Stochastic Vehicle Mobility Forecasts Using the … · Mobility Model Report 2 Extension ......

AD-A26B 79 Technical Report GL-93-15C ) 11111111July

199

US Army Corps h Jul 1.

of EngineersWaterways ExperimentStation

Stochastic Vehicle Mobility ForecastsUsing the NATO ReferenceMobility Model

Report 2Extension of Procedures and Applicationto Historic Studies

by Allan Lessem, Richard AhivinGeotechnical Laboratory

Paul Mlakar, William Stough, Jr.JA YCOR

Approved For Public Release; Distribution Is Unlimited DTICS SEP 01191

9332041

Prepared for Headquarters, U.S. Army Corps of Engineers

The contents of this report are not to be used for advertising,publication, or promotional purposes. Citation of trade namesdoes not constitute an official endorsement or approval of the useof such commercial products.

PRINTED ON RECYCLED PAPER

Technical Report GL-93-15July 1993

Stochastic Vehicle Mobility ForecastsUsing the NATO ReferenceMobility Model

Report 2Extension of Procedures and Applicationto Historic Studies

by Allan Lessem, Richard Ahivin

Geotechnical Laboratory

U.S. Army Corps of EngineersWaterways Experiment Station3909 Halls Ferry RoadVicksburg, MS 39180-6199

Paul Mlakar, William Stough, Jr.

JAYCOR1201 Cherry StreetVicksburg, MS 39180

Report 2 of a seriesApproved for public 'ease; distribution is unlimited

Prepared for U.S. Army Corps of Engineers

Washington, DC 20314-1000

Under Task No. AT40-AM-01 1

Monitored by Geotechnical LaboratoryU.S. Army Engineer Waterways Experiment Station3909 Halls Ferry Road, Vicksburg, MS 39180-6199

US Army Corpsof Engineers_,Waterways Experiment N

aterasEprmetSainCtaoion-ublalnDt

OWTIC" I N4al1 INAC,

Stchatc eilembiiyfoeassuin heNT Referece Moblity ~C

neers, moUordbyGoecnca abrtry .S r y Engineer

WANMWAUMqlIm IAInO

17,. l. 8c.- Tehia ec;G-31 rept. : 2) l~41

n lud biligrphca rfeenes

I'"WAWPWMffAW

Waterways Experiment Sitatiorcasloging-in-Publication Data

Stochastic vehicle mobility forecasts using the NATO Reference MobilityModel. Report 2, Extension of procedures and application to historic stud-

ies / by Allan Lessem ... [et al.]: prepared for U.S. Army Corps of Engi-neers, monitored by Geotechnical Laboratory, U.S. Army Engineer

Waterways Expoeiment Station.170 p.: Ill.; 28 cm. -- (Technical repo.-; GL-93-15 rept. 2)Includes bibliographical references.1. Transportation, Military -- Forecasting - Statistical methods. 2. Ve-

hicles, Military -- Dynamics -- Data processing. 3. Stochastic processes.1. Lessem, Allan. It. United States. Army. Corps of Engineers. Ill. U.S.Army Engineer Waterways Experiment Station. IV. Series: Technical re-

port (U.S. Army Engineer Waterways Experiment Station) ; GL-93-1 5 rept. 2.W34 no.GL.-93-15 rept.2

Contents

Preface ......................................... v

Executive Summary .................................. vi

Conversion Factors, Non-SI to SI (Metric) Units ................ vii

I-Introduction ..................................... I

Background ..................................... 1Purpose ....................................... 2

2-Summary of Prior Work ............................. 3

The Components of the Stochastic Forecast ................. 3Sensitivity Analysis ................................ 3Quantification of Sensitivity ........................... 5The Screening Process .............................. 5The Error-Magnitude Scenario ......................... 5The Monte Carlo Speed Simulation ...................... 6Speed Profiles ................................... 6Mission-Rating Speeds .............................. 8

3-Extension of Procedures ............................. 9

Origin of the Fingerprint Anomalies ..................... 9Change of Compnuting Environment ...................... .14Parameters and Curve-fits Considered .................... 14A Global View of NRMM Parameter Sensitivity .............. 16

4-Application to a Historic Study: HMMWV Candidate Vehicles .... 23

Introductory Remarks .............................. 23The Historic Study ................................ 23Scope and Methods of the Application .................... 24Fingerprints and Speed Profiles ........................ 33Missions and Mission Rating Speeds ..................... 55Comparison of Historic and Stochastic Mobility Fore:4asts ........ 57

5-Application to a Historic Study. FOG-M Candidate Vehicles ...... 65

The Historic Study ................................ 65Scope and Methods of the Application ..................... 66Sensitivity Study .................................. 70Fingerprints and Speed Profiles ........................ 72

iii

Missions and Mission Rating Speeds ..................... 131

Comparison of Historic and Stochastic Forecasts .............. 134

6-Conclusions and Recommendations ...................... 143

Conclusions ..................................... 143Recommendations .................................. 143

References ........................................ 145

SF 298

0

iv

Missions and Mission Rating Speeds...................... 131Comparison of Historic and Stochastic Forecasts ................ 134

6-Conclusions and Recommendations....................... 143

Conclusions...................................... 143Recommendations......... ......................... 143

References......................................... 145

SF 298

iv

Preface

Personnel of the U.S. Army Engineer Waterways Experiment Station(WFS) conducted this study during the period February 1992 thre!lgh July1992 under ZdTE Work Unit No. AT40-AM-011 entitled "Stochastic Afizy-sis Methodology for NRMM."

The study was conducted under the general supervision of Dr. William F.Marcuson m, Director, Geotechnical Laboratory (GL) and Mr. Newell R.Murphy, Jr., Chief, Mobility Systems Division (MSD). Dr. Allan S. Lessemdevised the stochastic methodology, guided the development of software byMr. Richard B. AhIvin, and made one of the applications to historic studies.Dr. Paul Mlakar and Mr. William Stough, JAYCOR, contributed statisticsexpertise and made the other application to historic studies. The report wasprepared by Dr. Lessem with the exception of Chapter 5 which was preparedby Dr. Miakar and Mr. Stough.

At the time of publication of this report, Director of WES wasDr. Robert W. Whalin. Commander was COL Bruce K. Howard, EN.

Aecession For

IRIS RA&IDTIC TABUnanmounced dJustifioation

ByDIotribut lon/Availability $94su

Avail and/or

Dist Special

V

A_____________________________ I _________

Executive Summary

This report is the second in a series that documents the conversion of theNATO Reference Mobility Model (NRMM) from a deterministic code to astochastic one. The first report described basic concepts and procedures anddealt with relatively small quantities of data as methods were "bootstrapped"into existence. This report describes the extension of the procedures to realis-tic quantities of data.

In addition, the methods are applied to two historical mobility assessmentsthat were influential, years ago, in the procurement of some Army vehiclesnow in the inventory. When the studies were originally done, NRMM wasused in a completely deterministic way and no consideration was given to theperturbing effects of uncertain data and of scatter associated with field-data-based algorithms. The analyses were repeated (albeit with greatly limitedscope) using stochastic mobility forecasting methods as a means of introducingthe NRMM user community to these methods in an inherently interesting way.

The original studies provided rankings (either explicit or implied) of vehi-cles against certain specified operating missions. The outcomes of therepeated studies suggest that those rankings were basically correct even in thepresence of data uncertainties and that the stochastic methods provide helpfulinsights into the risks assuciated with accepting such rankings at face value.

vi

Conversion Factors,Non-SI to SI Units ofMeasurement

Non-SI units of measurement used in this report can be converted to SI unitsas follows:

Multiply By To Obtainhorsepower (500 ft-lb (force) per see per ton 83.82 watts per Iklonewton

(force))

inches 2.54 oentimeters

miles (U.S. statute) 1.609347 kilometers

miles (U.S. statute) per hour 1.609347 kIlometers per hour

pounds (force) 4.448222 newton.

pounds (fore7q) per square inch 6.894757 ktlopesoals

vii

1 Introduction

Background

This report is the second in a series intended to describe the conversion ofthe NATO Reference Mobility Model (NRMM) from a deterministic vehiclemobility forecasting resource to a stochastic one. The following remarks,drawn from Report 1 (Lessem, AhIvin, Mason, and Mlakar 1992), may behelpful to the reader as an orientation.

NRMM is a computer code. used to characterize the ability of ground vehi-cles to move in various operational settings. Based on many years of fieldand laboratory work by the USAE Waterways Experiment Station (WES) andthe Army Tank-Automotive Command, and containing contributions fromNATO members, NRMM considers many terrain, road, and tactical-gapattributes, vehicle geometries, and human factors (Haley, Jurkat, and Brady1979). Its fundamental output is a mobility forecast based on speed predic-tions keyed to specific areal units of terrain and to specific lineal portions of aroad network.

Like many other mathematical models of broad scope, NRMM requires theassembly of a comprehensive dataset. Users of NRMM understand that confi-dence in results is governed by data quality. Informal trials are often made toinfer the effects of variation in important data elements. In addition, it isessential to remember that the algorithmic basis of NRMM is founded mainlyon empirical field studies having unavoidable errors associated with experi-mental control and measurement.

In addition to its service in user communities concerned with vehicledesign, war-gaming, and strategic planning, continuing developments in com-piuter technology are creating an opportunity for NRMM to serve a tacticalrole on the. battlefield. The battlefield setting requires high-resolution data andexpedient dataset preparation. Adaptation of NRMM to this role requires thatits users come to grips with the effects of errors in vehicle and terrain dataand of inherent algorithm errors.

Years of experience with NRMM have resulted in qualitative impressionsof unusual or unanticipated aspects of vehicle performance, both measured andpredicted, as ranges of terrain attributes are studied. It is now desired toformally quantify the variation performance of the model. By "variation

Chwitor I Introduction

performance" is meant the responses of NRMM when some dataset elementsare represented, individually or jointly, as random variables. Random vari-ation can arise from errors of measurement or judgment, and from intentionalvariation in the context of design studies. In addition, errors associated withregression-line representations of empirical data contribute to variations inNRMM outputs.

NRMM is an equilibrium model: supply it with all the numbers it needs tomake a speed prediction and its prediction is applicable to the one terrain unitand vehicle represented by those numbers. No neighboring terrain units exertan influence; no past prediction influences the present one. Each terrain-unit/vehicle combination has a unique equilibrium speed. Considered in a map-wide context there are many such equilibria, and no characteristic predictionpatterns emerge. Our approach to the determination of NRMM variationperformance is, therefore, project-specific. Each time NRMM is called uponto make a speed prediction, its variation performance is determined for thatterrain unit and that vehicle. The trick is to make this determination effi-ciently, to state outcomes clearly, and to integrate meaningfully over the manyterrain units that compose a mobility map.

Purpose

WES has undertaken the task of making NRMM capable of deliveringstochastic mobility forecasts in which the impacts of data and algorithm uncer-tainties, large and small, are clearly evident in the model's predictions ofvehicle speeds. The principal benefit will be the presentation in numericalterms of the quality of NRMM vehicle speed prediction products. With thisinformation, tactical decisions which depend upon vehicle performance can bemade with pertinent assessments of risks.

The purposes of this report are to extend the basic procedures presented inReport I of this series and to apply the procedures to two historic studies thatinfluenced vehicle procurement decisions in the past. These applications areintended to attract the attention of the user community to the significance ofstochastic mobility forecasts and to see whether or not the mobility assess-ments that influenced prior decisions would undergo substantive changes whendata and model uncertainties are considered explicitly.

2 Chaptoe 1 Introducuon

2 Summary of Prior Work

The Components of the Stochastic Forecast

The products delivered as a stochastic mobility forecast consist of fouritems: an expected value speed map, a "fingerprint," an expected value "mis-sion rating speed," and a range for the mission rating speed. The speed mapis a graphical presentation of nominal predicted speeds for one vehicle operat-ing according to one sceario on one terrain map (consisting of hundreds tothousands of terrain units) made under the assumptions of error-free data andan error-free model. It is the product obtained from NRMM at the p.esenttime. See Figure la for a representative example. The fingerprint is a graph-ical presentation of the error performance of NRMM specific to the one vehi-cle and the one terrain map and is capable of quick visual comprehension.Each nominal predicted speed is associated with related minimum and maxi-mum predicted speeds. The fingerprint plots the minimum and maximumspeeds against the nominal speeds. The greater the departure of the finger-print from a straight line of unit slope, the greater the error associated withthe speed predictions. Clustered errors are easy to spot. See Figure lb. hemiasion rating speed is a concept used by NRMM analysts who postulate amission "profile" expressing a mixture of on-rosd and off-road percentages anavoidance of worst terrain percentages to arrive at a one-number measure ofvehicle performance on the terrain map. This usef'i concept is preserved andextended by expressing its range thereby indicating in an integrated and quan-titative way the quality of the entire NRMM speed map. Extended discussionsof the concepts underlying the products of a stochastic mobility forecast arepresented in Report 1 of this series. The descriptions which follow are inten-tionzlly brief.

Sensitivity Analysis

The first step in a stochastic mobility forecast is the determination of thesensitivity of NRMM to uncertainty in its numerous data elements and tovariation in its experimentally derived algorithms. Sensitivity is found to varywidely among vehicles and terrains and is thus very much project-specific. Infact, it is tempting to require sensitivity to be determined on a terrain-unit-by-terrain-unit basis within a given map, but consideration of the large number of

3Chaptat 2 Summery of Prior Work

$30 AREA. LAUTERBACHVEHICLE. HlYIVV-t1S98

-- ft SEASON- WETCONDITION- SLIPPERY

0 ~SPEED (MPHO

.Jlm- 0. - 5-0

a. The speed map

M998 on 5322 (wwet-slipr-y)Spoed roge for 9 errescm

70- +

60-

E

v 50-

* 40-

E00

20

44

units to be found in a map (1500 to 5000, typically) quickly dispels notions ofsuch detailed resolution.

Sensitivity is determined parameter by parameter with each being treated asindependent of any other. Issues of joint sensitivity were examined and foundto be of secondary importance. An especially uncomplicated procedure,called "3-point extremum analysis," was devised therein to examine sensitivitybased on nominal, maximum, and minimum values of each independent vari-able parameter. The maximum and minimum values are tken as the nominalvalue plus and minus 10 percent, respectively. For the experimental curve-fitsextensively used in NRMM, sensitivity is expressed in terms of nominalregression values and maximum and minimum values taken as 14 percent ofthe nominal values, respectively.

Quantification of Sensitivity

In order to prepare the way for a screening process that discards insensitiveparameters, a means for numerical specification of sensitivity was devised thatis at once simple and effective. Nominal, maximum, and minimum NRMMpredicted speeds corresponding, for the most part, to nominal, maximum, andminimum parameter values are combined as the ratio of maximum minusminimum speeds to the nominal. This ratio is not permitted to exceed 2.When the ratio is evaluated for a given vehicle in each terrain unit of the mapand averaged over all the terrain units, it yields numerical values that varywidely among the parameters considered and discriminates quite well amongsensitive and insensitive parameters. The ratio was termed a "rank indicator."

The Screening Process

Evaluation of the sensitivity of NRMM parameters in the project-specificsetting of a given vehicle and a given terrain yields as many rank indicators asparameters examined. By selecting the maximum rank indicator and multiply-ing its value by 20 percent, a threshold value was defined below which cor-responding parameters were viewed as insensitive. During all subsequentprocedures in the stochastic mobility forecast, sensitive parameters are treatedas random variables and insensitive parameters are treated as constants.

The Error-Magnitude Scenario

An error magnitude scenario is a list of the sensitive pazameters and curve-fits and the actual nature of the variation to be assigned to each. During thescreening process, each parameter was varied plus and minus 10 percent ofnominal and each curve-fit was varied plus and minus 14 percent of its regres-sion value. During the subsequent Monte Carlo simulation, the opportunity is

5Chapter 2 Summary of Pror Work

provided to specify the actual variation type and ranges on an individual basisfor the parameters and the actual standard deviations for the curve-fit errors.

The Monte Carlo Speed Simulation

The Monte Carlo analysis of predicted ibpeeds, wherein the screenedparameters and curve-fits are vaiied jointly and independently and probabilitydensities are determined for the speeds predicted for each terain unit, is themajor element of analysis leading to the stochastic mobility forecast. The per-terrain-unit speed probability densities and data specifying the mission profileare the raw materials from which an analysis of mission rating speeds andtheir ranges can be made. Other outputs from the Monte Carlo simulation area listing of nominal speeds by terrain unit from which the speed map isobtained and maximum and minimum speeds by terrain unit from which thefingerprint is made. For conceptual simplicity and to bound NRMM errorperformance, initial work with the mission rating speeds was based on maxi-mum and minimum terrain unit speeds rather than the speed probabilitydensities.

