GPSIINS uses Low-Cost MEMS IMU

Alison K. Brown
NA VSYS Corporation

ABSTRACT

A description of the design, operation, and test results of

a miniature lowcost, integrated GPS/inertial navigation
system that uses commercial offtheshelf Micro-Electro
Mechanical System (MEMS) accelerometers and
gyroscopes. The MEMS inertial measurement unit (IMU)
is packaged in a small size and provides the raw IMU data
through a serial interface to a processor board where the
inertial navigation solution and integrated GPS/inertial
Kalman fIlter is generated.
The GPS/inertial software integration is performed
using NAVSYS' modular InterNav software product. This
Fig. 1. Silver Fox UAV [1]
allows integration with different low-cost GPS chip sets or
receivers and also allows the integrated GPSfmertial
navigation solution to be embedded as an application on a
customer's host computer. This modular object-oriented
architecture facilitates integration of the miniature MEMS
GPS/INS navigation system for embedded navigation
applications and is designed to handle the large errors
characteristic of a low-grade MEMS IMU.
Test results are presented showing the performance of
the integrated MEMS GPS/inertial navigation system.
Data is provided showing the position, velocity, and
attitude accuracy when operating with GPS aiding and also
for p eriods where GPS dropouts occur and alternative
navigation update sources are used to bound the MEMS
inertial navigation error growth.

INTRODUCTION Fig. 2. Unmanned Ground Vehicle [2]

The advent of low-cost, MEMS accelerometers and navigation for a wide variety of new applications, This
gyroscopes offers the opportunity for applying inertial includes navigation and guidance oflow-cost. small unmanned
air vehicles (UAVs) or unmanned ground vehicles (UaVs)
at the 11th Saint Petersburg International such as shown in Figures 1 and 2, While the principles of
Conference OD Integrated Navigation Systems, Saint Petersburg, Russia, 24-26 May 2004,
'This contribution was originally presented

appeared in the Proceedings of the conference, and is reproduced here wilh the kind
permission of the Copyright owner, the State Research Center of Russia Eleklropribor, 2004.
IEEE AESS has been one of the cosponsors of this conference for many years.
14960 Woodcarver Road, Colorodo Springs. CO 80921, USA.
Author's Curremt Address:
NAVSYS Corporation,
Based on presentation at the 2004 81. Petersburg Conference.
0885189851051 $17.00 2005 IEEE
inertial navigation are well-understood, the challenge, when
working with the current generation of low-cost MEMS
instruments, is to develop a robust navigation capability that
can deal with the large instrument errors experienced with
these low-grade accelerometers and gyroscopes. The software
product described in this paper, InterNav, was developed to
include this capability. This paper describes the approach
taken for providing inertial navigation with low-grade MEMS

IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005 3

 Table 1. Il\fU Gyroscope and Accelerometer Parameter Comparison

UNITS H0l700 (71 Crista [81

. Parameters
Type Ring Laser Gyro MEMS
Size 33 cu in 1.6 cu in
Weight 320z 0.7oz
Power 8w 0. 7 w
Gyroscopes
Operating Ranae D/s 1000 300
Scale factor accuracy (l 0) ppm 150 25000
Scale factor linearity I 0 to 800 Dis ppm 150 N/A
BissO 0) /hour 2 500
Axis alignment stability (I 0) prad SOO 3000
Axis alignment stability, non-ortho20nality (1 0) "rad 100 N/A
Output noise {I oof 10,000 samples} I1rad 80 80
Angular random walk max. lRthr 0.1 3
Accelerometers
Operating Range g 50 10
Scale factor accuracy (l 0) ppm

300 25000
Scale factor linearity (1 0) ppm 500 N/A
Bias (I 0) mg LO 15000
Axis alignment stability (I 0) "rad 500 3000
Axis alignment stability, non-orthogonality (l 0) "rad 100 N1A
Output noise ( ) 0 oflO 000 samples) mls 0.0024 0.00031
Velocity random Walk (ugIRt-Hz) 150 400
I. Accelerometer includes filtering in sampled signal

instruments and presents initial test results showing the type of

navigation performance that can be expected using current
generation MEMS devices.

