.DA5 6. RESULTS OF OPERATIONAL TESTING OF LRSS-11 (LITTON /AUTO-SURVEYOR SYSTEM) SYSTEMS(U) DEFENSE NAPPING AGENCY

WRSHINOTON DC L PFEIFER ET AL. SEP 85

UNCLASSIFIED F/G 8/5 N

IEEE "id

los

1 .0~ 328

1.2 1" 1.6

NA1111 O 1.1 uoFS~~olo -

(•UNL UNSSIFIED"/ASEUryCLASSIFICATION OF TH15 ,MT

/| ~ ~REPORT DOCUMENTATION PAGE , xpDT.Yun),89F.Agpoe

IrDERFPORT SECURITY CLASSIFICATION b. RESTRICTIIE MARKINGS

(JuNCA IFED N/A.SECURITY CLASSIFICATION AUTHORITY 3. DISTRIBUTION /AVAILABILITY OF REPORT

• "N/A Approved for Public Release;DECLASSIFICATIONI/DOWNGRADING SCHE'DULE Distribution unlimited

OPERFORMING ORGANIZATI'ON REPORT NUMBER(S) S. MONITORING ORGANIZATIONREPORT NUMBER(S)

T'N/A N/A

NAME OF PERFORMING ORGANIZATION 6b. OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATIONefense Mapping Agency (If applicable)

ADDRESS (Oty, State, and ZIP Code) "7b. ADDRESS (City, State, and ZIP Code)

Washington, D.C. 20305-3000 N/A

NAME OF FUNDING/SPONSORING 8b. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBERORGANIZATION (If applicable) N/ADMA Hydrographic/Topographic GSST

8c. ADDRESS(City, State, and ZIP Code) 10. SOURC : 0r rONIING NUMBERSPROGRAM PROJECT ITASK IWORK UNIT

Washington, D.C. 20315 ELEMENT NO. NO. | NO. ACCESSION NO.

n/a n/a n/a n/a

* 11. TITLE (include Security Classification)

Results of Operational Testing of LASS-I Systems

12. PERSONAL AUTHOR(S)L. Pfeifer and R. Tyszka T 4 DA E OF R P R (Y a o t ay S . AG C UN13a. TYPE OF REPORT 13b TIME COVERED 114. DATE OF REPORT (Year Month, Day) 115. PAGE COUNTFINAL FROM Aug 84 ToMay 85 September 198 13

16 SUPPLEMENTARY NOTATION

3rd International Symposium on Inertial Technology, 16-20 Sep 85, Banff, Canada

17 COSATI CODES 18. SUBJECT TERMS (Continue on reverse if necessary and identify by block number)

* FIELD GROUP SUB-GROUPInertial Surveying Litton Auto Survey System II (LASS II

19. ABSTRACT (Continue on reverse if necessary and identify by block number) i

The Defense Mapping Agency has been active in the field of inertial surveying since 1975.

More recently, delivery was taken of two Litton Auto-Surveyor System II (LASS-If) units.

"This paper descirbes the testing carried out with these two systems over theieyenne IPS Test Course and presents a statistical analysis of the results obtained. New

*thematical models for the RMS error of interpolated position and height as a functionof traverse length are proposed.

* SEP 20

20 DISTRIBUTION /AVAILABILITY OF ABSTRACT 21. ABSTRACT SECURITY CLASSIFICATIN A

0 UNCLASSIFIED/UNLIMITED Q SAME AS RPT. [3 DTIC USERS UNCLASSIFIED22a NAME OF RESPONSIBLE INDIVIDUAL 22b. TELEPHONE (Include Area Code) 22c OFFICE SYMBOL

T.,ajil- PfaF r 307-775-311 . GSST

DD FORM 1473, 84 MAR 83 APR edition may be used unti exhausted SECURITY CLASSIFICATION OF THIS PAGEAll other editions are obsolete

...........-.... %.. %...........