Speed Profiles

The mission rating speed is approached through the "speed profile," auseful concept worked out early in the history of NRMM. A speed profile isspecific to a given vehicle/terrain/scenario combination. NRMM is used toform a sequence of records each of which shows the area and the predictednominal speed for individual terrain units. These records are sorted indescending order by speed thus identifying the terrain units in which vehicleperformance is "best" and 'worst." The sum of terrain unit areas from thefirst record (which represents "best") to the Nth record divided by the sum ofall areas defines the fraction of map area represented by the first N records.When the sorted speeds are plotted against this fraction, tLr result is a speedprofile based on terrain unit speeds. NRMM calls it a "speed-in-unit" profile.See Figure 2a for an example. When the area-weighted averages of the firstN speeds are plotted against the fraction, an "average speed profile" is pro-duced, as in Figure 2b. Assuming that tactical usage of the vehicles willstress deployment over the "best" terrain units, the profiles allowquantification of what is meant by best. The plots of Figures 2a and 2b showthat

"as more terrain is used, or as the challenge level goes up, themore difficult the terrain becomes, and the average speed thatthe vehicle can attain over that terrain, and its average speedon that terrain and all bctter terrain, decreases. At somepoint, the challenge level is so high that the vehicle encountersvery difficult terrain, and NOGO's occur, shown as 0.0 mph."(Unger 1988)

6 Chapter 2 Summary of PNor Work

M1 13 on 5322 (off-rood) Ml 13 on 5322 (off-rood)

30 . -o *01 - - ,eI

a. Deterministic speed-in-unit b. Deterministic average speed

M113 on 5322 (off-rood) M113 on 5322 (off-rood)*e "-o - ~ ~ 0

30 *

21.' so.js

.. C , - . . ..

0 02 94 O4 O0 02 A. 04 060

S.. . .0- P. S

c. Stochastic speed-in-unit d. Stochastic average speed

Figure 2. Speed profiles

Stochastic orientation of NRMM requires the development of stochasticspeed profiles based not only on the nominal predicted speeds but also on theminimum' and maximum' predicted speeds. The very same computationalprocedures are used and result in plots like those shown in Figures 2c and d.In effect, range limits that bound NRMM error performance are placed on thetraditional speed profiles.

Anywhere these terms appear, it is possible via Monte Carlo simulation to replace them

with mean ± one or two standard deviations whichever the user wishes to calculate.

Chapter 2 Summary of Prior Work 7

Mission-Rating Speeds

Speed profiles form the basis for the calculation of the mission ratingspeeds. A mission rating speed is, as mentioned earlier, a one-number mea-sure of vehicle performance that factors in the parameters of a tactical missiondefined on a terrain map. The parameters are (a) percentages of total operat-ing distance spent on-road and off-road and (b) percentages of the best terrainchallenged and road units so occupied. Thus, a mission might be character-ized as 80 percent on-road in the 75 percent best road units and 20 percentoff-road in the best 10 percent areal units, and the mission rating speed wouldconvey an overall speed for these percentages by entering the on-road speedprofile graph at a total length fraction of 75 percent and the off-road profilegraph at a total area fraction of 10 percent and appropriately combining thetwo speeds read from the profiles. There are several ways to make the com-bination depending on the depth of resolution desired. For example, are roadsto be considered separately as primary, secondary, and so forth; are prede-fined "tactical mobility levels" to be considered; are time penalties for cross-ing linear fear ,s to be considered? See Robinson, Smith, and Reaves (1987)for insights and typical applications.

NRMM applications make use only of the average speed profiles to com-pute mission rating speeds and leave the in-unit profiles for other purposes.Stochastic mission rating speeds are derived from the stochastic average speedprofiles by evaluatiTg the defining equation three times: first using the nomi-nal values of speeds taken from the speed profiles, and then the minimum andmaximum values. These define the nominal, minimum, and maximum mis-sion rating speeds. The range in the mission rating speeds so computed fromminimum to maximum constitutes, together with the nominal mission ratingspeeds, a measure of NRMM error performance for the given terrain/vehiclecombination and the given mission.

8 Chapter 2 Summary of Nror Work

3 Extension of Procedures

Work discussed in Report I developed procedures and illustrated them interms of 19 parameters and curve-fits for off-road terrain units, and 16 foron-road units. There are, of course, many more to be found in NRMM; thus,the main thrust of procedural extension was to be in the direction of greaterquantities. Subsequent discussion will detail how 90 parameters and curve-fitsare now accommodated in a supercomputer environment. But before gettingto that, it would be of interest to present some work that went far to illumi-nate the rather strange attributes of the fingerprints. It was the need to under-stand these attributes that consumed developmental energy following the workof Report 1.

Origin of Some Fingerprint Anomalies

By examining in detail the performance of NRMM in the vicinity ofassorted spikes and other distinctive features of some of the fingerprints, anidea came to mind that an important contributor to the anomalies was therather nonlinear tractive-force versus speed (TFS) curve specified for eachvehicle. For example, Figure 3a shows the TFS curve for the Ml 13. Indi-vidual curved segments correspond to different transmission gear ratios.Figures 3b and c show the fingerprints for the M 113 on the Lauterbach andSchotten quads in a highlands area in West Germany. The first is an off-roadsetting; the second, on-road. The clustered nature of NRMM speed predic-tions is quite apparent.

Motivated by a hunch that the abruptness of the nonlinearities in the TFScurve were important to this behavior, the curve was smoothed (it is, ofcourse, still nonlinear) as shown in Figure 4a. The fingerprints of Figures 4band c resulted. Much of the clustering in the form of bulges and spikes waseliminated. Looking at another vehicle, original and smoothed TFS curvesand resulting fingerprints for the MI are shown in Figures 5 and 6. Most ofthe clustering was thus accounted for.

9Chapter 3 Extension of Procedues

Troctive Force vs Speed9113

161

17"W-I?°

-4.-

l'"Is

S 2-

7.

32.

otO s 4 a.

'U"' @.Y

a. Tractive force versus speed for M1 13

M113 on 5322 (Louterboch)p- -$. fr* OOW

40

f 44

20 +St * -

i.0So4

0 0

0 ;0a

b. Fingerprint for M1 13 on Lauterbach Quad

Ml13 on 5520 (Schotten)'U. .W Ub 4 g .e

46.

40, +

* 0

30. .- .!+:0I ,o. *. 4 o°

i i

* Is-

o 40 -"G 4

c. Fingerprint for M113 on Schotten Quad

Figure 3. Clustering anomalies in Ml 13 fingerprints

10 Chapter 3 Extenion of Procedures

Troctive Force vs Speed

17

2

14

I"

o to f1 40

OP." -96

a. Smoothed tractive force versus speed for M1 13

Ml 13 (constant power) on 5322

|="ftf t

24. 4e

'ft

0 4 aI 12 11 ;10 4 2

b. Fingerprint for M1 13 on Lauterbach Quad

M1 13 (constant power) on 5520

404'

-40- 40

- O

o ** * , , 4 V ,

o . ol o4

c. Fingerprint for M 113 on Schotten Quad

Figure 4. Elimination of most fingerprint anomalies for Ml13

11Chapter 3 Extension of Procedures

Troctive Force vs Speedm

140

130.!2

I'-

10

40

1 0 1 0 W 4

a. Tractive force versus speed for M1

M1 on 5322 (ILeuterboch)40 .

30-

1do

0 O 30 30 40

b. Fingerprint for M1 or Lauterbach Quad

M1 on 5520 (Schotten)

4.-

40-*

310-4. $**oG

3+

Figure 5. Clustering anomalies in M1 fingerprints

12C

Chapon 35520ns(Sco hPootdune

Tractive Force vs Speed

40

SP-a. Smote rciv oc.ess pe o

3-

I0-

a. Smoothed tractive force versus speed for M1

M1 (constant power) on 5322I~.d .q I 4 u N

415-

40-

36-

I4

b. Fingerprint for M1 on Lauterbach Quad

M1 (constant power) on 552045-

++ 0

** ,*** • ** o°

I

c. Fingerprint for M1 on Schotten Quad

Fi ?ure 6. Elimination of most fingerprint anomalies for M1

13Chaptet 3 Extension of Pr,_cedures

Change of Computing Environment

The work described in Report 1 was done on a VAX 8800, a fully compe-tent mainframe computer. Considerable computational intensity is a part ofstochastic mobility forecasting and even for the relatively small number ofparameters and curve-fits dealt with in that work, computing times of 90 to180 minutes were not unusual when 2000 to 3000 terrain units were involved.In anticipation of the expansion of procedures to include greater numbers ofparameters and curve-fits, operations were transferred to a CRAY YMPsupercomputer. An effortless speedup by a factor of about 5 resulted. Workthen proceeded to include more parameters and curve-fits with no specialattention paid to optimization.

It is appropriate to remark that dealing with a supercomputer may seem tohave little to do with the tactical setting seen as a motivation for the develop-ment of these procedures. One can hardly drag a supercomputer around abattle zone. But using a supercomputer to develop procedures is not a com-mitment to use it to perform the procedures in the field. The supercomputerallows one to deal with realistic ouantities of data as means are sought tooptimize code and to discover, if only by trial and error, those proceduralshortcuts that can lead to short analysis times in the field. For example, oneprocedural shortcut has been glimpsed that will surely yield a substantialsavings in analysis time in the future. Figures 7a to d show what happens to arepresentative fingerprint as fewer and fewer terrain units are involved in theanalysis. The figure shows that essential features are preserved despite reduc-tion in the number of terrain units from the original 2707 to as few as 250.There is a real potential for a streamlining of procedures if reduced numbersof terrain units combined with the 3-point extremum analysis for sensitivitycan be shown to give results identical to those obtained through Monte Carloconsideration of all the terrain units. The streamlined procedures could thenbe expected to perform well with fast desk-top machines in tactical settings asoriginally desired.

Parameters and Curve-fits Considered

The intent of this work was to make each "analog" parameter and, later,each curve-fit available for variation. An analog parameter is one that can beassigned any value over a continuous range. An example is RCIC, the soilstrength parameter that can range from about 10 to about 300 rating coneindex. In contrast to this, "digital" counting parameters are unsuitable forvariation. An example is NAMBLY, the number of traction element assem-blies (i.e. wheels or tracks) to be found on the vehicle.

Work went ahead in stages to incorporate more and more parameters intothe stochastic forecasting procedures, testing along the way. In most cases,the additional parameters were easily merged into the procedures. The onlysignificant problem arose when parameters associated with the computation of

14 Chapter 3 Extension of Procedures

2707 Terrain Units 2000 Terrain Units

40 OW 0 0s- . -. ,

" ! . !*WH~I l•* e .11 * 0

[ "0

"so.

4o 4 l

0 to • a 40 0 to ; I•

-ftO'l Velold- _ _ -- l tod, *.

a. 2707 Terrain Units b. 2000 Terrain Units

1000 Terrain Units 250 Terrain Units

• 0 " 0 0•

... ..

ot ox •I

0 *4 3D* 0 to 0 I

c. 1000 Terrain Units d. 250 Terrain Units

Figure 7. Reduced-dataset fingerprints

TFS curves from engine torque-speed characteristics and from torque con-verter speed ratio characteristics were dealt with. Fingerprints arising fromvariation of these parameters were clearly anomalous. However, when it wasrealized that no historic usages of NRMM have exercised the option to gener-ate "TS curves internally but have accepted the ITS curve as vehicle inputdata, it was decided not to spend the effort to track down the problem.

Instead, all ITS curves would be viewed as certified by the vehicle manufac-turers and not be subject to variation.

Finally, it was realized that the tables of values contained in the vehicledata files which represent the speed constraints due to rough terrain and due toobstacle-induced acceleratis are actually curve-fits in their own right. How-ever, unlike th other curve-fits which consist of coded equations inaccessible

Chap t FS en uson of Procedures 15

to the user, these consist of tables of values that are very accessible. Becauseof their tabular form, it was decided to manage data variation in a mannerfundamentally different from that of the equations. Sensitivity of NRMM toerrors in the ordinate and abscissa values of these tables was studied by hav-ing all values in the table vary by identical amounts. This was tantamount tohaving the entire speed-roughness and speed-obstacle height curves moving upand down or side to side as a manifestation of uncertainty in the tabularvalues.

Parameters and curve fits finally installed as random variables in thestochastic mobility forecasting procedure are listed in tables 1 to 4 as follows:

Vehicle Parameters: Table 1Terrain Parameters: Table 2Scenario Parameters: Table 3Curve-fits: Table 4

A Global View of NRMM Parameter Sensitivity

Having installed in NRMM the capability to treat the aforementionedparameters as random variates, it was of interest, then, to repeat some of thesensitivity analyses discussed in Report 1. That work was done, as mentionedearl:er, with only 19 parameters selected by engineering judgment with theexpectation that these were the model's most sensitive parameters. Wouldthey be upstaged, so to speak, by any of the 71 new parameters under consid-eration? The analyses of Report 1 involved 4 vehicles and 4 terrains. Thevehicles were the M998, a light wheeled utility vehicle; the M977 a heavywheeled transporter; the M113, a light tracked personnel carrier; and the M1,a heavy tracked main battle tank. The terrains were 5322, an off-road quad inthe highlands region of West Germany (WGe); 2726, an off-road qr-d in theplains region of WGe; 3254, an off-road quad in a Jordan desert region; and5520, an on-road quad in the highlands of WGe. Figures 8 to I 1 illustrate thevalues of the sensitivity rank indicators of 87 of the 90 parameters. (The lastthree, numbers 88, 89, and 90, had not been installed at the time the analysiswas performed and were later found to be relatively insensitive.) It was foundthat, indeed, the original 19 parameters whose sensitivities were studied inReport I were among the most sensitive parameters but that the most sensitiveones (in at least some of the cases) had actually been overlooked. These werethe parameters associated with the speed constraints due to ground roughnessand obstacle traversal. Note that these parameters are NOT ALWAYS themost sensitive ones, pointing out once again the wide range of outcomes ofwhich NRMM is capable. The study showed that it is possible to make a gen-eral statement that certain independent parameters will dominate performancepredictions in all terrain unit-vehicle combinations.