COMPARISON OF INERTIAL MEASUREMENT

UNITS AND INSTRUMENT PERFORMANCE

In Table 1, a comparison of a commercial off-the-shelf

(COTS) MEMS IMU and a low-cost COTS Ring Laser Gyro
(RLG) IMU is shown. The RLG IMU used for the comparison
is the HG1700 produced by Honeywell (see Figure 3). This is
available as a commercial unit and has been integrated into a
number of different GPS/inertial products produced by
NAVSYS [3,4]. The MEMS IMU used for the comparison is
the Crista IMU produced by Cloud Cap Technology (see
Figure. 4). This is built using a triad of Analog Devices
accelerometers (5] and gyroscopes [6]. As seen from the
performance figures in Table 1, the instruments used by the
Crista IMU, while significantly smaller, lower-cost, and lower
power, are also almost a factor of 100 times less accurate than
the HG1700 instruments. While future MEMS technologies
promise to provide improved performance levels, approaching
those of the HG1700 instruments, the challenge today for low
cost navigation applications is to design an integrated system
that can perform inertial navigation using these existing low Fig. 3. Honeywell HG1700 IMU
grade MEMS instruments.
and Kalman filter functions used to combine inertial
INTERNAV SOFTWARE measurement unit raw data from the gyroscopes and
accelerometers (Lla, LlV) with other sensor data to provide an
The MEMS inertial navigation integration and testing was integrated inertial navigation solution. The software includes
performed using NAVSYS' InterNav integrated GPS/inertial the functions illustrated in Figure 5. The inertial measurements
software product (9]. InterNav includes the inertial navigation are integrated using a quaternion integration algorithm to

4 IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005

 0 ... -
I
0"'''''''-
. __ .....n
."D\fl\JIfIitT.q:ST

Fig. 5. InterNav Software Architecture

Fig. 4. Cloud Cap Crista IMU

propagate the inertial states. Periodically, updates are

performed to these states to calibrate the inertial state errors
and instruments. A Kalman filter is used to perform these
updates.
The InterNav software includes different Kalman filter
configurations which are designed to perform both basic Fig. 6. GI-Eye Test Fixture
aligriment of the inertial states, calibration of the inertial errors,
and advanced functions to facilitate integration with a wide alignment mode of operation prior to transitioning to the
variety of different aiding sources of data. The basic InterNav Kalman filter fine alignment mode. In the rough alignment
filter states are shown in Table 2 [10]. To align the inertial mode, the initial position, velocity, attitude, and rough
navigation states, updates are required to observe the initial calibration parameters for the inertial instruments are set. In
inertial error state. This is accomplished using a rough the fine alignment mode, the filter calculates the best estimate

Table 2. Basic InterNav Kalman Filter Navigation States

State Meaning
1-3 Position Error (navigation frame)
4-6 Velocity Error (navigation frame)
79 Body Attitude Error (navigation frame) (Til, Tv,a)
10-12 Accelerometer bias error
1315 Gyro bias error
16 GPS Clock bias error
17 GPS Clock frequency error
1826 Accelerometer misalignment & scale factor error
27-32 Gyro misaliJtnment & scrue factor error

IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005 5

 Ground Tl'Kk
39.085 OIf111",I'I(:' In Posion ComparlSon$ (E.sI): Kin. Align. HGHOO

HG1700
39.08
HO 1700 UI'I(:,I'I.inly
MEMS

,
I
_ .
2

39.075
-- MEMS Unc.rt.irlty
. . .
.

39.07
i
. . . .

'2.
39.065 -, -.- - -. - ,
i
39.06 - , - , " , -

39,055

J9.05

39045 L--"---"----""",-""""-"-...-
- -",,-,,,,
,,,
longnudt (deQ)
"04.9-104.1195104.1111'04.111$104.11&104.1115'04.11];'04.1'11$ 104.16104.8$$ 104.115

Fig. 7. Truck Test Route

Fig. 10. East Position Error

2140

2t30
OIl1erenc. in HG1700
2120
Roll: MEMS .