RESULTS OF OPERATIONAL TESTING

OF LASS-11 SYSTEMS

by

L. Pfeifer

R. Tyszka

Defense Napping AgencyWashington, DC 20305-3000

* USA

Aoops9sOl o

1TS GRA&IDTIC TAB UJ

Unan~nouncedjutifictioI-

Availability Codes

F Avail and/or

DisL

Spucial

4 SE298D 3401

ASS 920 006 *

1. INTRODUCTION

The Department of Defense, Defense Nap ping Aqency (DA) hasoperated inertial surveying systems since f979 when a Litton Auto-Surveyor System (LASS) unit was acquired. Designated the InertialPositioning System One (IPS-1), this unit is still operational in itseleventh year of service and Is frequently utilf:rd for routine sur-

, veying tasks. Experience was also gained with extensive testing andlimited operational use of a prototype of the Honeywell GEO-SPIN sys-tern, known in DNA as IPS-2 (1979-82), and with the standard U.S. ArmyPosition and Azimuth Determining System (PADS) of which two unitswere acquired in 1982.

These two PADS units have now been upgraded to the configurationof the Litton Auto-Surveyor System 11 (LASS-II) and bear the DMA des-ignation of IPS-3 and IPS-4. The purpose of this paper is to presentthe results of operational testing of these LASS-11 systems carriedout in the span of the past year over the Cheyenne IPS Test Course.

2. LASS II

-The Litton Auto-Surveyor System II is an upgraded version of thePosition and Azimuth Determining System intended for applications insurveying and geodesy. The hardware features which set a LASS-Iapart from a PADS are (1) gyros screened to tighter performance spec-Ifications, (2) modifications to the control/display unit (CDU), (3)addition of a digital tapedeck and associated interface electronics,and (4) a coat of paint change from olive drab to off-white which,aside from giving the LASS-I a distinctive look, has the functional-

Figure 1. Litton Auto-Surveyor System it (LASS-Il)

%~~ ,% ,.

* 5 *~**~ * , ~~w. ~ S 9 . -~;** ~ **k .

a . * -'c

S Am

Figure 2. Typical Lass-Il installation in a survey vehicle.

Figure 3. Control/display unit (CDU) of the LASS-11 system.

,,purpose of aiding with the management of heat dissipation.,.Figures 1and 2 show the LASS-11 unit, and Figure 3 shows the CDU. "

3. THE TEST TRAVERSE

Cm

9m

KILOO UTERI

46

* Gas

* 45

dei uvySudo G S loae tFE are i oc aeI

Fgr 4.Cheyenne IPS Test Coursndtetettavre

d Th SureoSuaron of(G SS -locat s oedra FShoWa rn t ArForce Base InChynde,1 Wymi ng. Aee alibrationan test cors ;TCon(1isting, at thepresenth imeer-49suvee paintsrais beenolad both aogThe oadec

tive of the present test was to ascertain the performance of theLASS-11 system over a traverse line of operationally significantlength and configuration. The traverse selected reflects the of ten-encountered operational scenario of a generally straight line ofprogress with majo' meandering along the way. Comprising 24 controlpoints, It extends in a generally north-south direction from Cheyenneto just south of the town of Chugwater, with a path length of 86 kmsee Figure 4. The elevation range along the traverse route is 230 a.

SI%

W

Since part of the overall objective was to document system accura-cy under extreme conditions, the test traverse is the longest trav-erse which could be laid out within the confines of the Cheyenne IPSTest Course end also fully conform to the inertial traverse designcriteria specified in LITTON (1982). These criteria require that (1)the traverse be fully contained within the radius of 100 km from thestarting point, and (2) that no point of the traverse deviate later-ally from the straight line connecting the endpoints of the traverseby more than one-third of the distance to the nearer endpoint. TheLitton criteria, in effect, require that the traverse be containedwithin a diamond-shaped area no longer then 100 km and no wider thanone-third of its length. Such diamond-shaped area which contains thetest traverse is also indicated in Figure 4.

4. DESCRIPTION OF THE TEST

As is the usual case with acceptance testing, the acceptance testsof the IPS-3 and IPS-4 systems were preceded by exacting bench androad calibrations of the respective hardware, and were carried out bypersonnel possessing special skills and extensive experience with theoperation of LASS-11 systems. In contrast, the intent of operationaltesting is to carry it out as one would any other routine assignment,i.e., without any special preparation of the equipment and utilizingroutinely trained personnel. Accordingly, the test runs made for thispurpose were accoqlished whenever the systems were available and with-out recourse to selected operators.