Table 1

Vehicle Parameters Capable of Being Treated as Random Variables

Vehicle Armonym

1 Aerodynamic drag coefficient ACO2 Area of one track shoe (per assembly) ASH3 Average cornering stiffness of tires AVC4 Interaxle spacing (per assembly) AXL5 Hydrodynamic drag coefficient CD6 CG height above ground CGH7 CG lateral distance from center line CGL8 Horiz distance, CG to rear axle CGR9 Displacement of each engine (per engine) CID10 Minimum ground clearance CL11 Minimum ground clearance (per assembly) CLR12 Tire deflection (per assembly, par case) DFL13 Undeflected tire diameter (per assembly) DIA14 Combination vehicle draft DRA15 Driver eyeheight above ground EYE16 Final drive gear ratio (1) and efficiency (2) FD17 Combination maximum fording depth FOR18 Height in obstacle height vs speed array FVA19 Speed in obstacle height vs speed array FVO20 Roughness in roughness ve speed array FRM21 Speed in roughness ve speed array VRI22 Track grouser height GRO23 Total net engine power (per assembly) HPN24 Vertical chassis to axle distance (per asembly) HRO25 Maximum pushbar force POF26 Height of pushbr above ground PSH27 Projected vehicle frontal area PFA28 Maximum torque (per engine) QMA29 Avg suspension stiffness (per assembly) RAI30 Tire rim diameter (per assembly) RDI31 Tire revolutions per unit distance (per assembly) REV32 Eft. track bogie radius (per assembly) RW33 Tire undeflected se.tion height (per assembly) SEH34 Tire undeflected section width (per assembly) SEW35 Engine to torque conv. gear ratio (1), eff. (2) TCA36 First-to-last wheel center distance TL37 Tire ply rating (per assembly) TPL38 Tire pressure (per assembly, per case) TPS39 Length of track on ground (per assembly) TRL40 Track width (per assembly) TRW41 Transmission gear ratios and efficiencies TRA42 One-pass fine-grain VCI (per assembly) VFG43 Vehicle maximum fording speed VFS44 Slope sliding speed limit VSL45 Max. swim speed w/o aux. propulsion VSS46 Max. swim speed w/ aux. propulsion VSA47 Slope tipping speed limit VTP48 Tire max. speed limit VTI49 Vehicle length (per unit) VUL50 Winch capacity WC51 Water depth to allow aux. propulsion WDA52 Max. combination vehicle width WDT53 Vehicle weight (per assembly) WGH54 Ratio of ground weight to fording weight WRFi5 Tread width (per assan.bly) WT

56 Min. width betw. traction elements (par assembly) WTE57 Combination vehicle braking coefficient XBR

Chaptor 3 Extensioo of Procedures 17

Table 2

Terrain Parameters Capable of Being Treated as Random Variables

Terrain

58 Surface roughness ACT59 Sol depth to bedrock DBR60 Superelevation angle EAN61 Terrain elevation ELEV62 Terrain dope GRA63 Obetade approach angle OBA64 Obtac height OBH65 Obstacle width OBW

6 Obstacle length OBL67 Obstacle spacing 0BS68 Road radius of curvature RAD69 Soil strength RCI70 Recognition distance RDA71 Stem spacing (per class) S72 Depth of standing water WD

Table 3

Scenario Factors Capable of Being Treated as Random Variables

Scenario

73 Driver safety factor, sliding & tipping COE74 Cohesion of snow COH75 Max. braking deceleration driver will accept DCL76 Specific gravity of snow GAM77 internal friction angle of snow PHI78 Dnver braking reaction time REA79 Max. recognition distance RDF80 Amount of available braking driver will use SFT81 On-road visibility-limited speed VBR82 Off-road visibility-limited speed VIS83 On-rod speed limit VLI84 Mn. vegetation override speed VWA85 Snow depth ZSN

Table 4Regressions Capable of Being Treated as Random Variables

Regre lone

86 Drawbar pull coefficient versus excess rating cone index FDO87 Powered/Braked motion resistance coefficient versus excess rating cone index FRP88 Towed motion resistance coefficient versus excess rating cone index FRT89 Slip verus pull coefficient FSL90 Mobility index vs several vehicle parameters FXM

Chapter 3 Extension of Procedures

NRMM Parameter Sensitivity NRMM Parameter Sensitivitye~e eelS 27j6

oXI - 4,2

aft,

jo Af

05 M 63 63 5 40 07 S4 6 s s3 0A

*7 00o. * - a', e

a. Lauterbach (WGe) b. Winsen (WGe)

NRMM Parameter Sensitivity NRMM Parameter Sensitivity

.. 3 - 614

62-

s~ &2

*2.S. .

c.- Mafra (Jra)d chte W

Figur 8. Gobalview f parmete senstivit forM99 nfurtran

Chate gurxeno of Globaedvrew 19prmtrsniiiyfo 98i ortran

NRMM Parometer Sensitivity NRMM Porameter Sensitivity.0 6=2 i f2nd

0.3' 032.03 S.i•

4264 114.

02O- *k Ore

62 amml -w ymn Nam

at#- 9.14- 41

014,

a46

itt IL e..

5.1 SOi 93 04 0N 07 O5a 4 00 0 0 7

7 own

a. aftrah (Jorda) . Schoten MWe)

FigureNM 9.Goa Piwof paraetr S esitivity forM M97i oraerinsitvt

201 C t -i o

a. ::.

SAM

as-

oas-

i 0C3 -i 1 2

42-

6 01 5M A Q4 00 04 07 09 OI W 36 0 flU ~.6flUWL L 4 7 f

c. Mafraq (Jordan) d. Schotten MWe)

Figure 9. Global view of parameter sensitivity for M977 in four terrains


NRMM Poromexer Sensitivity NRMM Porometer Sensitivity::3:ItO ,

*411 so

4114 140*

@24

at MM 4WL. L It .. . L

a t ; A A A ia 0 55 so - a a U isA a 46 6 19611 M NJ 2-042 U 70 A A

V 4111

a. Lauterbach (WGe) b. Winsen (WGe)

NRMM Pirometer Sensitivity NRMM Parometer Sensitivity

&2 - o

3it-I l -i I I

...+ ,H l+?

4 04 -

L I ... ........... .... ..6 1C 16 21D •6 30 AJ 4b W 06 40 46 W IS O A 10 It 30 Ai 3D, 3i6 40i 416 W~ llI k 0 ? O

c. Mafraq (Jordan) d. Schotten (WGe)

Figure 10. Global view of parameter sensitivity for M 113 in four terrains

21Chapter 3 Extension of Procedures

NRMM Parameter Sensitivity NRMM Pororneter Sensitivity

so* dab.??

W.41 - f A

iSi. -am-

a-0".

am.N

&" S

-m tMILel

so, 0 a 1 0f 001'S4M

A-

at

c.- Mafra (Jrn d. -cotn W

Figur Parameteie o pratr S ensitivity forM Pai ouerameeSnntvt

22 Ch pe 3l Exe5.0f rce ue

4 Application to a HistoricStudy: HMMWV CandidateVehicles

Introductory Remarks

In this part of the report and in Chapter 5 to follow are discussed twoapplications of the stochastic mobility forecasting procedures to importantstudies accomplished several years ago. Both studies influenced decisionsabout vehicle procurement. Both were accomplished using deterministicmobility forecasts (called "assessments"). It is of considerable interest to seeif conclusions reached at that time still hold up when inevitable uncertaintiesin the data, ignored then, are now accounted for. It is also a helpful way topresent the subject of stochastic mobility forecasting to the user communitybecause the subjects of the historic studies are now in the current inventory ofvehicles and their replacements are on the drawing boards and will requiresimilar studies in the future.

The Historic Study

This study was part of a report entitled "Ride and Shock Test Results andMobility Assessment of HMMWV Candidate Vehicles: (Green, Randolph,and Grimes 1983). In 1982, specifications for a "High Mobility MultipurposeWheeled Vehicle (HMMWV)" called for a common chassis capable of accept-ing several body configurations to accommodate weapons systems, utility, andambulance roles. The HMMWV was to be 4x4 wheeled vehicle capable ofoperating off-road, on trails, and on primary and secondary roads with a1-1/4 ton payload at speeds comparable to the then-current high-mobilityM561 vehicle.

Three military vehicle manufacturers were awarded contracts to buildHMMWV prototypes for test and evaluation. Each manufacturer provided2 prototypes; one was configured as a utility vehicle and the other, as a weap-ons carrier. For purposes of analytic mobility forecasting using NRMM,vehicle data files were prepared for 11 vehicle configurations: 3 for each of

Chapter 4 Application: HMMWV Candidate Vehicles 23

the 3 prototypes and one each for the M151 jeep and M561 (GAMMAGOAT) utility vehicles then in service. The study considered the utility vehi-cles in both loaded and unloaded conditions and the weapons carriers inloaded conditions.

Study terrains were chosen in the Fulda, Lauterbach, and Schotten areas ofWest Germany and in the Mafraq and Az Zarqa areas of Jordan. Seasonalconditions included dry, wet, Rnd slippery surfaces, and gave special con-sideration to snow conditions in Germany and sand conditions in Jordan.Performance assessments were based on mission rating speeds for severalstandardized and specialized HMMWV mission definitions. Details arespelled out in the original report.

That report did not state a "winner." It did not attempt to formulate aranking. It merely stated comparative outcomes on a variety of terrainsaccording to a variety of missions. Army procurement specialists acceptedsuch comparisons as advisements during their deliberations. Accordingly, itwas unnecessary to completely rerun the study with the stochastic procedures.It sufficed for the purpose of demonstration to deal with just a portion of theoriginal study.

Scope and Methods of the Application

Three vehicle data files were prepared to represent the fully loaded utilityconfigurations of the candidate HMMWV vehicles. The vehicles were desig-Dated the AU7 (AM General Corporation), GUF (General Dynamics, Inc.),and TUQ (Teledyne Continental Co.). Photos of the vehicles are presented inFigure 12. Photos of the M151 and M561 comparison vehicles are shown inFigure 13, A listing of geometric characteristics is given in Table 5.

At the outset, each vehicle was studied individually on each of 4 terrains.The terrains were 5322, the Lauterbach off-road quad, and 5520, the Schottenon-road quad in WGe, and 3254A, the Mafraq off-road quad, and 3254R, theAz Zarqa on-road quad in Jordan. In all cases only a "dry, normal" seasonwas used. The sensitivities of each vehicle/terrain combination were deter-mined for the 90 parameters of NRMM whose variability could be controlledat that time. As few as 2 parameters and hs many as 12 parameters out of the90 studied in each case were found to be sensitive. Considered together, only14 parameters out of 90 were identified as sensitive in any vehicle/terraincontext studied. Figures 14 to 18 identify the sensitive parameters. In allcases, the 14 parameters are listed along the horizontal axes. Note that verti-cal scales differ among the plots. See Tables I to 4 to identify the acronyms.

Following the screening of sensitive parameters for all vehicle/terraincombinations, Monte Carlo speed prediction simulations were performed inwhich the sensitive parameters were treated as random variables and the otherparameters were held constant. ErTor-magnitude scenarios were composedbased on judgments of the statistical attributes of the random variables. As in

24 Chapter 4 Application: HMMWV Candidate Vehicles

a. AM General 1-1/4-ton utility vehicle, AU7

b. General Dynamics 1-1/4-ton utility vehicle, GUF

c. Teledyne Continental 1-1/4-ton utility vehicle, TUQ

Figure 12. The candidate HMMWV vehicles

25Chaepter 4 Application: HMMWV Candidate Vehicles

a. M 151 A2 Truck, utility, 1 -1/1 4-ton

b. M561 Truck, cargo, 1-1/4-ton

Figure 13. The comparison vehicles

26 Chapter 4 Application: H-MMWV Candidate Vehiclas

Table 5Vehicle Dimensions

VIDTH

CL

Fiaure VehicleGetomet ry 8eferencc AU7 ClIF "1130 MlSlA2 MS(,!

Vehicle/unit longth, in. VL 185.0 139.0 192.0 132.0 228.0

Front axle to end of vehicle. in. AIELl 163.0 163.0 171.0 113.0 203.0

Front hitch to front axle, in. FLL 22.0 25.0 24.0 19.0 20.0

Front axle to rear hitch, in. ELL 163.0 163.0 171.0 113.0 203.0

Vheelbese, in, VI 130.0 134.0 130.0 85.0 165.5*

Front axle to CC, in. iCC 77.1 74.6 74.7 44.5 90.1

Front axle to driver's station, in. - 61.0 64.0 63.0 .. ..-

Vehicle/unit height, in. HO 71.0 73.0 72.4 71.0 92.0Driver's eye height, in. gyEiiT 62.0 68.0 59.0 62.0 64.0

Driver's station height, in. - 34.5 40.5 30.0 .. ..-

Front pushber height, in. P881 24.8 25.5 17.5 18.G 27.0

Front hitch height, in. FI4 24.8 25.5 17.5 18.0 27.0

Approach angle, deg VAA 90.0 63.0 61.0 66.0 62.0

CC height, in. CCI 33.5 33.5 32.25 24.2 34.7

Minimum chassis clearance, in. CL 11.3 10.0 9.25 8.8 13.5

tear hitch height, in. 1t8T 24.9 25.0 26.25 20.0 15.8

Departure sngle, deg VDA 58.0 51.0 40.0 37.0 52.0Vehicle/unit width, in. h 84.0 84.0 82.3 64.0 85.0

Lateral distance: center plane to C1, in. C.LAT 0 0.3 0 0 0

If COLAT 5f 0 lateral CC location fro center planelooking forward: right (i) or left (L) -- I . 1. . .

Swim speed, mph (0 for non slme er) -- 0 0 0 0 2

Fording depth, in. (measured fro1 3round) orWimln depth, in. .-- 30.0 30.0 30.0 21.0 49.0

F teel track, in. 1r(l) 71.6 72.3 69.8 53.0 72.0

Fround clearanee under axle, In. C IN(1) 11.3 10.0 9.25 8.8 13.5

Lateral clearance between inner tires, in. 111E(1) 59.1 59.8 57.3 46.0 61.0

* First to last axle.

Chapter 4 Application: HMMWV Candidate V hicle 23

AU7 on 5322 AU7 on 5520

Oil

004 t S

oats -

a. Lauterbach (WGo) b. Schotten (WGe)

AU7 on 3254o AU7 on 3254r

A 0" S84m

c. Mafraq (Jordan) d. A Azarq (Jordan)

Figure 14. Parameter sensitivi of AU7 in four terrains

28 Chapter 4 Applietion: HMM~WV Candidate Vehilee

GUF on 5322 GUF on 5520

i 0*s0 -

022 02

OW w*

Ol, ail~w m I. PO w lol , w .M .M . w. S MU

a. Lauterbach (WGe) b. Schotten (WGO)

rGUF on 3254a GUF o~n .3254r

0.201020-02-

llas ]

M -l'WI . . Ow a- on..M. X. 0.al-0

c. Mafraq (Jordan) d. Al Azarq (Jordan)

Figure 15. Parameter sensitivity of GUF in four terrains


TUO on 5322 TUO on 5520

SAO

VVA M W W M MM2 S M

S'Sep

a. Latrbc (We .a.ote W

TU on. 324 o 24

114-

0*. Z,

a. atrac (Wrd) d. Slhotten (orda)

Figur 16. 3254ete sesiivt on 3254rfurteran

30 C apte 4 pp!*cston: MMW Canidae Whc5a

M151 on 5322 MI51 on 55'.C

j I of

0 *

-. . --- go a90 .2, go S

a. Lauterbach (WGe) b. Schotten (WGe)

M151 o17 3254o M5 on 3254r

C*

, /

* 0 oS,/

oS A L,

-o-..o - -./-


Figure 17. Parameter sensitivity of Ml151 in four terrains

31Chapter 4 Application: HMMWV Candidate Vehicles

1561 on 5322 M561 ort 5520

IF

m~~~o -- if oI

a. Lautwbach (WGe) b. Schotten MeG)

U501 on 3254o M561 on 3254r

63i

am w w9I ii

-, , o a . s '0n . . . . .0 0* C . . . . . .to


Figure 1G. Parameter sensitivity of M561 in four terrains

32 Chapter 4 Application: HMMWV Candidate Vehile

Report 1, these error-magnitude scenarios were conjectural and must ulti-mately be strengthened by the involvement of mobility-data-collection profes-sionals. Table 6 summarizes the characteristics speculatively ascribed to the14 parameters that were identified as sensitive in this study.

Table 6

Conjectured Statistical Attributes of Parameters Selected for theError-Magnitude Scenario

Parameter DietbutionAcronym Type Range

ACT Uniform Plus and minus 10 percent of nomind

FDO Gaussian Standard deviation 5 percent of regression value

FRM Uniform Plus and minus 5 percent of table midpoint

FVA Uniform Plus and minus 5 percent of table midpoint

FVO Uniform Plus and minus 5 percent of table midpoint

GRA Uniform Plus and minus 5 percent of nominal

OBA Uniform Plus and minus 5 percent of nominal

OBH Uniform Plus and minus 5 percent of nominal

OBS Uniform Plus and minus 5 percent of nominal

RAD Uniform Plus and neus 5 percent of nominal

S Uniform Plus and minus 3 percent of nominal

WDT Uniform Plus and minus 2 percent of nominal

WGH Uniform Plus and minus 10 percent of nominal

VRI Uniform Plus and minus 5 percent of table midpoint

Table 7 repeats the identification of sensitive parameters with side-by-sidecomparisons among vehicles and terrains. Note that where few parametersare sensitive, those parameters are very sensitive and others have beenscreened out; conversely, where many parameters are sensitive, no oneparameter is dominant.

Fingerprints and Speed Profiles

The basic information generated by stochastic mobility forecasts is con-tained in fingerprints and speed profiles. Postprocessing this informationyields mobility rating speeds and speed ranges which are the factors of great.-est interest to the user community. In this section are presented numerousfigures which show the stochastic NRMM speed predictions for the HMMWVcandidates. They are presented as fingerprints, in-unit speed profiles andaverage speed profiles for each vehicle on the off-road terrains (5322 and3254a) and as fingerprints, and average speed profiles for primary roads,

33Chapter 4 Application: HM, MWV Candidate Vehicles

Table 7Identification of Sensitive Parameters

Terrains:3254r - Az Zarqa Scenado: Dry. nomd5520 - Sohotten3254.- Mafraq Vehicles: AU7, GUF, TUQ. M151, M5615322 - Lauterbach

ParameterAcronyms:

F F F V W W A G 0 0 0 R S FV V R R D G C R S B B A DA O M I T H T A A H S D 0

Soreeninge:5322 AU7 0 * 0 a a a *

GUF * * aTUG Q * *M151 * * * *

M561 0 # 0 *

3254. AU7 * * * 0

GUF * * aTUG * 0 aM151 * *

M561 a *

3254r AU7 * a

GUF " aTUc * aM151 * a

M561 a *5520 AU7 a a

GUF # aTUG a aM151 * *

M561 a a

secondary roads and trails for the on-road terrains (3254r and 5520). Thisinformation is presented in Figures 19 through 38. It will be helpful to dis-cuss two of these figures in detail as representatives of all the others.