2110 MEMS, HG1700

3 HG 1700 Uncertainty
-

_ 2100 - MEMS Uncel1ainly

!1. 2
2090 "

2010
or

O------'f
GPS T."" ",nn '0 2lI7tl6 (..tct 2

Fig. 8. Altitude
-3

400 IlOO 1000 1200 1400

01.,.1'1(:' In Positkm Comp.1lrlSons (Noll"): MEMS - HG 1700

,
2. "
. . . . .
- - - . . . . . . - - - - - - .

Fig. 11. Roll Error

for these parameters using measurement updates from a GPS

,
....
. ..-".,..-.-... -.. --.- .."., --- ... >... ...-.. y-_._"----'- ...-.__ .. .

-.
t I
receiver or other source of aiding information,

MeMS HG 1700 The InterNav software is designed to accept inertial
H017DO \)n1;'I'I.iI1Iy
" -

f!
0
" navigation updates from a variety of different sources as
illustrated in Figure 19, For the purposes of the MEMS IMU
-3
" " - MEMS Une.., ..""

d
..
performance testing, we used a GPS receiver as the

S
sensor-aiding source, A discussion of how the different sensor
,- ,
"
inputs can be applied to aid the MEMS inertial navigation
.. , solution in the event of GPS dropouts is included in this paper.
1
-

GPS AIDED MEMS IMU FIELD TEST RESULTS

- -

200 400 eoo aoo 1000 1200

COPS TIme 29761 (eJ;1
1.00
i,,C! '0

To test the MEMS IMU, we instaI1ed it on the GI-Eye test

fixture shown in Figure 6, This unit includes the HG 1700 IMU,
Fig. 9. North Position Error a Novatel OEM-4 GPS receiver and antenna, a camera used as

6 IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005

 OiIft e in Plcll: MEMS HGt700
...

500 ..........j. . . ;......... i.. MEMS NoYAIel GPS

2 : MEMS Unttlt_,
. . .. . . . ..

: :
GI
-

..........} . . .... . . i

..
. . ..
I
.
..
..
.
..
..

... . . ... t ..
4

t
f

Xl)

.......... .. .. .
. i . ...
.
. .... . . .

.. j .
...... ..
.
.. . .

............... .. ... ."'

.

,
. .

200 .. ..
. .

.. . .... .. . .. ...... . .. - - -. .-- . ........ ... . .. . + ...

: tOO . . .
..
.. . .. .. .. ...... . .. .
.,
. .

r

r
.

i
.

f ..

1"
'"

0
'"

: ..----'------
'"
.,

.. ..

6 00 ..... . . . .
'" 1 '. ". - -. - t ..

a 1
"' .. _ .... f ........... ... .. f .. .............. .
- f - ... ...... ..... . .. . ..

- Mlellgn HG1700 r,.. r .. t

'"

200 .... f
u 4 '

2 HGl100 Unc:eMInty
'"
.. ..... .. .. .. .. .. .. .... .. ........... .. .. ...... .. ...... ...... .. .... ..
0. .

. MEMS Unc:en./nlJ 300

.......... .. . ....:. ............. -)............ - .. " . . '" ... ............... ; . ... ..... ...

: : : : :
........... ....
.. ,

t t
. ..

3 '.
-0)
....... .. t .. ...... .. .. .... "' .... .. .. ... ........

.... .. ,. ..

.. ..
-SXI ................. . f _!-- .-
.
................. ....... .. .... .. ..... .......... - .. - .-.-. --_ .. . ...... - ..

200 400 800 100 1000 1200 1400 35D -0) .t5O
uPS lime &Ince 10 4291e/\&ttC) G?S lime ,inti !a = -419767 (uc)

Fig. 12. Pitch Error Fig. 14. MEMS North position

during GPS drop-out
0IIIwenc:.1n Huding: MEMS HGl700
ni -- --
- MEMS HG1700
----

MEMS INERTIAL PERFORMANCE

HG1700 Unc:mtaInIy
15 . <.- MEMS alntJ
UnceJt
FOLLOWING GPS DROP-OUTS

10 .
,
.
. .
Because of the poor quality of the MEMS inertial
,

instruments, the navigation performance degrades extremely

rapidly following loss of the OPSaiding data. To demonstrate
this, a OPS drop-out was forced (post-test) in the data and the
errors on the freeinertial propagated solution was compared
{ .$
against the OPS "truth" solution. The error growth during the
dropout period is illustrated in Figures 14 to 17. As can be seen,
10
..
x

the position error grows very rapidly without the OPS aiding,
.15 exceeding 10 meters within 20 seconds of the start of the
drop-out. To maintain navigation performance when GPS is
not available, another source of aiding information is essential.