Four forward-and-reverse "double" runs were made over the testtraverse with each system, using 3.6 minutes as the nominal zero-velocity update (ZUPT) interval. In every case, the traverse run waspreceded by a one-hour alignment of the system and by a short "dummy"traverse leg. The time required to run the 86-km traverse in onedirection ranged from I hr 35 min to 2 hr 25 min and averaged I hr 56min.

IPS-3 FORWARD RUNS

-.-

Neea SA"t of travel ad inppe-~te "-,ae (ha)l

Figure 6.

jaw,

6a -R '-.Cin

IPS-3 REVERSE RUNS

- --- POSITION ~"

30 3s

General direction of travel and epro.I.-to Matance (ha):

Figure 6.

_______RUNS SMOOTHED

Figure 7.

- - - - - -e-- -

5. IPS-3 DATA

Figures 5 and 6 show the JPS-3 unsmoothed data in the form of themagnitude of position and height deviation of the forward and reverseruns, respectively. Temptation is here resisted to call these "raw"data, as the data in question has been acted upon, and thereby irre-versibly altered, by the Kalman filter built into the LASS-1I on-linesoftware; raw inertial data is not e,:cessible. In each case, a 3-kmdummy leg was run due south between tle second and first poirts oftraverse, whereupon the traverse proper was started due .1orth.

Figure 7 shows the magnitude of position and height error in the,corresponding smoothed (i.e., adjusted) data of the eight (four for-ward and four reverse) "single" runs. The strikingly better perfor-mance of the vertical channel is immediately apparent.

6. IPS-4 DATA

Figures 8 and 9 show the IPS-4 unsmoothed data. Of immediatenote are the much larger position deviation ramps of the forward runs,not duplicated in the reverse runs, while the height deviation rampsare not significantly different from those of IPS-3.

ISE

IPS-4 FORWARD RUNS

-- POSITION-

S. -_ _ -------

•-- HEIGHT -- : ,

-$ -IA 1. . .L. UM

Figure 8.

i -- j

.:€ 4 " ...... . _- , *. . *.. . . . ... . .. .

IPS-4 REVERSE RUNS

- -- -- POSITION-

Caneral. dirctio of a troal end approximate distance (ba)t

i~~~~! 6!s 0l ~ ~3 'uF 3I. 1 Val. 49

Figure 9.

P'N\

POSITION

HEIGHTFigure 10.

Here different operators are Involved and, instead of running astraight 3-km course as was the case with IPS-3, the du leg was runfrom a control point near the place of alignment over approximately 5 kmof tortuous route to the starting point of the traverse. This and thelikely existence of a relative position error between the two controlpoints involved may have caused an azimuth error to be introduced intothe system, as opposed to removing residual azimuth error left afteralignment, which is the whole purpose of running a dumW leg.

Another possible explanation is that the large ramps in positiondeviation are caused by an intrinsic characteristic of the IPS-4hardware, such as a significantly greater gyro drift. Additionalcontrolled test runs with IPS-4 are needed to resolve this issue.

Figure 10 shows the corresponding smoothed data, with position

errors somewhat greater than those of IPS-3, while height errorsappear to be smeller than those of IPS-3.

7. RNS ERROR OF A SINGLE RUN

Combining the smoothed IPS-3 and IPS-4 data (total 16 single runs),root-mean-square (RMS) errors in position and in height were computedfor each of the 22 intermediate points along the test traverse. Theseappear plotted as a function of path distance in Figure 11 togetherwith the respective RMS error models given in L[TTON (1982) and thecorresponding new, "elliptical" RMS error models proposed as a result

of this analysis.

RMS ERROR OF SINGLE RUN

331

EIIIIra ct o

Jao Fiur I

-I-.- ---. Im ~ -

POSITION

H EIGHT "'ft f of " -"wm

Figure U1.