Figure 19 shows fingerprint and profiles for the AU7 on 5322. The fin-gerprint, remember, conveys a global impression of the impact of data uner-tainties in the speed prediction of NRMM for this vehicle, this terrain, and theassociated error-magnitude scenario. The greater the departure of the finger-print from a straight line at unit slope, the greater the effect of the uncertain-ties. Clusters are characteristics of the vehicle. Note that the density ofpoints gives clues about the importance of the clusters. There are 2707 pointsin this figure. Above about 20 mph, most of them actually fall close to theunit slope line. Below 20 mph, most points occupy a dense cluster whosewidth is fairly well defined. Points that can be distinguished individually rep-resent events of relatively infrequent occurrence. In other words, relativelyfew of the 2707 terrain units are repiesented by individually distinguishablepoints.

The in-unit speed profile contains the same speed data as the fingerprintbut plots speeds versus an area fraction obtained by sorting a list of nominal

34 Chapter 4 Application: HMMWV Candidate VaNolec

AU7 on 5322 (off-road)

a. Figeprn

W4 "o rs

f3 3

I 4* 4

b. ~ 4 Inui sedprfl

A7 on 532(ofrod

o 02 a4 D

0 -- 00 40-m -

Figure 19~~ ~ . Fingerprintsadsevpoiefr U onLurbc

Chapter ~ ~ ~ ~ AU on 5322aton (off-roadte)a~l 3

AU7 on 5520 (on-rood) AU7 on 5520 (on-rood)a F b Prry r4o ae s

U a,

t* .-p *... °

" "

AU7 on 5520 (on-road) AU7 on 5520 (on-road)

4 404

64-4

I. .I b .. ...

a 62 -"4~I

c. Secondary road average speed profile d. Trails average speed profile

Figure 20. Fingerprint and speed profiles for AU7 on Schotten

36 Chot2r 4 Appli(rtion: HMMWV Cndidate Veh(ole

AU7 on 3254 (off-road)-

'01

0 '4

40 4

4,

a . Figrrn

~4-

S1.

0 40 Oka

. Fuingserpri

AU7 on 3254 (off-road)7 0 -r - - - - - --- -

40.

SO %

c. Avrgsedprfl

Figure~~~~b 21nFigepnitsan speed profile U nMf

Chapter ~ ~ ~ ~ AU 4n 3254aton (off-road)e ehile3

AU7 on 3254 (on-road) MJ7 on 3254 (on-rood)

s

a. Figrrn b. Prmr odaeag pe rfl

AU o 324 (nro) n35 o-odW MM W - "w01

c.Scna. Foangaerprin bpe rfl .Prayloa average speed profile

Figre22 oigrnt 254 (op-eod oflso AU7 on 325 (on-ood

U.8 Chatar Appliation HM W Canddat!____

GUF on 5322 (off-road)

90-

f 4

0~4 4444,

NO' PM * * 00

40.

4* * 4

0 a2 CA ION --1 "MW W

a. Fnuingserprof


40

ao-

0

. Averagt speed profile

Figure ~ ~ ~ GU 23 igrrnsadsedof5322 (ofroad)nLatrbc

- 39Chaper 4Appicaton:WAMWCanddat Vehcl0

GUF n 520 (n-radsGUF on 5520 (on-roodj

U.

a. Figrrn .Piayra vrg pe rfl

GU on520(nrod UFo 52 o-ra)

aiue2 . Fingerprint Piardod vrg speed profiles o U nShte

G40 onate 552 (ppliaooon GUFW onn5520 teon-road


40.

* 10..

v 40.3 *

0 40

10

0

a. Inuingserprtl

GUF on 3254 (off-road)7g.

3 0.

2I0

*0

. Aneragt speed profile

Figure ~ ~ ~ GU 25 igrrnsadsedof3254 (ofroGo afa

Chapter~7 4 plcto:-VCaddt eils4

GUV on 3254 (on-rood)' GUV on 3254 (on-rood)

S .0..Q

97 *A"c.Scnayra vrg pe prfl . Tal vrg pe rfl

iue2.Fnepitadsedoie for GU nA a1

42 hate 4 Apliaton H MW Cnddae e1le

TUG on 5322 (off-rood)

40.

61

is i

a ~ 4 - 0

i . Fingerprint

TUG on 5322 (off-rood)

to-

3.

b. In-unit speed profile

TUO on 5322 (off-rood)

10

c. Average speed profile

Figuiu 27. Fingerprints and speed profiles for TUQ on Lauterbach


TUO on 5520 (on-road) TUO on 5520 (on~-rood)

U -~ --

*A Go 9

-- OOPA. ~,_O~-*

a .A. Figrpit .imayra vrg pe rfl

411-

S 2 64 U 2 8SO

c. Secondary road average speed profile d. Trails averace speed profile

Figure 28. Fingerprint and speed profiles for TUQ on Schotten


TUQ on 3254 (off-rood)3 awm ll-wd/ k-

i I - 'm u . f.I0

0 0D4

S 0

s

I a] J .-

20.

o a a

a. RIngerprint

TUO on 3254 (off-road)

2-4

10.D

IC-

o 0,4 CA


TUO on 3254 (off-road)

' I'a


Figure 29. Fingerprints and speed profiles for TUQ on Mafraq


TUO on 3254 (on-rood) TUQ on 3254 (on-rood)

f

Wa

a.Fnepitb mr oaTvrg pe rfl

TI n35 o-od U n35 o-od

Fiue3 . Fingerprint b.Paarnoddvrg speed profiles o U nA aq

46 n35 onro)TOo 35 o-od

46 Chapter 4 Application: HMMWV Candidate Vah~cles

M151A2 on 5322 (off-rood)

~~4 .0 -

70 ;

"..w-. -

a. Fingerprint

M151A2 on 5322 (off-road)

do

40.

1 10-

0



- 40-

j 20-

0 0

c, Average speed profile

Figure 31. Fingerprints and speed profiles for M1 51 on Lauterbach


M151A2 on 5520 (on-rood) M151A2 on 5520 (on-rood)

I . .:

a . Figrrn b. Prmr.odaeag pe rfl

aiue3 . Fingerprint b.Piarnoddvrg speed profiles o 5 nShte

48A onate 552 (on-rood)on MMA ondia20 (onod


ID;

2-* 4.0

1 00

0 0 L

b. Inui spedpofl

M151A2 on 3254 (off-road)

0-

* 20-

0 0. 04 04 06

. Anerait speed profile

Figure ~ ~ ~ M1A 33 igrrnsadsedof3254 (ofroad) n af

Chate 4. Aplcto:H M VCaddt eils4

M151A on 3254 (on-rood) M151A2 on 3254 (on-roo)

31 ;0 0. 1~~U- -

-4-4

c.Scnayra vrg pe rfl . Tal vrg pe rfl

Fiur 3. inerrit ndspedprfiesfo M 5 o A Zr.

50Ch ptr A plcaio : MM V anid teVeiIo

M561 on 5322 (off-road)

a. Fingerprint


I -

. ..

S to-

40I



i4-

4 ,°


Figure 35. Fingerprints and speed profiles for M561 on Lauterbach


M561 on 5520 (on-road) M561 on 5520 (on-road)

I -- U- - ..f________omp __________.

~ . ~ ..6000 ,

a. Fingerprint b. Primary road average speed profit.

M561 on 5520 (on-road) M561 on 5520 (on-road)

-a.. -4 - 62 4

c. Secondary road average speed profile d. Trails average speed profile

Figure 36. Fingerprint and speed profiles for M561 on Schotten


M561 on 3254 (off-rood)

• m 4. 410

a. ingerpint

M561 on 3254 (off -rood)3-

+4

f .| 4.

I -.

i



I d V lI to,

0 0.2 0.4 04 O.

. Averaie speed profile

Fiur 3. inerritsan see pofe frM561 on 325a(of-rad

Cheter4 Apliatin: MMW Cadidte ehile -534

M561 on 3254 (on-rood) M561 on 3254 (on-rood)

*SN

0- U-

aiue3 . Fingerprint b.Piarnoddvrg speed profilesfrM6onA aa

546 onpts 325 Appn-catio) HM O ndi e (on-oo

speeds and accumulating, from highest to lowest speeds, the areas associatedwith the terrain units having those speeds. See Report 1 for details. As oneviews this plot from left to right, more and more terrain units are being accu-mulated whose nominal speeds are lower. Thus, the left-most terrain units are"better" for vehicle mobility than the right-most ones, an identification forwhich the speed profile was devised. In-unit speed profiles usually have aclearly defined backbone. The abrupt step-like dropoff on the right-side of theprofile Ocrives from the "worst' terrains in which zero speeds (NO-GO) arepredicted.

The average speed profile, once again, uses the same speed data as thefingerprints and the same accumulated areas as the in-unit profiles, but usesthe areas, in addition, to weight the speeds. Again, Report I has the details.The average speed profile contains three traces, one each for nominal, mini-mum, and maximum predicted speeds, respectively. The bottom-most tracearises from the minimum predicted speeds. The irregular character of thetrace derives from the occurrence of low or NO-GO minimum speeds in areaswhose nominal speeds are much higher. These speeds can be seen individu-ally in the in-unit profiles.

In Figure 20 are shown the fingerprint and average speed profiles referredto road type, There is no difference in concept between on-road and off-roadfingerprints and speed profiles. But NRMM distinguishes several road types(super-highway, primary, secondary, and trails) and allows speed profiles tobe similarly identified by simply separating the speed predictions by roadtype. Road-type-specific speed profiles are of central importance in the defi-nitions of mission rating speeds.

Missions and Mission Rating Speeds

The average speed profiles are the data resource from which mission ratingspeeds are calculated. The historic study defined several missions to evaluatethe candidate vehicles. These are shown in Table 8.

Data from Table 8 are worked into mission rating speeds as follows.

Let MRS = Mission rating speed.P = Off-road operations percentage.P, = Primary road operating percentage.P, = Secondary road operating percentage.P, = Trail operating percentage.C = Percentage best off-road terrain.

PP = Percentage best primary roads.PS = Percentage best secondary roads.PT = Percentage best trails.V, = Speed corresponding to C on the off-road average speed profile.

VP, = Speed corresponding to PP on the primary road average speedprofile.


VP, = Speed corresponding to PS on the secondary road average speedprofile.

V, = Speed corresponding to Pr on the trail average speed profile.T = Average time in hours spent crossing gaps and steams per mile

of off-road terrain traversed (0.101 in Germany and 0.025 inJordan under dry, normal conditions).

then

RS - 100

+ + eT + + + +rc rp- V- V-

able 8HMMWV Evaluation Missions

Perint Total Peroent of "as"Operating Distance n Terraln/Roid Units

Moblty rn1;V -NI f . Idnay8ondrFYLevel Foas oads 1Tras inod jRoads i eads Trails Road

Federal Republic of Germany

Tactical High 10 30 10 50 100 100 100 90

Tactical Standard 20 50 15 15 100 100 100 80

Tactical Support 30 55 10 5 100 100 50 50

On-Road 35 0 5 0 100 100 10 -

HMMWV 30 30 30 10 100 100 80 80

- -Jordan

Tactical High 5 20 25 50 100 100 100 90

Tactical Standard 1 35 35 15 100 100 100 80

Tactical Support 20 40 35 5 100 100 80 50

On-Road 30 40 30 0 100 100 50 -

HMMWV 30 30 10 30 100 100 80 80

The V's are gotten by entering the appropriate speed profiles at the valuesof the "percentage best" terrains. For example, entering the average speedprofile for the AU7 on 5322 (Figure 19) at C = 80 percent (or as the fraction0.8) gives V, = 8.0, 18.5, and 21.0 mph as minimum, nominal, and mlxi-mum values. Entering at C = 50 percent gives 15.0, 25.0, and 28.0 mph.Finally, entering at 90 percent gives 2.0, 6.0, and 6.0 mph. When values forthe V's are taken from the nominal speed profiles, a nominal mission rating

56 Chapter 4 Applicaton: HMMWV Candidate Vehicles

speed results. When values are taken from the minimum and incximum pro-files, minimum and maximum mission rating speeds are obtained.

Note that the difference, or *range,* between minimum and maximummission rating speeds can be taken as a figure of merit for the performance ofthe given vehicle on the given terrain against the given mission. Two vehicleshaving the same computed nominal mission rating speed can be rankedaccording to range. A smaller range indicates less susceptibility to the influ-ence of data uncertainties.

Mission rating speeds for the 3 candidate vehicles and 2 comparison vehi-cles are shown in Figures 39 to 43. Nominal, minimum, and maximum val-ues or the mission rating speeds are clustered by vehicle. One can think ofthe maximum and minimum speeds as defining an error band about thenominal.

Comparison of Historic and Stochastic MobilityForecasts

The first point of comparison is between the nominal mission rating speedsof the stochastic procedure and the mission rating speeds of the historic study:they should be the same. Figure 44 shows representative comparisons madefor the "HMMWV" mission. The comparison is, in fact, favorable. That thehistoric and nominal values are not identical is reasonable because NRMM hasevolved since 1983 (and continues to evolve) as computational refinements areimplemented on a continuing basis.

It should be remarked that the three candidate vehicles are actually verymuch alike. It comes as no surprise that their nominal mission rating speedsare largely the same. This provides an opportunity to see if the stochasticprocedures can discriminate among these vehicles.

Consider Figure 43. Of all the missions considered, the HMMWV missionis probably most representative of the demands made on the vehicles. Look-ing at the nominal speeds, one sees that all candidates outperform the M151and M561, the high-mobility vehicles of the time. Thus, if scores are beingkept, it is clear that the M151 and M561 are out of the picture at the outset.Based on nominal values of the mission rating speed, little would separate theHMMWV candidates. However, when ranges are considered, the TUQcomes up short as the stochastic analysis procedures have captured consider-

pradictions for the TUQ entails more risk than for the AU7 or the GUFbecause the predictions cover a wider range of values. Finally, between theAU7 and GUF is seen a classic tradeoff situation. In Figure 43a the vehiclewith the higher mission rating speed (good) has the higher range of speeds(bad). Thus, a procurement committee, looking at this scenario only, thatmight have been tempted to rank the GUF above the AU7 based on missionrating speeds would realize that another factor is at work. The uncertainty in

57Chapter 4 Applhcaton: HMM4WV Candidate Vehicles

RATING SPEEDS: Tactical High missionLisAudwVSchdM Oy flsm*

a-

m ~ ~ ~ ~ rw Ewffm umf

! 3-7min

2 -minfm

1 min

AU7 GIF 7wO 161 541

HUMMWV Ca~dd". V~hc1

RATING SPEEDS: Tactical High missionMWfvuaJAa-Zatq Dry .Wm

17-14-

Is- ffm

14-MA

a 12-

to- min

c 7,

5 4

4 Z32-

AU? OW 1.3 711 641

Figure 39. Mission rating speeds for the Tactical High mobility mission

58 Chapter 4 Application: HMMWVV Candidate Vehicles

RATING SPEEDS: Tactical Stondrd mission21 m20-fl

Is-17- W

E 147i 13-

12-

C0

6-

2.