AIDING SOURcEs DURING GPS DROP-OUTS

Fig. 13. Heading Error
a truth reference, a PC-I04 computer stack and associated
power supplies and interface circuitry. Data was collected
from these components while driving through the test route
shown in Figures 7 and 8. The HO 1700 integrated GPSlinertial
solution has been extensively tested as a TSPI reference system
[3] and so was used as the truth reference. The comparison of
the position and attitude solutions between the H01700 and the
Crista MEMS integrated navigation solution is shown in
Figures 9 to 13.
The OPS-aided Crista solution performed well in observing
the position and tilt of the vehicle. As expected, the heading
accuracy was much poorer than possible with the HO 1700.
Our previous testing with the HO 1700 has demonstrated
alig nment accuracies of better than 0.06 degrees (1 mrad) [3].
With the Crista !MU, the heading accuracy was only 2 degrees
(35 mrad).
The alternative sources of aiding information that the
InterNav software is designed to accept are illustrated in Figure
19 and described below.
Altimeter
The vertical error growth in the inertial solution can be
bounded by the addition of a baro-altimeter. This provides a
measure of the change in altitude which is used as an update by
the InterNav software.
Dead-Reckoning
InterNav will accept updates from dead reckoning sensorS
that can provide speed and heading information. An example
of this type of back-up mode would be using a digital magnetic
compass integrated with speed data either from an odometer
(e.g., in a ground vehicle) or air speed indicator for (e.g., in an
air vehicle). Applying these updates limits the inertial

IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005 7

 0rIe. in Po5dion Comp .... on (NOIth): totEMS No.IIael GPS 06renee in PO$IIion Complirison (Ent): MEMS eI GPS

oQ"l- 4 . ..... -1'"" .. : : . MENS NovAtIi GPS

. ..:r . ... : I MENS lJrII;.naint'f I
. .
. .. 4 .. . .... .... .... ..

20 . . .
.... ..... ..

T .. .
.

xn y f .................. ! -... -.... -.. ! ...... -...

-

. .. . . . . .

: :
. .

.. . ..-

'
... - ...
.

15 - - - ...
. .

.. .. . .. --+ .... _c .. . * . ... .. 4 ..

............. ... + ............... "'

."
.
.

.
... -- .. ........ -



.
.

.
.
.

15 ..............:- . .. 1.. :
.

10 . .. f- . - ............ -- --
. .

..... : .... - ............. ........... . .. . . . .. ....... ... . -- ... .. ... +

. .
... ... . ..
. .. ..... ..
4- . ... - .... ... . . -

-Y ...:
.. ..

....... . ... .. -.. ..

I
t
5 -.. -.- ..

+ . ... ... .. . ... . . .. . . .. . . .:
:
.. .. .... .. --. !...

:
-- ...
. -
L E
.

10 .---.. +-.. -- - - . --: -.----.............. .. .

.

!"
c

0 . ... .. - ........ -
.. .. ! ... . . .. . . .. - -.. -

- ... : : !
- ... .. .... z .. ..... "..&..... ... .. .. ..
. .... - --

a
.... .. .. ..

i... --
-5 r--- : . . .
r,;

1:"-. ...).. ,,- :, . ...-

-
,. .

.. .
-

:
.. . .}

:
. .. . .........

. . . .

:
. .. . . 1 . ... . .

:
. ..

; :
! : : :
10 ......... "'! ........... -: -...... . . .. . . . .
I

. . . . ....
r. 5 1 f .: t
.. . . . .

.
... .... .
. .
.
. .. ..
..............

f.. f . .. . !.
. .................. ... ...... ...... .... : .. ..

