stressed that this error term can arise even when using
perfect accelerometers, !eing purely a function of the
inaccuracy in the resolution process. If, for example, ' ?
,>g, ? >.,. and the phase difference () is #ero, the
resulting acceleration !ias is around H milli-g.
Si#e effect also introduces errors during the
navigation. or example, in the presence of cyclic angular
motion of amplitude >.,., and of fre(uency B> )#, for
example, a !ias of J ,.B milli-g arises for a displacement of
,> cm, i.e. the distance !etween the center of gravity of the
vehicle and the location of the sensor.
4 7e3eren'es
0,2 4avid ). Titterton K Lohn 9. =eston, Strapdown
Inertial Navigation Technology, +
nd
:dition, 0,, Radar,
Sonar, and /avigation Series, +>>C.
0+2 4r. Ing. :. )inue!er, MIf you investigate in an inertial
measurement system... which technical data you should
analyse and compare !efore ma"ing your decisionN, iM$R
'ec%nica! /otes.
0-2 '. 4. Oing, MInertial Navigation % orty Pears of
:volutionN, G,7 Review, Aol.,-, No. -, ,HHF.
0C2 )aiying )ou, M*odeling Inertial Sensor :rrors
Qsing 'llan AarianceN, 87G, Report R +>>,, 4epartment
of 5eomatics :ngineering, Qniversity of 6algary, +>>C.
0B2 6hing-ang 9in, *odern Navigation, 5uidance, and
6ontrol ;rocessing, -rentice 9a!! Series, ,HH,.
0@2 Sergei '. Lere!ets, M5yro :valuation for the *ission
to LupiterN, 0,,, $erospace 7onference, +>>G.
0G2 8. ;. 5. 6ollinson, Introduction to 'vionics, irst
:dition, 7%apman : 9a!! pub!ications, ,HH@.
0F2 4r. Poung 6. 9ee K 4aniel 5. 7E9aughlin, M'
;erformance 'nalysis of a Tightly 6oupled 5;S/Inertial
System for Two Integrity *onitoring *ethodsN, 7enter for
$dvanced $viation System .eve!opment (6''S4)
Technical ;aper, *IT8: 6orporation, ,HHH.
0H2 MInertial Navigation SystemN, &i!;am F!ig%t
'raining 7enter Technical Notes.
>
6
>