AU7 our 7iO 161 6A

#4JMMWV Con&dot VehoIe

RATING SPEEDS: Tactical Standrd mission26

MM24-

22-

20-m

16-

* 14

S 12- i

Z100

0

AU U w 5 6IUMVCn& *Vhl

Chapter ~ ~ ~ ~ 4Jhi 4~i~ Aplcain-HM aniat ehces5

RATING SPEEDS: Tactico' Support mission

26

26

124

2

SM 16

26

24

277oh

AO1 GLF ruo 151 541

.www Cw aev.hK f

FioureRTIN SPEDS Mno ekspfTacrOw tical Support miion

60~ ~ ~ ~ ~ ~~~~Je2q Ctpr 4Ap~dn LANCFW" ei

RATING SPEEDS: On-Road mission

40-

* 22

207C

36

10-

10 - 0 7

U oW iJ 11 61

IJU Ceii V" i

FigureRTIN 42 iso aigSPEDS:rh On-Rod mbt mission

Chapter ~~~U m/D-Zq 4v AplctinuvmdCnodt hils6

RATING SPEEDS: HUMMWV mission~LaftSioxse" Dry l-

24m

2

10

AU? OLF lUG6156

HUWVWV Con~dieVtc

RATING SPEEDS: HUMMWV mission26' :m f-wo.Dyrf~

24-

22

20-_m

j 16 M

* 14

12 -min

C 10

4

2-min

AU7 GUF 5UGIl M

leUMMWV C.'6j. V.1*1 2

Figure 43. Mission rating speeds for the HMMWV mission

62 Chapter 4 Appiiaation: iIMMWV Candidete* V011cle6

RATING SPEEDS: HUMMVW missionLfbO h/S ft' Cwy ~W, -6

1,1X

a 2

C 10

C

6" /4

3

AU7 Ou TUO 151 1

CIIUMW caftake Veh.:e

RATING SPEEDS: HUJMMWV mission24UMfM./ft-Zarq,. Oy fmcvmd

22 /

2

/ \\\AU 184-K TnO 11

WWWW /&+7e t-

Figure~~~~/ 44/oprsno itrc(hs)adsohsi oia "o"

miso raigsedsfrteHMW iso

12 / Nsn63

Chapter ~ ~ ~ ~ ~ ~ */ 4Apiain M W CaddtVecs

the speed of the GUF exceeds that of the AU7. Whether this factor is signifi-cant, or whether the vehicles are actually too close to call becomes a judgmentcall in its own right. In Figure 43b, this tradeoff situation does not arise.

In a similar manner, one can step through all of Figures 39 to 43 and notethat the AU7 and the GUF are the only real contenders and that in someinstances, the AU7 outperforms the GUF and in other instances the GUFoutperforms the AU7. Clearly, other scenarios would need to be studiedbefore a ranking of the vehicles on the basis of predicted speeds could beaccomplished.

Will deterministic mobility assessments undergo substantive changes whendata and model uncertainties are considered explicitly? The answer to thisquestion is a definite "maybe." This is actually not meant to be amusing: theanswer depends, like the stochastic mobility analysis itself, on the particularcombination of vehicle and terrain. In the present instance, the answer isprobably no: there will probably be no substantive change in the outcome ofthe HMMWV candidate rankings with stochastic forecasting compared withthe historic deterministic methods. In this case, the historic methods werequite satisfactory. But let the vehicles be other than those considered, and letthe terrains be other than those considered and one must ask the questionagain.

The real benefit of the stochastic approach is that, in project-specific con-texts, sensitivity of NRMM to its parameters and uncertainties in data andalgorithms are quantified and illustrated along side of traditional deterministicoutcomes. The additional information that flows from this procedure allowsspeed prediction risks to be clarified and faced by decision makers.

64 Chapter 4 Application: HMMWV Candidate Vehiole

5 Application to a HistoricStudy: FOG-M CandidateVehicles

The Historic Study

In 1987 mobility studies were performed for the Forward Air DefenseSystem (FAADS) using the Army Mobility Model (AMM). The purpose ofthese studies was to determine the best candidate vehicle, based on mobilityassessments, for the Fiber Optic Guided Missile (FOG-M) component of theFAADS.

Candidate vehicles were required to perform at the mobility levels outlinedin the Non-Line-of-Sight (NLOS) Required Operational Capabilities (ROC)document of July 1987. This required the FOG-M vehicle to be equallymobile as the other vehicles within the supported forces. An additionalrequirement that the vehicle operate within 2 to 7 km of the Forward Line ofOwn Troops (FLOT) put additional limits on the chosen vehicle.

Using these criteria, the U.S. Army Missile Command (MICOM) selectedtwelve candidates for the FOG-M vehicle. Of the twelve, four were trackedvehicles and eight were wheeled. Three terrains were selected for study,those being the Federal Republic of Germany, Southwest Asia, and theRepublic of Korea. The vehicles were evaluated on dry, wet-slippery, andsnow surface conditions. These scenarios are analogous to studies done byFAADS for its Ground Based Sensor Component.

Each vehicle in the FOG-M study was evaluated under 'onditions toapproximate the tactical operation requirements of the FOG-MA mission. Inorder to quantify mobility performance, WES, in conjunction with theU.S. Army Training and Doctrine Command (TRADOC) developed fivelevels of tactical performance. These levels include high-high, tactical high,tactical standard, tactical support, and on-road. Each performance level isbased on percentage of travel performed off and on-road, as well as the sever-ity of traveled terrain.

In the FOG-M analysis, the candidate vehicles were required to operate atno less than the tactical high mobility level and many times at the high-high

Chapter 5 FOG-M Candidate Ve~cles 65

level. In order to select the best performer of the candidate vehicles, thetactical high performance level was averaged over the three scenario areas.Based on this average, the M993 Fighting Vehicle System was selected as thebest candidate vehicle for FOG-M applications.

Scope and Methods of Application

The vehicles used in the stochastic analysis will be limited to the four topperformers as determined in the original non-stochastic FOG-M study. Thesevehicles, in order of their selection, are the M993 (FVS), the M977(HEMTT), the M113A3, and the M1037. This group consists of two trackedvehicles as well as two wheeled vehicles. As in the original analysis, thesevehicles will be evaluated at fully rated combat payload and a 12 watt limitingride speed.

The M993 is shown in Figure 45 and is a fully-tracked armored carrier,fighting vehicle system (FVS). It is one of the heavier vehicles used in themobility study with a weight of 56,200 lb. The M993 has a length of 275 in.and a width of 117 in. The tracks themselves are 174 in. long and 21 in.wide. Minimum ground clearance is given as 17 in. This vehicle has apower rating of 17.8 hp/ton which provides it with a maximum velocity of35.7 mph.

The vehicle rated as second in the original analysis is the M977 HeavyExpanded Mobility Tactical Truck (HEMrI'), Figure 46. This, the heaviestvehicle considered, has a gross weight of 60,375 lb. It is one of the twowheeled vehicles considered. The length of the M977 is 361 in., and itswidth is 96 in. Its wheel base is 210 in., and the minimum ground clearanceis 13 in. This vehicle can deliver 14.3 hp/ton and obtain a maximum velocityof 63 miles per hour.

The MI13A3 Figure 46, is a fully-tracked, armored, personnel carrier andis one of the lighter vehicles considered. This vehicle weighed 27,200 lb andhas a length and width of 205 in. and 106 in., respectively. The track lengthis 108 in.; and the width, 15 in. The minimum ground clearance required is16 in. This vehicle is rated for 20.2 hp/ton and obtains a maximum speed of41 mph.

The fourth vehicle considered in this analysis is the M1037 High-MobilityMultipurpose Wheeled Vehicle (HMMWV), which is pictured in Figure 45.This is one of the wheeled vehicles considered, and it is the lightest with agross weight of 8,660 lb. The M1037 is 188 in. long and 85 in. wide, andthe wheel base is 130 in. The minimum ground clearance is specified as11.3 in. The M1037 can deliver the most power and highest velocity of thecandidate vehicles with a rating ot 34.6 hp/ton and a maximum speed of66.8 mph.

66 Chapter 5 FOG-M Candidate Vehicles

a. M I13A3 Armored Personnel Carrier

b. M993 Fighting Vehicle System (FVS)

Figure 45. Tracked candidate vehicles for FOG-M

Chapter 6 FOG-M Candidate VohirIes 6

'.1 ° , : . -.

a. M1037 High Mobility Multipurpose Wheeled Vehicle (HMMWV)

b. M977 Heavy Expanded Mobility Tactical Truck (HEMTT)

Figure 46. Wheeled candidate vehicles for FOG-M

The stochastic study is limited to subsets of the three terrains considered inthe original FOG-M analysis. These selected terrains represent the typicalareal and road conditions that are expected for tactical vehicle deployment.The digital terrain data bases used represent quads in West Germany, South-west Asia, and the Republic of Korea. In Germany, the Lauterbach Quadprovided off-road data; and the Schotten Quad, on-road data. For SouthwestAsia, the Arzhan region provided both on and off-road data; and in Korea,Cheorweon areas were used for terrain data. Refer to the original FOG--Mreport for terrain detail.

In this stochastic study only the dry-normal scenario conditions were con-sidered. The dry-normal condition represents the lowest soil moisture content


of the terrain surfaces - thus, the highest soil strength. This condition isrepresentative of the driest 30 day period of an average rainfall year. Theassumption that no rainfall has occurred within the last six hours is also inher-ent to the dry-normal scenario.

The principal tool used in the stochastic mobility analysis is the NATOreference Mobility Model (NRMM). This computer code is based on years offield and lab work and considers terrain factors, road factors, vehicle geome-try and human factors in speed predictions. The NRMM analysis provides aspeed prediction specific for one vehicle, one terrain unit, and a specific linealfeature of the terrain unit's road network. These speed predictions do not takeinto consideration any neighboring terrains nor past speed predictions. TheNRMM software is an equilibrium code: therefore, there is a unique equilib-rium speed for each specific analysis.

This feature of the mobility model allows for each input variable to bevaried independently or simultaneously in order to show the influence of data.With the input variables modeled as random, the impact of data and algorithmuncertainties are made very apparent. These results help to quantify the qual-ity of speed predictions by providing a range rather than a single number forpredicted speeds. The stochastic approach is clearly advantageous to themaking of tactical decisions.

For the stochastic analysis of the FOG-M candidate vehicles, the version ofNRMM used was release H. This software was executed under a CRAYsystem. The first application of the software was the performance of thesensitivity analysis of the input variables. These inputs consisted of 57 vehiclerelated variables, 15 terrain, 13 scenario, and 5 curve-fit variables, seeTables 1 through 4. Initially these variables were varied independently; and amaximum, minimum, and nominal speed was computed for each terrain unit.

After the computation of these velocity ranges, a second software packagewas employed on the CRAY to generate rank indicators for each variable oneach terrain unit. This program also computed the arithmetic average of theindicators over all of the terrain units. The maximum rank indicator was thendetermined, and a threshold of 20 percent of its value was established. Anyvariable with a rank indicator below this threshold was eliminated from fur-ther analysis.

With the sensitive variables isolated, the full Monte Carlo component ofthe NRMM analysis was then performed. In the previous sensitivity calcula-tions, the variation of each variable was plus and minus 10 percent of itsnominal value while the curve-fit variables had a range of plus and minus14 percent. In the Monte Carlo analysis, however, each variable is given aspecific variation based on past observations and expert opinions. In thisNRMM analysis, each sensitive variable is allowed to vary both independentlyand simultaneously. In the Monte Carlo routine, 200 velocities are calculatedfor each terrain unit and a speed probability density is obtained.

Chapter 5 FOG-M Candidate Vehicles 69

The maximum, minimum, and nominal speeds for each terrain unit gener-ated by NRMM was transported to a PC environment to facilitate the genera-tion of the necessary speed profiles and mission rating speeds. Spreadsheetswere developed to sort the velocities and terrain units and compute the aver-age cumulative velocities. A separate spreadsheet was created for each vehi-cle on each terrain set for each road type; therefore, speed profiles werecomputed for off-road terrain, primary roads, secondary roads, and trails.

Data obtained from the speed profiles was then extracted to compute theapplicable mission rating speeds. For the stochastic analysis, the tactical high,tactical standard, and tactical support mobility levels were computed. Thesemobility levels were developed by representatives of the TRADOC WHEELSstudy group in the mid-1970's. Each terrain has a different definition of themission rating speeds for each tactical level as the requirements for tacticalmobility change with the environment.

Sensitivity Study

Initial calculations in the stochastic analysis required the determination ofvariables for which the vehicle performance was sensitive. Sensitivity andscreening were performed for the four vehicles studied on each terrain set.Within each terrain, the sensitivity was computed independently for theon-road network as well as for the off-road. Of the 90 variables available forthe NRMM forecast, only 21 were determined sensitive in the evaluation ofthe FOG-M candidate vehicles.

Of these 21 variables, 10 were vehicle variz>.Ns, 9 were terrain variables,and 2 were regression curve-fit variables. These relevant variables are listedin Table 9 and their speculative statistical distributions in Table 10. Theseerror-magnitude scenarios again were based on judgments of the statisticalproperties of the random variables, and the conjectural scenarios must bestrengthened by mobility-data-collection professionals. All of the distributionsused in the FOG-M study are identical to those applied in the HMMWVcalculations. Graphs of the sensitivity rank indicators per vehicle per terrainare shown in Figures 47 through 58.

An examination of the sensitive variables for each vehicle on each terrainprovides insight into the factors that affect -4 vehicle's performance. The fre-quency of occurrence of sensitive variables is given in Table 11, and completeresults of parameter screening sorted by terrain and vehicle are shown inTable 12. In comparing the variables that are sensitive for the FOG-M studywith those found to be sensitive for the 11MMWV, some trt.nds are apparentas all of the variables found to be sensitive for the HMMWV candidate vehi-cles were found to be sensitive for the FOG-M vehicles as well. Some vari-abl; found to be sensitive for the FOG-M analysis, however, were neversensitive for HMMWV vehicles. These variables exclusive to FOG-M includeEYE, TL, TRL, WT, OBW, RDA, and FRP; but only FRP was found to besensitive more than one to three times.


Table 9Sensitive Parameters for FOG-M Vehicles

Vehicle: AcronymDriver eyeheight above ground EYERoughness versus sped array FRMObstacle height versus speed array FVALimiting speed for obstacle height FVOFirst to last wheel center distance TLLength of track on ground per assembly TRLIUmiting speed for roughness VRIMaximum vehicle width WDTVehicle weight per assembly WGHTread width per assembly WT

Terrain:Surface roughness ACTTerrain slope GRAObstacle approach angle OBAObstacle height OHObstacle spacing OSObstacle width OBWRoad radius of curvature RADRecognition distance RDAAverage vegetation stem spacing S

Curve-fits:Drawbar pull coefficient versus excess rating cone index FDOPowered traked motion resistance coefficient versus excess

rating cone index FRP

A closer examination of the distribution of variables reveals several inter-esting facts. Although many of the same variables were sensitive for differentvehicles and different terrains, there were some exceptions. Two of the vari-ables, for example, driver eye height above ground, EYE, and recognitiondistance, RDA, were determined to be sensitive for only one vehicle and oneterrain-the M1037 on off-road terrain in the Federal Republic of Germany.

Some other parameters were also determined to be restricted to few vehi-cles or terrains. One example is the obstacle width, OBW, which was deter-mined to be a sensitive variable only for the M113A3. Other variables thatwere sensitive strictly for the M1 13A3 were track length on ground, TRL,and tread width, WT. The variable representing the first to last wheel centerdistance, TL, was found to be sensitive exclusively for the analysis of theM1037. These two vehicles, the Ml13A3 and the M1037, the lightest of thefour study vehicles, were also the only vehicles to exhibit sensitivity to surfaceroughness, ACT, and obstacle spacing, OBS. Two variables were found to besensitive for all vehicles but on only one terrain, the off-road areas of theWest Germany. These parameters were the maximum vehicle width, WDT,and the vegetation stem spacing, S.

71Chapter 5 FOG-M Candidate Vehicles

Table 10Conjectured Statistical Attributes of Parameters Selected for theError-Magnitude Scenario

Partmelsr DhvtrlutcAcronym Type Range

-10I

ACT Uniform Plus and minus 10 percent of noninal

EYE Uniform Plus and minus 5 percent of nominal

FDO Gaussian Standard deviation 5 percent of regression value

FRM Uniform Plus and minus 5 percent of table midpoint

FRP Gaussian Standard deviation 5 percent of regression value

FVA Uniform Plus and minus 5 percent of table midpoint

FVO Uniform Plus and minus 5 percent of table midpoint



OB Uniform Plus and minus 5 percent of nominal

095 Uniform Plus and minus 5 percent of nominal

OBH Uniform Plus end minus 10 percent of nominal

RAD Uniform Plus and minus 10 percent of nominal

RDA Uniform Plus and minus 10 percent of nominal

S Uniform Plus and minus 3 percent of nominal

TL Uniform Plus and minus 10 percent of nominal

TRL Uniform Plus and minus 10 percent of nominal

VRI Uniform Plus and minus 5 percent of table midpoint

WOT Uniform Plus and minus 2 percent of nominal

WGH Uniform Plus and minus 10 percent of nominal

WT Uniform lus and minus 10 percent of nominal= - -

Fingerprints and Speed Profiles

Following the isolation of sensitive variables and the establishment of theirerror magnitude scenarios, a measure of their influence on the vehicle's per-formance can be determined. This determination is done by Monte Carloanalysis in which the sensitive variables are treated as random variates and theinsensitive ones are held constant. Speed predictiens are repeated 200 timesin each terrain unit, When the maximum and minimum speeds are plottedversus the nominal speeds, a fingerprint of the vehicle is generated for a givenvehicle ou a given terrain.