15 r-- . . .. . .
. .

. ..
. .. . .. .

-.+-
...

. "
. . . -

,
. . . .. .. .. . . . . . .

iII- ;I 4 " .
. . .. .. .. :

-- :
.

, ..-

.

.. .. i........ .
.

........ .... --.... --..... !.-- .. . .. ---.--.-....-............

.

. ------.... j. : .... .J
D
.

.. ... ... .

\
-20 -:- ............ . . . .

. : i '
. ...... ..

3!iO .coo 450 SXI :m 3!iO -G) 450 SIXI

I I

D)
GPS Tm. line '0 ..29161 (uc) .. GPS Ttme s'nr;/t 1(1 4297ft7 (GJlC} .-:::

Fig. 15. MEMS North position during Fig. 17. MEMS East position during
GPS drop.out (expanded) GPS drop-out (expanded)

Different, in Polition Comparison (Ees1): MEMS NcJ.6lel GPS

500
:

:
-
: . MEMS NorAaei GPS

400 ...... . f .. . .. .. f .... r" - MEMS Uncm .

. .

. . .......
.. .

"' .. . .. .. . ... . ...

,
"'
.

..
.. ... . . . ... .. .
.
...

. .... ... . .... . .......... ...... . ..
.

I
.. ...
.
.. . . ... ... . . ......

. .:- .. . -......! ......... -

. .
..

. . ... : ........ ,... ... .. ! ... : .

I
... . ... .
.. . .... . .. .. .. .. . .. -_ ..... ..

,
... .... . ... ....
. ... -.

. ... . .... . .. .. . ..... ..... . . .. .. . ..

. ..

. .. ..
t . ..
.

... . ... t
. ... . ..... ..

,
.... .. . ........ .. .. . .. .... . .. . .. . ....
t
..
..
..

.
. . .

D
. .
..
. .

0.100 ,
.............. ............ --:-..

... ....... -....... ....... -.. . : ... ....

. . .
...... .. . .. .. . .. .. .

: :
.

-200 ..... .. ........... ..

...........
...

... .. . - .. ... ..
.

"""
.

;,

.... . .
. ..... , ...... -...... --" ...... -.. . . .. -.-- .. .. .
.

..

t
t
t
,
.
.. . .. .... . . ..... .
.. I

.lJ)
..

.
. 4
... I
..
..

...... .. ...
.. . . ........ .. .... ... .... oo\ ... ,
..01> ... ... . . ..........
. . .. .. .
... .... ... ...
. ..

f 1 l 1 1
. .

.... 7 ........... -: .. .. ...... : .................... .. ... : ... ... ............. . :. .. ... _ .. ... ..
! : ! : :
.

-5ID --

......... ........... .......... ........... .... -- ... . . ..... . . .

l!il .coo .c5O 500

GPS Tim. tIMe 10" '29787 (cee)

Fig. 16. MEMS East position

during GPS drop-out

navigation error drift rates to the accuracy of the aiding data

source.

Video Sensor Fig. 18. Example of Video Model Offset Detection

InterNav is designed to accept updates of position offsets
and relative motion updates using image processing functions Communication TOA Updates
operating on video input data [11, 12]. A model generation InterNav is designed to accept time-of-arrival (TOA)
approach is used to track the motion of objects in images and updates when they are available from a communications link.
also observe the inertial derived position offset from reference As an example, Link 16 provides TOA information that can be ...

landmarks, as illustrated in Figure 18. These video updates can used to damp the inertial error growth following GPS
also be used to provide position updates and velocity updates drop-outs [l3]. We are currently implementing TOA aiding
to bound the inertial error growth during GPS drop-outs. within a Software Defined Radio (SDR) [14] that will provide

8 IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005

 IMU
IMU &v,.68
Measurements

Altimeter
Altitude
Measurements

Air Speedl
Wind Estimation
Velocity Estimate

Compass
Heading
Measuranent Output
Integrated ....
Nav
-
-
Nao.;gator -
Solution
Video
Sensor
Pial
NWgator Optimaly Combines ..., Available Measurements
Video Updates
into Robust Navigation Sdution