16

The Litton RMS error models, giving the respective RMS errors(RMSE) in meters, are formulated as follows:

RMSEpostion 0.15 + ' (1)

S

F44SEheight = 0.12 + S (2)

where S is the straight-line distance from the point of Interest tothe nearer endpoint of the traverse, in meters. In the case of posi-tion error, the Litton model has been found to be too optimistic forpoints close to the endpoints of the traverse and too pessimistic forpoints in the middle of the traverse, while in the case of heighterror the Litton model turns out to be grossly pessimistic, by a fac-tor of 4 or 5, as is readily apparent from the lower graph of Figure11. Also, being defined in terms of the straight-line distance S,the Litton model is not appropriate for meandering traverses.

The proposed, elliptical RMS error models, giving the respectiveRMS errors in meters, are defined as follows:

RMSE 1 d(D - d) (single-run) (3)position 20,000

RMSEhelght = ,0 / d(O - d) (single-run) (4)

where d is the path distance from either endpoint of the traverse tothe point of interest, and D Is the total path length of the traverse,both in meters. The appropriateness and goodness of fit o! thesemodels may be judged from the graphs of Figures 11 and 13. Where themodel curve falls below the observed RMS error, it must be kept inmind that the data has not been purged of outliers. In each instance,the rejection of the largest deviation in the set of 16 at each pointbrings the observed RMS error in line with the proposed model.

Also shown in Figure 11 is the magnitude of the position andheight error of the means of all 16 (8 forward and 8 reverse) singleruns. This is indicative of the noise present in the geodetic con-trol; perhaps half of this error is attributable to uncertainties inthe conventionally surveyed positions and elevations.

8. RMS ERROR OF A DOUBLE RUN

It is the standard practice to double-run inertial traverses.that is, to follow the "forwaird" run imiediately with a run made inthi nppo-,ite (rtrpetion, touchinq uopon the traverse points in revor%-'

P; en e" U * Po ".v*' -rs'rre' r1,n Tt -,-An o'~fl to,,P resr'( t I ,1

smoothed latitudes. iungituwies. and .eijhts are then tahen as im-

proved resultS. in the rase of )eiqht error. this Improvement fol-lows the laws of statistics for random n.,mbers. in that the RMS errorof the mean is reduced by the factor i12.

In the case of position error, as is readily seen In the uppergraph of Figure 13, this improvement is significantly greater. TheRMS error of position (where the position errors are computed as thevector sum of the deviations of the corresponding mean latitude andmean longitude pairs) is reduced by the factor 1/2 as a round figure.

MEANS OF FORWARD-AND-REVERSE

DOUBLE RUNS

Q*-u 0.5n

POSITION

HEIGHTFigure 12.

RMS ERRO F OBE RUN

POSITION

HEIGHT

Figure 13.

LN-A

W ".".II z 7 1 7 7 T:

Note that the letter E is used in Figures 13, 14, and 15 to iden-tify the elliptic RNS error model curve given by equation (3) for RKSerror in position and by equation (4) for RNS error in height. Theother R14S error model curves shown are those corresponding to E/V7and E/2, and are so identified.

To complete this investigation, single runs were pair-wise com-bined and the RNS errors of the respective means were computed forruns made (1) in the same direction with the same equipment, (2) inthe same direction with different equipment, and (3) in opposite di-rections with different equipment. The results are shown in Figure15, where the RHS error of a single run (see Section 7) has also been

. plotted for comparison.

An . ./ "" Y. I2 .om

_-5.

A; ._, Ol.e- t the := dir ction with the een 7yste

i the e drction with different systeus

i opposite directions with dilferent systems

POSITION

a,0.

.:. HEIGHT

Figure 15.

The following may be abstracted from Figure 15: (1) Running thetraverse twice in the same direction with the same equipment producesonly a negligible improvement in either position or height accuracy.

" •(2) Running the traverse twice in the same direction with differente is not much better for improvingpos tion accuracy; forilT9q.hFowever, it achieves accuracy Improvement comparable to thatof the forward-and-reverse double run. (3) Running the traversetwice in opposite directions with different equipment does achieveabout three-our hs of the position accuracy improvement of the for-ward-and-reverse double run and height accuracy improvement, again,comparable to that of the forward-and-reverse double run.

It Is apparent that the errors affecting the interpolation ofposition with an inerti&l surveying system are predominantly system-atic, made up of components which are both equipment-specific and

d% .