Sensitivity for M993W on 5322.A90

wr . ,

FROIm in:"

COCuM

v - m m. ms

Raning

Sensitivity for M993W on 5520.R90

F=

"I-

0 0.02 0.04 0.06 0.(06 0.1 0.12 0.14 0.16 0.18 02

Ranking

Fiur 4.Sensitivityi for M993 W on 5520terRai

w t*GRAME

ACD

CUWI

FAD

FVR20E 0.0 6 0.8Rnig 1 .14 0.16 0.18 0.20 0 .0 4 !8 OI Or2 0

Figure 47. Sensitivity analysis for M993 on WGe terrain

Chapter 5 FOG-M CamOm~ie Vehicles 7

Sensitvit for M977W on 5322.A90

"e :

Sensi for M97Won 552O.R97

_ - - . S

MAW

cs t . s u

>°

ci

Figure 48. S3nsitivity analysis for M977 on WGc terrain

74 Chaptist 5 FOG-M czrvdadate verwoi.,

Senstivity for M1 13A3W on 5520.R90

ACTO

> W

Sens"t for M1 I 3A3W on 52.9

0 401 C04 069 0111 1 . 1 2 0.14 (06 0.16 0.2

Figure 49. Sensitivity analysis for M1 1 3A3 on WiGe terrain

Cr~ater rFOG-M CE~rd~date Vt.a .jA 75

Sensitivity for M1 037W on 5520.R90

FW""

* I - p : •

0 bl *A asoo 0.1 .12 &$4O, 0,16

- -- :1

* , LU ,0 LU LU L L .2 L .14 LW LW

Fgure 50. Sensitivit analysis for M1 037 on WGe terrain

76niniu

Cmmmio 5 Om a~~eVto

Senstvit for M993W on 6349.A90

room ~imin

0 M, 0IW . 8 mim:4 0.1 01

jwrft

~WS

WEACTFFSUJPX M

RCICuloffFDOWPB I II

FRTOWPB 1 1fACTRM> GRADE 1E.mt

PFATRAKWD

CLRMIN

0 c~b2 O.4Ob 601 .1 0.14 0.16

Figure 51. Sensitivity analysis for M993 on SWA terrain

77Chpo FOG M Ca.-ndatt Vetuc-low

Senstvt for M977W on 6349.R90FSUPX

FRTOMP

jSECM ~ :

SECiW:.ITS.

AVW.DRLCT

UtAW]

0~M O dl 3aSd2 0 . 3

Sens"vt for M977W on 6349.A9

Figure 52. Sensitivity analysis for M977 on SWA terrain

78 Chaptef 6 FOG-M Candsdate Vzh.Q4oc

Senstivity for M1 I 3A3W on 6349.A90

hM

ACT"Jf

0 0.6 0U4 0O0 0.0 0.1 0.12 0.14 0.6

Senstivity for M1 1 31A3W on 634 9.R90FSUPX.

R=IFDOWPB

FRUMWPRAWO -

ACTRMI:-

~TRA(WPFA

TRAJ(L CutoffACD

CLRMINASHOE

FMRDE 1_______1 go 0.b2 064 6.6 0.08 0.1 0.12

Figure 53. Sensitivity analysis for MI 1 3A3 on SWA terrain

Ch~apter 5 FOG-M Candtdate VaNdes 7

Sensitivity for M1 037W on 6349.A90

S- OM 6 & - O . 2 0.14

Sensiivity for M 1037W on 6349. R90

swks

RWA

Fourm 54. Sensitivity analysis for M 1037 on SWA terrain

80 Ch atr 5 FGGM C nddat V ath e#

Sm - u mu.m -. =.. .


sYc

Rwif

p% I 't -r

0 .0 0.04 0.06 0.06 0.1 0.12 0.14 0.16 0.1i; 0.2


* - . * * il -

AF °X

wrGRME 1 i

T"C> PFA

TP*J(WV

o 0.02 0.04 0.06 0.06 0. 0.129 0.14 0.16 0.16

Figure 55. Sensitivity analysis for M993 on ROK terrain

Chapter 5 FOG-M Cenci0date Vehicles 8


WTht

mom

m-m m pa

uciw ~Rutot


8 2 Ch p e 5 *O - CCm i a e e c

*!m Sin!1:: uic : :

W m i : . '

' wi ' ii i iiX

in 1 mmim l l~I 1u:nm1:

. ..m e e o . u 14 O e1 ! : ig

Fiue5.Snii i nlyi fo M97 on RO ter

82 Chate 5 FO-MC. !iei Vhile

p

Sensitivity for M1 13A3W on 32223.A90REACT

R=C

cimffFHVO

0 0.5 0. 0.15 22 025 0.3 026 0.

Sensitivity for M1 13A3W on 3222.R90

GROMI i ' 1

UWWA

:Cutff.

P V-o m 0.04 0.0 0.6 0.1 0.12 0.14Ranking

Figure 57. SensitiviW analysis for M1 13A3 on ROK terrain

Chapter 5 FOG-M Candidete Vehicle. 3

0t

Sensitivity for Ml 037W on 32223.A90

a : . . - . .:

: :

Sensitty for M1037W on 3222. R90

GRAM.:

AC6M.

WOH* m . :

i-i, 1

Cuof

a a .04 00 0.6 01 1 2 0 .1 0.16 0.16

-ens i vity ar sy~F ' w Mfo03 7 o ROK terrain

Chvtpt 5 FOG-M Condidets Vahicles

RE~T* ~ 4* ~ a .

Table 11Frequency of Occurrence of Sensitive Variables

Numb. of Tim. Numb.r of Times

Name Senwive Most Senive

ACTRMS 6 0

EYEHGT 1 0

FDOWPB 13 0

FHVALS 4 0

FRMS 14 1

FRTOWPB 13 0

FVOOB 8 2

FVRIDE 14 7

GRADE 21 6

OBAA 9 1

OBH 9 0

OBS 4 0

OBW 2 0

RADC 10 3

RDA 1 0

S 4 0

TL 2 0

TRAKIN 2 0

IDTH 4 0

WGHT 23 4

WT 3 0

These fingerprints are shown for each vehicle for each terrain and roadtype in Figures 59 through 102. This graph is important as it gives a view ofthe error performance of the given vehicle on the specified tewrain. If noerrors were present in data and algorithms, the fingerprint would merely be astraight line. Deviations from a straight line indicate the effects of variationsof the input par. ieters.

For the four vehicles studied, some general patterns were found in thefingerprints. In most cases, the data points were clustered or certain bulgesand spikes appeared indicating the nonuniform error performance of NRMMfor the range of vehicle speeds. The few cases in which data points can beindividually identified indicate rare occurrences in the overall perfo ,ance.

Chtof 5 FON-M Cendidato VuNclea 85

Table 12Identification of Sensitive Parameters

Terrains: 5322 - Lauterbech Scenedo: Dry, Normal552( - Schotten3349a - Arzhan Vehicles: M993, M77, M113A3, M10376343r - Arzhan32223 - Cheorweon3222 - Cheorweon

E F F F T T V WWWA G 0 00 0 R R S F FYV R VL R RDGTCR BBBBAD DRE A MO L I T H TAAHSWDA O P

Screenings5322 M993 06"

M977M113A3 " 6 0 " 6 O 0

M1037 " " 0 0

5520 M993 "M977 °M113A3 0* 00 0 0

M1037 * 6

6349a M993M977M113A3 * 6 0 0 0 6

M1037 " 0 0

6349r M993 0 0 0

M977M113A3 "M1037 "

3223 M993 a a "M977 "M113A3 P " 0 0

M1037 " " 6 "

3222 M993 0 6 0

M977 6 6 6

M113A3 " 0 6 0

M1037 0 6 • " 06

When the fingerprints were analyzed for the M993 on off-road terrain,however, the deviation between maximum and minimum speeds was not asgreat as with the other three vehicles on this type of terrain. Ibis indicatesthat the M993 is not affected as greatly by data errors as are the other candi-dates. Similar results were found for the M1037 on all on-road terrains;hence, the effect of uncertainty for on-road performance of the M1037 is lessdramatic.

The same data shown in the fingerprints can be more clearly defined by anin-unit speed profile. These in-unit profiles Figures 59 through 102, areobtained by plotting speeds versus percentage of total area or distance crossed.This percentage is based on a list sorted by nominal speed from best to worstand accumulating the terrain corresponding to each speed. These plots show

86 Chpter 5 FOG-M Candidate Vchicle*

Stchsic Spein-U i

M993W on 5322.A90

Promnt Towm Off-Road Ame

Stochastic Average Speed- M993W on 5322.A90

i L-.- - -.

S0 w0 30 w 0 7 O U I

Percan Tota Off-Road Area

Figure 59. Fingerprint and profiles for M993 on WGe off-road

Chnmnor 5 FOG-M Ciid~dt VaNclas 87

Fingerprint

M993W on 52O.R90 -Prmary

4-4

0 0 x 6 i U 30 o 4D

Stochastic Speed-in-UnitM993W on 5520.R90 - Primary

ii

PwcWmnToW Distanc

Figure 60. Fingerprint and profiles for M993 on WGe primary roads

88 Chapter 5 FOG-M Candidate Vehicle

FingerprintMM9W on 5520.R90 - Secondar

Stochasbc Spee -U

SI

Stochastic Speed-en-Unit* M993W on 6349.R90 - Secondary

3tu

Pwt Tomi Distmn

Figure 6 1. Fingerprint and profiles for M993 on WGe secondary roads

Chaptar 5 FOG-M Candidate VaNclot 8

FingerprintM993W on-55M.R9 -Trails

0 io iss i asW" Pm*WS SpWd (noh)

Stochastic Speed-in-UnitM993W on 5520.R90 - Trabl

Pwomn Totmil D~stano

Stochastic Average SpeedM993W on 5520.R90 - Trails

.j.........

a 1 0 20 is k 5 io is iii 00

Peont ToWa Distano

Figure 62. Fingerprint and profiles for M993 on WGe traits

90 Chapter 5 FOG.M Candidate VehCIcle

10 iNom*W Pfds SPWe 00~)

Stochastic Average SpeedMW on 5322A90

Peimt Total Off-Road Area

Stochastic Speed-in-UnitM977W on 5322A90

-. -- ---

wm TotajOff-Road AMe

Figure 63. Fingerprint and profiles for M977 arn WGe off-road

91Chptar 5 FOG-M Candidate Veh~claa

Fingep rintM977W on 5 20 R90 - Primary

Nordn Am~ Spewd WO)

Stochastic Seed-in-Unit

0 , 30 i |li ,o ,

Stochastic Averd-in-UnetMV97W on 5520.R90 - Primary

140-A --

12. .E

J 0 "' o 'D x so s

Stochastic Average SpeedM977W on 5520.R90 - Primary

Percent Total Distance

Figure 64. Fingerprint and profiles for M977 on WGe primary roads


In

It

1 s 30 i s '

Stochastic Sped-in-UnitM977W on 5520.W Secondary

0 O io i *i 0s

4 Pern TotlDistance

Stochastic Average Speed- M977W on 5520. R90 - Secondary

Percent Total Distne

Figure 65. Fingerprint and profiles for M977 on WGe secondary road

Ch~apter 5 FOG-M Candidate Vehicles 9

Fineprnt

NUT"n Rm~e SPeed(RO)

Stochastic Speed-in-UnitM977W on 5520.R90 -Tras

Stohati AvraeSpe

!,;,.4-7W on 5520. R90 - Trails

Percent Towa Ditne

Figure 66. Fingerprint and profiles for M977 on WGe trails


FingerprintMl 13A3W oni 5322.A9

Nw . S SOO

Stochastic Speed-in-UnitNil 13A3W on 5322.A90

Ml 13A3W n'3A 9

PNNMft TOWa OntRa AWW k

Figure 67. Fingerprint and profiles for M1 1 3A3 on WGe off-road

Chapter 5 FOG*M Candidate VaNclee 9

I

Fined-intMl 13A3W on 5520.R90 - Prmary

is io i 3' i 4

Stochasic Averae SpeednMl 13A3W on 5520.R90 - Pnmary

AS

3e .......

Percen Total Desanoe

Fiur 6. igeprntan ruis rMl 13A3 on p rimary rod

96 d

1 3&0 10 3

Figure~ ~ M13,3 6.Fnepitadpulsfor Ml 13A30o- Primarimrod

96°

Stochasti 5vrg SpeednddaeVeice

Fingerprint

M1 13A3W on 5520.R90 -Secondary

4&

0 S 4 4i c i c i

Percst Tota Distnce

Stochastic Average SpeedMl 1 3A3W on 552O.R'-On - Secondary

....... ._...-

Pern Total Distac

Figure 69. Fingerprint ahid profiltis for Ml 13A3 on WGe secondary roads

Chaptar 5 FOG*M Candidea VsNclos 9

M1 3u -n5OR0 Trails

m C441 4.

No" ftmt~ Sps~d (Mph)

Stochastic Speed-in-UnitM1 13A3W on 552.R90 -Trails

~0 il i 0 D O 5 70 0 0 C

PorcWn Tomi Distnce

Stochastic Average SpeedM I 13A3W on 5520.R90 - Trails

~~.................... ................

2 0 10 20 30 40 00 2 70 60 i 0 i O O

Percent TotW Dtsance

Figure 70. Fingerprint and profiles for M1 13A3 on WGe trails


FIngerprint

Ml 037W on 5&12MA

Stochastic Speed-in-Unit- MM1037W on 5322A90

10 U M TOW f-I a Am

Figure 71. Fingerprint and profiles for M1 037 on WGe off~roadl


Fi naerirtM1037W on .FRO - Pumwy

Stochasi vr SpeediM1037W on 5520.90 - Ptnaty

100 ~~~ CtOapattr 5vrg SpeednddeVeice

UU

{---m 6 aa oo oo 0*

Figre72 Fngrprntan pflesfoM07on We p-P r imary rod

Fiue120inep0n n poiesfrMl07onW epr 5FOr -r oad dt ehce

Finemrint

miom on 56D.g - sox

Stochastc S_/e_ :-in-UnftM1037W on 5520.A90 - Secondary

PWOWm TOW 01btw=

Stochastic Average SpeedM1 037W on 5520.R90 - Secondary

0 I 3 3 4 0 U S 0 0 OPotm TbtW Disan

Figure 73. Fingerprint and profiles for M1037 on WGe secondary roads

Chpte, 5 FOG-M Candidate VeNclos 101

FingprnM1037W o520R90 - Traib

Stochastic Speed-i(-Un)

Stochastic Speed-in-UnitM1 037W on 5520.R90 - TaIts

Percent TotW Distanc

Figue 7. Fngeprit Oad 7pofiles 2o.r0o ietra ls

1020

Chpe 5_____aniat__~ca

Stochastic Speed-in-Unit

M977W on 6349.A9

PmWOt TOW Off-RoW NA

Stochastic Average SpeedM977W on 6349A90

o o 2 30 40 50 12 70 i50 g0 ia

Permnt TotaJ Ofl*Road Area

Figure 75. Fingerprint and profiles for M993 on SWA off-road

Chapter 5 FOG-M Candidate, Vehiclas 103

FingerprintM993W on 6349.R90 - Secondary

1;

.a •, •.