GPS
GPS PR/DRlCPH
or Positianf
Velocity updates
COMM
TOA Measurements

Fig. 19. InterNav Alternative Sensor Inputs

a similar function to perform aiding of the inertial solution
following GPS drop-outs. In this mode of operation,
participating units on the communications network that have
access to the GPS signals. Detection can augment the
performance of units without access to GPS by providing three
TOA updates through their communication links.
CONCLUSION
In conclusion, our testing has shown that it is possible to
perform inertial navigation using a low-grade, inexpensive
MEMS IMU when GPS-aiding is available. The InterNav
software includes the capability to observe and calibrate the
MEMS inertial errors and can align the inertial navigation
solution to provide the vehicle's pitch, roll, and heading in
support of guidance and control operations.
The low accuracy of the MEMS inertial instruments means
that the accuracy of the navigation solution degrades rapidly
following loss of lock of the GPS signals. However. InterNav
allows other types of navigation updates to be applied to bound
the inertial error growth. The combination ofGPS and back-up
sensors for inertial aiding. as illustrated in Figure 19, allows
inexpensive MEMS instruments to be used as a low-cost
inertial navigation system.
REFERENCES
[1 J Silver Fox UAV.
http://www.acrtuson.comlUAV/silverfoxl.

IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005 9

 [2J IAI Unmanned Ground Vehicle,
http://www.i-a-Lcom/view.asp?type=view&PassageID:::116&
SubID= 13&TopiclD= I.
[3] M. Nylund, A. Brown and 1J. C lark,
Kinematic GPS-Inetial Na vigation on a Tactical Fighter,
Proceedings of ION GPS/GNSS 2003, Portland, OR,
September 2003, http://www.navsys.comlPapersl0309004.pdf.
[4J GI-Eye Precision Georegistration System,
http://www.navsys.comlProductslgi_eye.htm.
[5] Analog Devices Accelerometers,
http://www.analog.com/UploadedFilesIData_Sheetsl5739IB736AD
XLISO_250_0.pdf.
[6J Analog Devices Gyroscopes,
bttp://www.analog.comlAnaloootJsitePage/mainSectionHome/
0.2130,leveI4%253D%25251D1 %01 %2526Language%25
3DEnglish%25261 eveVl %253D212%25261eveI2%253
0310%25261 eve13%253D%252520 1.00.hOOI.
[7] HG1700 Specification,
http://oontent.honeywell.comldseslassetsldatasheetslds7
hgI700_imu.pdf.
[8] Crista IMU Specification,
http://www.cloudcaptech.comlcrista_imu.httn.
[9] http://www.navsys.comlProducts!intemav.hOO.
10
[10] A. Brown and D. Sullivan,
Precision Kinematic Alignment Using a Low-Cost GPSIINS System,
Proceedings of ION GPS 2002, Portland, OR, September 2002.
http://www.navsys.comlPapersl02-09-002.pdf.
[11] A. Brown and D. Sullivan ,
Inertial Navigation Electro-Optical Aiding During GPS Dropouts,
Proceedings of the Joint Navigation Conference 2002,.
Orlando. FL, May 2002,
http://www.navsys.comlPapersl0205001.pdf.
[12] S. Pender, D. Boid, D. Sullivan and A. Brown,
Video Updates During GPS D ropouts Using Navigation and
ElectrofOptic Sensor Integration Technology,
Proceedings of the 58" Annual ION Meeting,
Albuquerque. NM, June 2002,
http://www.navsys.comlPapersl020600I.pdf.
[13] A. Brown and P. Sack,
Navigation Using LINK-I 6 GPS/INS Integration.
Proceedin gs of ION GPS/GNSS 2003,
Portland. OR. September 2003,
http://www.navsys.comIPapersl0309001.pdf.
[14] F. Carpenter,
PCJI04 Test-Bed for Software GPS Receiver (SCR) and Software
Defined Radio (SDR) Applications,
MPRG's Wireless Communications Symposium,
Virginia Tech., June 2004,
http://www.navsys.comlPapersl0406001.pdf. .4
IEEE A&E SYSTEMS MAGAZINE, SEPTEMBER 2005