. 4 *j 4 * ... %w! .'*-.....

Note that the letter E is used in Figures 13, 14, and 15 to iden-tify the elliptic RNS error model curve given by equation (3) for RMSerror In position and by equation (4) for MS error in height. Theother RNS error model curves shown are those corresponding to E/V2and E/2. and are so identified.

To complete this investigation, single runs were pair-wise com-bined and the R14S errors of the respective means were computed forruns made (1) in the same direction with the same equipment, (2) inthe sam direction with different equipment, and (3) in opposite di-rections with different equipment. The results are shown in Figure15. where the RNS error of a single run (see Section 7) has also beenplotted for comparison.

/ /.J.

.S or b t sC m

Is the se directlo Witb aiflerent systemsim opposite diCecCtInG with different systems

POSITION

0..30

.. HEIGHT

Figure 15.

The following may be abstracted from Figure 15: (1) Running thetraverse twice in the same direction with the same eguipnent producesonly a negligible improvement in either positiTon-ir height accuracy.(2) Running the traverse twice in the same direction with differente 9ot is not much better for improving position accuracy; forheighthowever, it achieves accuracy improvement comparable to thatof the forward-and-reverse double run. (3) Running the traversetwice in opposite directions with different equipment does achieveabout three-fouths of the position accuracy Improvement of the for-ward-and-reverse double run and height accuracy improvement, again,comparable to that of the forward-and-reverse double run.

It Is apparent that the errors affecting the interpolation ofposition with an Inertial surveying system are predominantly syst-.tic, ulde up of components which are both equipment-specific and

Es

%'b I

trajectory-specific, and of such nature that tlheir optimal reductiontakes place when means are taken of the smoothed results of runs madeIn opposite directions with the same equipment. This would seem to

;o..A' indicate that the trajectory-specific component Is dominant, with thegravity disturbance field the likely source. The same is observed toapply, albeit to a significantly lesser degree, to the interpolationof height. Deeper understanding of this complex error propagation

*, must wait more data and another study.

9. SWUINRY

This paper has presented, in graphical form, the errors in posi-tions end heights interpolated with two DNA-owned LASS-1I systemsover a specially designed test traverse for the purpose of operation-a testing of the LASS-11 systems. Based on subsequent data analysis,new mathematical models for the RMS errors in Oosition and in heightare proposed, which are better suited for operational configurationsof inertial traverses, as well as more accurately descriptive of theRNS errors actually observed.

It must be kept in mind that the error models in question giveRMS (i.e., one-sigma) values, and hence isolated errors up to threeor four times their magnitude may occur. Also, these error modelsare specific to LASS-1I systems operated in a light truck type vehi-cle with nominal 3.5-minute ZUPT Intervals. Their validity for e.g.helicopter operation and/or longer ZUPT intervals Is yet to be veri-fied.

10. ACKNOWLEDGEMENT

The work on this paper was accomplished by the authors while as-signed to the Techniques Office (GSST) of the Geodetic Survey Squad-ron (GSS), F.E. Warren Air Force Base, Cheyenne, Wyoming. LudvikPfeifer is a geodesist and project engineer for the development ofinertial surveying systems. Captain Robert Tyszka, USAF, is a carto-graphic/geodetic officer. The inertial survey data used in this testand analysis were acquired by personnel of Geodetic Survey SectionThree (GSSOG3); their essential contribution to this project is here-by acknowledged. The Geodetic Survey Squadron is a division of theDepartment of Geodesy and Surveys, Defense Mapping Agency Hydrograph-ic/Topographic Center, Washington, DC.

REFERENCES

ITECH: The Litton Autosurveyor I (Dash II) - Test Results. Unpub-lished report. International Technology Limited, Englewood, Colo-rado. 1963.

LITTON: Technical Description - Litton Auto-Surveyor System II (LASS-II). Litton guidance ane Control Systems, Woodland Hills, Call-fornila, 1962.

M... . 4. N.: Evaluation of an Inertial Surveying System. The Aus-

tralin Surveyr, Vol. 32 (2), pp. 78-96, 1964.

77 e

11-8

DTIC~