.3 i ii

Ii

10 16 3 S 30 U 4i

Stochastic S e-in-UnftM993W on 6349.R9 - Secondary}:4 ____

Stochastic Average SpeedM993W on 6349.R90 - Secondary

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

i D i i i o sooe"I 0 30 30 40 5 0 6 0 80 30 ICC

Perent Total Distance

Figure 76. Fingerprint and profiles for M993 on SWA secondary roads

104Chapter 5 FUG-M Candidate Vah olas

mg onW9RK rf

N**Wn Prgd*g SPWe WO~)

Stochastic Speed-in-UnitM3W on 6349.R90 - Trails

0 ;0 o 30 io 30 k o ioo

Permit TOta MUManc

Stochastic Average SpeedM993W on 6349.R90 - Traits

3C

2 0 ao 9 0 twoPercent Total Distanc

Figure 77. Fingerprint and profiles for M993 on SWA trails

Chapter 5 FOG-M Candidate Vehicles 1 05

Finerprnt

M9 7w on 6349A9

Stocasti Argele Speed

M977W on 649A90

~ ~o7FPercen Toti Off-Road Area

Figure 78. Fingerprint and profiles for M977 on SWA off-road

106CtulPter 5 VOG-M Candidate Veh~cles

FingerprintM977W on 6349.R90 - Secondat

W77 onM4.R -Sem

StocasliISa-iU

M977W on 6349.R9 - Secondary

W. ZiZ iZ0 w~f* ;0. ' 16 io i____ _ we

Percnt Tbtal D6stan

Figure 79. Fingerprint and profiles for M977 on SWA secondary roads

Chapter 5 RF(A-M Can~didate Vo~clas 1 07

FingerprintM977W on 634§.R90 -Trails

-. £ ..& -*k% 6"

i i

Nomna P. S (mp )

Stochastic Speed-in-UnitM97"W on 6349.R90 -Trails

W

S 0 10 i0 30i5 0 0 0 o 100Percent TOtW Distance

Stochastic Average SpeedM977W on 6349.R90 -Trails

S 0 1,0 ;D i) ;0 i - 10 o

Percent Tota Distac

Figure 80. Fingerprint and profiles for M977 on SWA trails

108 Chaptri 5 FOG-M Candidate VONcIGG

FingerprintM113A3W on 6349A90

Nowin PmcSf.d SPWe (noh)

Stochastic Speed-in-UnitM1 13A3W on 6349A90

_0 "______ •..............------

0 0 '2 30 0 so U 3O S ODPOcnt TOWm Off-Road Ame

Stochastic Average SpeedM1 13A3W on 6349.A90

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

S0 10 2D 30 40 50 U 7'0 w s0 SOD

Pe-ent TotW Off-Rcad Area

Figure 81. Fingerprint and profiles for M1 13A3 on SWA off-road

Chapter 5 FOG-M Candidate VWNclas 109

FingerprintM1 13A3W on 6349.190 - Secondary

~4&

S @ is o is io w o 4No" P" SPe (W)

Stochastic Speed-in-UnitM1 13A3W on 6349.R90 -Secondary4&

-- - -- - - -...

10 iD i- k io io0 00

PeOM Total DiSa

Stochastic Average SpeedM1 13A3W on 6349.R90 -SecondaryW--, ....... ....... ..... ..... ........................ .............. .

S 10 20 30 ;0 i50 i 70 i0 ;0 100


Figure 82. Fingerprint and profiles for M1 13A3 on SWA secondary roads

1 1 0 Chpter 6 FOG-M Candidate Vehicles

FingerprintM1 13A3W on 6349.R90 -Trails

Nwnr" PmAs Speed (fO)

Stochastic Speed-in-UnitMI 13A3W on 6349.R90 -Trails

S 0 10 a; i0 40 5 iD 7;0 io 0 o OPercent ToWa Disanc

Stochastic Average SpeedM1 1 3A3W on 6349.R90 - Trails

C1. ............ .. ............ .. ..... ...... ...........--.

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

0 a 0 to 2D s o i iD i% iso i o oPercent TotaJ Distance

Figure 83. Fingerprint and profiles for M1 13A3 on SWA trails

Chapter 5 FOG-M Candidato Vahicles11

IM1 037W on 6349A90

aa

NPerentw oalOffRod (Mpha

StochasticAvr SpeedUiM1 037W on 6349A90

o 10 2 30 50 so 0 7* 0 Wo0Percent TOta Off.Road AMe

Figure ~ ~ ~ M 840igrpitad7rfWs o 1 on WAof-roa

112.......... r ....... C...d.......cl

M1 037W on 63S Secondary

Stochastic Speed-in-UnitM1 037W on 6349.R90 - Secondary

0 10 w 4 o i io io i iPero"n TOWa Disanc

Stochastic Average SpeedM1 037W on 6349.R90 - Secondary

'T o w w V

Figumj 85. Fingerprint and profiles for M1037 on GWA secondary roads

Chapwo 5 F0G-M Candidata Voniclas 11

M1Wo 9O Traft

M137 on64A0 - S 666

1141CEpe 5 l- widt OC0

Stochastic Speed-in-UnitM993Won 3m3A90

Stochastic Average SpeedM93W on 3223A90

-: -- -- -

0-. Co mo io 50 U 0 so S

Figure 87. Fingerprint and profiles for M993 on ROK off-road

Chapter 5 FOG-M Candidate Vehicles15

FlngerprintIM93 on 32.9 o-PM

In

n

a

4 4

o is is S "U Uo 4

Nan*Nd uAKind Speed (ffh)

Stochastic Speed-in-UnitM993W on 322.R90 - Prdmcy

4-

3w

22 10 io 0 40 iO i 7 i s

POWdtW Diswi

Figre88 Fngrpin an pofle9 fr399 on 322.90 pima ryrod

116 Chapter 5 FOG-M Candidate Vehicle

MM9W on3=R90 -SwrW

S ~ as :: -

Nawdni PIdm SPeed (m~h)

Stochastc S ed in-Unft

M993W on 3222.R0 Scndaiy

IC s i ~ i i io io s i

Percett TOWi Dlsmn

Figure 89. Fingerprint and profiles for M993 on ROK secondary roads

Chapter 5 FOGM Candidat Vecls17

-. Fi *-. grnt

Stochastic Average SpeedM993W on 22.R90 Trals

Fiur .0.Figepr~~an pofl frM993 on 322 OFi K Trails118

haptr 5 OG. Can~ dat VON~3S

FingerprintMg"7W on 32223.A90

o $ 1o 15 3o Ms aoS 4

Stochastic Speed-in-Unit[i: M977W on 32223.A90

4 - n - ---

o 0 0 3o 30 4is io is 0 80 1Pret Total Off-,oad Are

Stochastic AveraUe Speed

W7Mon A90

S 0 10 w0 30 40 s0 G0 so 0 MetO

M977W on 32223A 90

Prc0ri Total Ofi-rkoe Area

Figure 91. Fingerprint and profiles for M977 on ROK off-roid

119Chapter 5 FOG-M Candidate Vohcloe

Fi ngerprintM97, W on3 3=2 - PdrmW

1 0 is o i U wU Z5 4Nonn Pmdi.d SPOW (00)

Stochastic Speed-in-UnitM977W on 3222.R90 - Prmary

PereMt ToW Dstawe

Stochastic Average SpeedM977W on 3222.R90 - Primary


Figure 92. Fingerprint and profiles for M977 on ROK primary roads


Fingerpri nt

-. INM' mp2

1 10 ts o is i s40 INomri Prd Sped (nh)

Stochastic Speed-in-UnitM977W on 322.R90 - Secondary

-------------

4.

Peroent ToaJ Distac


---- ----

S 0 O O 10 2D D d0 iO 46 i o 10Percent Total Distac

Figure 93. Fingerprint and profiles for M977 on ROK secondary roads

Chapter 5 FOU.-M Candidate Vehicles 121

FingerprintM977W on 3222.R90 - Trails

NwvMn Prsdcd Speed (ffh)

Stochastic Speed-in-UnitM977W on 3222.R90 -Trails

00 1'0 io 30 405 io ;0 i6 0 100

Stochastic Average SpeedM977W on 3222.R90 - Trails

...........

S 0 r0 2 3 O O 40 X 70 ;0 0 w0

Pement TotI D~tance

Figure 94. Fingerprint and profiles for M977 on ROK trails


FingerprintMI 1 3A3W on 32223.A90

1ok P. 4PW(0

0 10 15 30 U so 0 so 6 0 0

Stochastic Speed-in-UnitM1 13A3W on 322A90

70 ......fiRod

p.o.. -. ........-

Percent TotaJ Off-Road Area

Figure 95. Fingerprint and profiles for MI 13A3 on ROK off-road

Chaptor 5 FOG-M Candidate Vehicles 1 23

FingerprintM1 13A3W on 3222.R90 - Primary

I~4L-I--- iC-

Nwr ftce Sp (ff)

Stochastic Speed-in-UnitM1 13A3W on 3222.R90 - Prmary

0 1 0 i pi D i D 3 c i 0

Stochastic Average SpeedM1 13A3W on 3222.R90 - Primary

A--N ------

Pent Total DistanceStocastcerage ToWDSpee

Figure 96. Fingerprint and profiles for M113A3 on ROK primary roads

124 Chaptr 5 FOGM Candidte Vehicles

Fingerprint

M1 13A3W on 3222.R90 -secondary

-4A

j£

Si i5 i3 1 6 40 4

NodPmded Total (nst h)

Stochasic Speed-in-UnitM1 13A3W on 3222.R90 - Secondary

-10 10 iD i3 i4 i3 i6 70 soPement Total Dsance

Figure 91. Fingerprint and profiles for M1 13A3 on ROK secondary roads

Chaptor 5 FOG.M Candidate VoNclas12

M113AWo~3~R9O~aiLs

M1 13A3W f322R90Tris

o a I 0 0 0 74I g 0

Ml 13A3 on 222 R90- Mailsk

S0 10 i 0 is 7A as0 is10

StocPertc entToi Dstaii

Figure ~ ~ ~ M 98.3 Fnepitadrofie fo22r 9 -l T3a3 nR ra

Ctocpaser Avrg SOpeCnddaeedce

FingerprintMl 037W on 3222A90

'Ac 0 ___ __________

Stochastic Speed-in-UnitM1037W on 32223A90

Pecant Total Off.Ro~d Area

Stochastic Average SpeedM1 037W on 32223.A90

Figur ~*'~Pement Total Off-Road Area

Fiue99. Fingerprint and profiles for M 1037 on ROK off-road


FingerprintMlQ037W on Y&2R90 - PnrnaryM1037W on 3222A90 - Pdmaw

, oow TM OM i o w

Stochas Averag -SpeeM1 037W on 3222.R0.- Pdmary

jii .. , -. ..-

00

Stochastic Speraei SpU

M1 037W on 3222.R90- Primary

Paor TbtW O*Mrceo

Figure 100, Fingerprint and profiles for M 1037 on ROK primary roads


FingerprintM1 037W on 322R90 - Secondary

I" "

Not P- Sped (0h)

Stochastic Speed-in-UnitM1 037W on 322R90 - Secondary

4 . i • iPamse Toli Dbmno


Percent Toti Disance

Figure 101 .Fingerprint and profiles for M1037 on ROK secondary roads


FingerprintMI 037W on 3222.1190 - Trails

0 0 iO dug iNmi Am~s SPIPW(nO)

Stochastic Speed-in-UnitM1037W on 3222AR90 - Trails

I0

to Percnt Tow Dfstr"c

Stochastic Average SpeedM1 037W on 3222.R90O- Trals

w t

Peroam Totai Distnc

Figure 102. Fingerprint and profiles for M 1037 on ROK trails

1 30 Chapter 5 FOG-M Candidae Vehicle,

that as a higher percentage of terrain data is accumulated, the nominal veioci-ties decrease. As the percentages continue to increase, the plots have a sharpdrop-off indicating where the speeds reach very low or zero magnitudes-aNO-GO condition. The in-unit profiles are depicted in Figures 59 through102 with their corresponding fingerprints.

The third speed profile that can be obtained from the NRMM output is theaverage speed profile. Plots of the average profile are included with theircorresponding fingerprints and in-unit profiles in Figures 59 through 102.Once again the same data as used in the fingerprints and in-unit profiles areprocessed for average speed profiles. Accumulated areas and distances com-pute for the in-unit data are again used, but this time the speeds are weightedbased on these areas or distances.

For the average speed profiles, three curves are generated: minimum,nominal, and maximum velocities. In many cases the minimum speeds arecharacterized by low-magnitude, irregularly shaped patterns. This representa-tion arises from the occurrence of very low or NO-GO minimum calculatedvelocities.

Missions and Mission Rating Speeds

Once the average speed profiles are generated, the data extracted fromthem can be used to compute the required mission rating speeds as defined inChapter 4. Required average speeds include speed off road, VC, speed onprimary roads, VP, speed on secondary roads, VS, and speed on trails, VT.Depending on the mission being considered, these speeds are computed atdifferent percentage of terrain challenged. These appropriate speeds are givenby terrain and vehicle in Table 13.

The three mission rating speeds have identical definitions in Germany andRepublic of Korea. The first mobility level, tactical high, has the highestrequirement for off-road utilization. For this level, 50 percent of the missionis operated off road, and 90 percent of the off-road terrain must be negotiable.Primary roads comprise only 10 percent of the tactical high mobility mission,but 100 percent of these roads must be successfully traversed by the vehicles.Similarly, 100 percent of the secondary roads and the trails must be negotiablebut 30 percent of the mission is assumed on secondary roads and 10 percenton trails.

The tactical standard mobility level reduces the requirements for off-roadtravel with only 15 percent of the total mission performed off road with only80 percent of the terrain negotiable. Primary roads, secondary roads, andtrails represent 20 percent, 50 percent, and 15 percent of the total missionrespectively. For these three elements of the network, 100 percent negotiationis required.

Chapter 5 FOG-M Candidato Vehiclea 131

Table 13Summary of Stochastic Average Speeds - MPH

West Germany

Minimum

Vehicle VC90 VC8O VCS0 VTIOO VT50 VS100 VP100

M993 0.351 2.327 10.626 10.316 18.809 17.735 23.589

M977 0.347 4.974 13.998 9.751 12.219 19.116 27.814

M113A3 0.485 2.942 4.705 12.021 17.854 21.794 28.871

M1037 0.221 1.730 1.203 13.166 19.681 35.966 43.780

Nominal

Vehicle VC90 VC80 VCSO VT100 VTS0 VS100 VP100

M993 9.906 12.s.2 17.890 12.456 21.114 20.289 25.970

M977 2.864 11.300 16.960 11.316 14.322 21.817 31.073

M113A3 10.914 13.983 19.708 14.394 20.884 24.759 31.399

M1037 0.673 18.697 25.149 15.502 24.313 37.774 45.694

Maximumm -i

Vehicle VC90 VC80 VCSO VTI0 VTSO VS100 VPI100

M993 11.549 14.446 20.184 14.633 22.304 22.968 28.264

M977 4.230 13.565 18.296 13.088 15.903 25.096 32.809

M113A3 13.523 16.559 21.979 17.285 24.099 28.112 33.333

M1037 0.686 21.453 27.776 18.283 26.684 38.993 47.170

Southwest Asia

Minimum

Vehicle VC90 VC80 VC50 VTIO0 VT8O VS100

M993 0.0538 0.118 0.509 4.778 23.288 12.785

M977 0.0352 0.0517 0.768 11.805 13.230 14.767

M113A3 0.0407 0.0659 1.354 16.463 21.533 16.173

M1037 0.0323 0.0445 2.801 17.629 21.341 28.000

Nominal

Vehicle VC90 VC80 VC50 VTIO0 VT80 VS100

M993 0.0688 0.260 13.019 5.882 26.594 14.393

M977 0.0400 0.0646 11.415 13.544 15.611 16.546

M113A3 0.0454 0.0812 14.547 19.968 25.950 18.289

M1037 0.0362 0.0530 21.156 22.062 27.661 29.895

Maximum

Vehicle VC90 VC8o VC5o VT1OO VT8O VS100

M993 0.0709 0.262 15.299 13.113 29.483 16.147

M977 0.0424 0.0676 12.893 15.542 18.076 19.242

MI13A3 0.0520 0.0892 17.47 1 23.678 30.099 20.583

M1037 0.0391 0.0574 26.838 27.663 34.962 32.866

(Continued)

1 32 Chapter 5 FOG-M Candidali Vehicles

Table 13 (Concluded)Republic of Koma

Minimum

Vehice VC90 VC80 VC5O VTo0 VTsO VS100 VP100

M993 0.128 0.405 0.800 1.755 24.196 16.477 22.220

M977 0.0355 0.052 0.739 3.503 13.544 18.145 28.284

M113A3 0.0402 0.0645 0.397 2.415 22.847 20.679 27.251

M1037 0.0274 0.0351 1 3.769 0.994 23.481 32.772 38.565

Nominal

Vehicle VC90 VC80 VC50 VTOO VT50 VS100 VP100

M993 0.252 6.605 8.453 2.332 26.694 19.023 25.048

M977 0.0438 0.0756 7.050 6.148 16.013 21.058 30.386

M113A3 0.0465 0.0854 9.150 5.533 26.395 23.401 29.929

M1037 ] 0.0318 0.0436 14.569 6.981 30.840 35.374 41.249

Maximum

VehCiclS VC90 VCS0 VC50 MOO0 VT50 VS100 VI0

M993 0.269 7.109 9.849 3.184 28.118 21.647 27.293

M977 0.0556 0.100 7.817 6.805 17.945 24.368 31.708M113A3 0.0901 0.187 10.401 6.981 29.415 26.326 32.044

M1037 0.0438 0.0631 16.283 9.423 36.489 38.312 43.503

The tactical support mobility level in WGe and ROK is only slightly influ-enced by off-road performance as only 5 percent of the mission is required tocover such terrain with 50 percent of the area successfully traveled. Primaryroads represent 30 percent of the support mission with 100 percent negotiationrequired, arid trails comprise only 10 percent of total scenario with 50 percentof the trails negotiable.

With no primary roads in the Southwest Asia terrain, secondary roadsbecome increasingly important. For the tactical high mobility level, 50 per-cent of the total mission is assumed to be off-road with 90 percent of the arearequired to be negotiable. Tactical high mobility weighs the secondary roadsand trails equally with each comprising 25 percent of the mission at 100 per-cent successful negotiation.

For the tactical standard level of mobility, the off-road requirements arereduced. In this case only 15 percent of the total mission occurs over off-roadterrain with 80 percent of the terrain negotiated. This time, secondary roadsmake up 50 percent of the mission and trails, 35 percent. Both the secondaryroads and trails must be negotiable 100 percent.

The influence of off-road operation is minimized while secondary-roadtravel is maximized for the tactical support mission rating. In this case, only5 percent of the mission is required to operate on off-road conditions with


only 50 percent of the terrain negotiable. The secondary roads, however,contribute 60 percent of the total travel required for support operations. Ofthe secondary road distance, 100 percent must still be successfully traveled.Trails in this computation represent 35 percent of the total mission with an80 percent negotiation of the terrain required.

With the zvailable data from NRMM, a maximum and minimum ratingspeed was computed for each vehicle on each terrain in addition to the nomi-nal. This gives an indication of the mission rating speed range to provide aclearer picture of the vehicle's performance on the given terrain. The missionrating speeds are depicted in Figures 103 through 105 by terrain and tacticalmobility level.

In order to facilitate comparisons of the vehicles, a global mission ratingspeed was developed. This global speed is computed as a weighted averagefor each tactical mobility level based on the percentage of time the vehicleswould be expected to operate on the examined terrains. The global scenarioassumed 50 percent of the vehicle's use would be on terrain approximated bythe Southwest Asia conditions, 30 percent on terrain similar to the Republic ofKorea, and only 20 percent on areas resembling those in West Germany.These global mission rating speeds are shown by mobility level in Figure 106.

Comparison of Historic and Stochastic Forecasts

In the historic evaluation of the FOG-M candidate vehicles, the vehicularperformance and ranking was based on studies of many different terrain typesas well as scenarios. The percentage of distance which the vehicle could nottravel, a NO-GO condition, was also used in making final decisions. In com-paring the stochastic results with the previous analysis, only the total area withthe dry-normal condition will be used. These comparisons are made based onindividual terrain performance as well as overall performance.

Within West Germany, the two vehicles, of the four considered, found tobe dominant in the original study were the M993 and the M1037. For all ofthe on-road travel, the M1037 was found to obtain the highest average speeds.For the off-road transportation, the M1037 also obtained the highest speedsexcept for the challenge level of 100 percent of the terrain negotiated forwhich the M993 had a slight advantage. Taking the computed average speedsinto consideration, the M1037 was chosen as the best candidate for tacticalsupport missions. Although the M1037 maintained the highest average speedsthroughout this region, the percent NO-GO's led to the selection of the M993as the best vehicle for tactical high and tactical standard levels ofperformance.

In the stochastic analysis of the candidate vehicles in West Germany, simi-lar results were found. The M1G37 clearly outperformed all other vehicles, inrespect to minimum, nominal, and maximum speed, on nearly all terraintypes. The only exception was for the off-road terrain. At the 90 percentchallenge level, the MI 13A3 maintained the highest minimum, nominal, and


Mission Raftn SpeedFRG - LMaStrach ISchotte

2 -

M113A3 M93 M1037 M9"7VN*.

Mission Rating SpeedFRG - Lauterbach / Scoten

325-

15-

M1I13AM Mm9 MIcX7 M7

Figure 103. Mission ratngspedsgo W ee

3135

Chptr OGM 25-d'-eVeice

Mission Rating SpeedSWA - 6349

0.164!

01.

1.6

1.2

I 0.8'

0.6" -

I-0.2"

OW

M113A3 mm993 MI7 M977Vehide

Mission Rating SpeedSWA - 6349

35"

1 1- . -

Ml13A3 M993 M1037 M977Vehid

Figure 1M04.Mission rating speeds for SWA

136 Chapter 5 FOG-M Candkdate Vehicles

Missio Rang Speed

ROK - 3=2

0.46

043.

~02.

M1133 M903 M1cO37 M977VeNce

Mission Rating SpeedROK -3222

M113A3 M993 M17 M977

Vehicle

Mission Rabrng SpeedROK -3222

~35- _ _ _ _ _ _

~25-

1:1

E5 "

M113A3 M993 M1037 MW977Vehicle

Figure 105. Mission rating speeds for ROK

Chaptat 6 FOG-M% Candidato Velvcles 1 37

Gob Missio Rabrig Speed

03-

Oz

M113A3 M993 M1037 M977Vewd.

GlJob Misn Rdng Speed

3-

25-

0. M113A3 M993 M1037 L1977

FigureGlba 106. Global mispnrainepdd

138chat~i 5 OGM Cndk~tuV40-u

maximum speeds. For all other off-road percentages, the M1037 producedthe highest nominal and maximum speeds, but the M977 had the largest mini-mum speed. Based solely on maximum and nominal mission rating speeds,the MI13A3 should be selected for tactical high missions in West Germanyand the M1037 for the standard and support tactical levels. If decisions werebased on minimum mission rating speeds, the M113A3 would still be the bestperformer at tactical high levels, but the M977 would be better for tacticalstandard and support requirements.

When the four vehicles of interest were originally compared on the South-west Asian terrains, again the performance of the M1037 and the M993 wasoutstanding. Also evident this time was the performance of the Ml 13A3. Inthese studies, the M1037 obtained the highest average speeds over secondaryroads as well as off-road at the 50 percent challenge level. When the vehicleswere compared for performance on trails, the M993 was found to have thehighest speed. For the more demanding off-road travel, 80 percent and90 percent terrain negotiation, the M993 and the M I13A3 performed equallywell. Taking into consideration mission rating speeds as well as NO-GOsituations, the M993 was again chosen as the best candidate for tactical highand tactical standard mobility requirements with the M1037 again selected fortactical support.

Stochastically, there was very little deviation from the deterministic analy-ses performed for Southwest Asia. As with the historical study, the M993achieved the highest average speeds on the more demanding off-road challengelevels, and the M1037 had the greatest speed at the 50 percent challenge level.The M1037 was also found to be the best performer over all of the on-roaddistance including trails. When the stochastic mission rating speeds werecompared, not surprisingly the M993 had the highest minimum, nominal, andmaximum tactical high aiid tactical standard rating speeds. Also in line withthe computed speed profile, the M1037 was found to be the best candidate fortactical support implementation with the highest minimum, nominal, and maxi-mum support rating speed.

The historic ranking of vehicles in the Republic of Korea yielded resultssimilar to those in the other two terrains. Once again, the M1037 was clearlythe most rapid vehicle for all on-road transportation as well as for the 50 per-cent negotiated off-road terrain. As before, the M993 was the clear choice forthe most difficult off-road manipulation. These performances led to the famil-iar choices of the M993 for tactical high and tactical standard missions and theM1037 for tactical support implementation. The stochastic results for theKorean terrain led to the same conclusions with the M993 dominating othercandidates for tactical high and tactical standard missions, and the M1037clearly besting the other vehicles for tactical support.

The goal of the original study was to select the best vehicle from a list ofcandidates to be utilized in the FOG-M component of the Forward AirDefense. Since the historic study was based on several terrain subtypes andsurface conditions, the vehicles could be ranked by examining the frequencyof their selection as best candidate for tactical-high mobility on each type of


condition. This criterion was further restricted by using the high-high mobil-

ity level as the determining factor in similarly performing vehicles.

The best vehicle selected in accordance with these stipulations was the

M993 F IS followed in order by the M977 HEMTT, th M1 13A3, and finally

the M1037 HMMWV. Since only the dry-normal surface condition and total

area were analyzed in the stochastic analysis, the vehicles' performance in this

scenario is used for ranking. To clarify the overall performance of the vehi-

cles, the results of the global mission rating speed calculations will be used.

When comparing the computed global mission speeds, the M993 is still the

candidate of choice for tactical-high mobility performance, and hence the

vehicle of choice for the FOG-M application. The second choice of the four

studied vehicles is the M I13A3 followed in order by the M977 and the

M1037. Although the vehicle of choice from the stochastic analysis is consis-

tent with that of the deterministic study, the order of selection is modified.

This is the result of evolutions in the deterministic NRMM between the origi-

nal study and the present. The stochastic approach has added depth to the

FOG-M mobility study by revealing the range of velocities the candidate

vehicles can be expected to achieve. This adds confidence to the ranking of

candidate vehicles.

140 Chapter 5 FOG-M Candidate Vehicle6

Table 14

Stochastic Mission Rating Speeds - MPH

Tactical High Tactical Standard Tactical Support

West Germany

M993 0.69 12.96 14.99 8.65 17.77 20.21 18.60 21.65 24.08

M977 0.68 4.99 7.10 12.68 17.89 20.60 19.49 22.32 24.99

Ml 13A3 0.95 14.69 17.72 10.71 20.96 24.08 19.28 25.58 128.58

M1037 0.44 1.32 1.34 8.58 28.32 31.00 14.54 36.73 38.44

Southwest Asia

M993 0.110 0.140 0.140 0.72 1.49 1.59 6.24 17.04 19.12

M977 0.070 0.080 0.085 0.34 0.42 0.44 7.56 15.88 18.37

M113A3 0.081 0.091 0.104 0.43 0.53 0.58 11.08 20.11 22.91

M1037 0.065 0.072 0.078 0.29 0.35 0.38 17.96 28.50 33.19

Republic of Korea

M993 0.251 0.489 0.524 2.02 8.24 10.14 8.81 19.78 22.21

M977 0.071 0.087 0.111 0.34 0.49 0.65 8.63 20.27 22.73

M113A3 0.080 0.093 0.179 0.41 0.55 1.18 5.96 23.38 26.00

M1037 0.055 0.063 0.087 0.23 0.29 0.42 23.76 33.90 36.95

Global Mission Rating Speeds

M993 0.162 0.233 0.242 1.16 2.61 2.82 8.01 18.61 20.85

M977 0.086 0.103 0.116 0.42 0.55 0.62 9.00 18.09 20.65

M1 13A3 0.099 0.114 0.154 0.52 0.67 0.89 9.44 21.97 24.78

M1037 0.073 0.085 0.100 0.33 0.40 0.49 18.44 31.41 35.23


6 Conclusions andRecommendations

Conclusions

Based on this study, the following conclusions can be drawn:

a. Extension of stochastic mobility forecasting procedures to encompassrealistic quantities of data requires intense computation. Ie proceduresdescribed in this report call for a supercomputer environment and, whilecapable of satisfactory service in a research setting, would be unsuitablefor the tactical setting that motivated their development.

b. A potential breakthrough was glimpsed that may provide an order-of-magnitude reduction in computational overhead. This involves thedevelopment of stochastic mobility forecasting products using far fewerthan the full complement of terrain units contained in current maps.

c. Unrelenting advances in computational speed and capacity of portabledesk-top machines makes the final attainment of tactical stochasticmobility forecasting highly probable.

d. Application of the procedures to two historic studies confirmed theearlier outcomes attained with the deterministic form of NRMM andillustrated how the stochastic components provided additional informa-tion useful in ranking candidate vehicle designs against specified operat-ing missions.

Recommendations

The following recommendations are made:

a. By means of analysis of existing data for the independent variables, per-sonal interviews or by the covening of a panel of expert mobility dataa-quisition personnel, obtain and use better information on the

Chapter 6 Conclusions and Recommendations 143

uncertainty of the independent variables to more closely focus on sensi-tivity thresholds and error-magnitude scenarios.

b. Follow up on the lead reported here suggesting that the products of astochastic mobility forecast for a quad can be obtained with far lessterrain data units than used heretofore. Transfer operations from thesupercomputing environment to a desktop machine at the cutting edge oftechnology. Upgrade that machine as technology advances. Convertsoftware from its present "breadboard" state to one of refinement,uniformity, and friendliness.

c. Document as Report 3 in the current series the many new and intricateelements of code used to convert NRMM to a stochastic model. Docu-ment also the postprocessing software used in both the supercomputingand personal computing environments to create the products of thestochastic mobility forecast.

d. Create tactical offensive and defensive localized scenarios, preparestochastic estimates of (1) travel time for several avenues of approachand (2) areas which must be defended against a mechanized attack.Explore means of expressing tactical decision parameters in stochasticterms.

e. Prepare and conduct classroom and field exercises using the militarypersonnel who would be responsible for deploying the stochastic versionof NRMM in a tactical setting. Learn from these activities howstochastic outputs are best presented to their intended audience.

f Undertake a limited program of field validation of these methods. Thiswould involve selection of several terrain units, characterizing them bymeasurements repeated at many different physical locations and travers-ing them with a vehicle many times using different physical paths. Inthis way, probability d.nsity functions can be developed for the terraindata and the recorded vehicle traversal speeds. Comparisons wouldfollow with NRMM responses to the same inputs.

g. Extend stochastic iurecasting methods to all applications that useNRMM as a foundation.

h. The mission rating speed ranges, based as they are on minimum andmaximum predicted speeds, depict worst-case error performance byNRMM. Users should be alert to the possible need to resolve NRMMerror performance more selectively by appealing to analysis of themission rating speed probability densities rather than the ranges. Thiswould allow the use of well-known and accepted statistical proceduresfor the formulation of confidence intervals and for the testing of thesignificance of differences among vehicles.

144 Chepter C Conclusions and Recommendations

References

Green, C. E., Randolph, D. D, Grimes, K. (1983). "Ride and Shock TestResults and Mobility Assessment of HMMWV Candidate Vehicles," Tech-nical Report GL-85-10, USAE Waterways Experiment Station, Vicksburg,MS.

Haley, P. W., Jurkat, M. P., and Brady, P. M. (1979). "NATO ReferenceMobility Model, Edition 1 User's Guide," Technical Report 12503,U.S. Army Tank-Automotive Command, Warren, MI.

Lessem, A., Ahlvin, R., Mason, G., and Mlakar, P. (1992). "StochasticVehicle Mobility Forecasts Using The NATO Reference Mobility Model.Report 1: Basic Concepts and Procedures," Technical Report GL-92-11,USAE Waterways Experiment Station, Vicksburg, MS.

Robinson, J. H., Smith, R. P., and Reaves, A. M. (1987). "Mobility Per-formance of Candidate Vehicles for the Fiber-Optic-Guided Missile Com-ponent of the Forward Air Defense System," Technical Report GL-87-26,USAE Waterways Experiment Station, Vicksburg, MS.

Unger, R. F. (1988). "Mobility Analysis for the TRADOC Wheeled VersusTracked Vehicle Study," Technical Report GL-88-18, USAE WaterwaysExperiment Station, Vicksburg, MS.

145Reorences

" Form Approved

REPORT DOCUMENTATION PAGE / F No 07proo,eubihh reporting burden for tis collection of information is estimated to average I hour per response, including the time for structions, searchig eisting date sourc40,

gatheing and maintaining the data needed, and completing and reviewing the collection of information Send comments re.ardng tins burden estimate or any other aspect of thiscollection of information, including suggestions for reducing this burden tu ivashington Headluarters Services. Directorate for information Operations and Reports. 1215 Jefferson

avis Highway. Suite 1204. Arlington, VA 22202-4302. and to the Office of Management and Budget. Paperwork Reduction Project (0704.0 18). Washington. DC 20503

1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED

I July 1993 Report 2 of a Series4. TITLE AND SUBTITLE S. FUNDING NUMBERS

Stochastic Vehicle Mobility Forecasts Using the Military RDTENATO Reference Mobility Model Task No. AT40-AM-011

6. AUTHOR(S)

Allan Lessem, Richard AhlvinPaul Mlakar, William Stough, Jr.

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATIONREPORT NUMBER

U.S. Army Engineer Waterways Experiment Station, Geotechnical Technical ReportLaboratory, 3909 Halls Ferry Road, Vicksburg, MS 39180-6199 GL-93-15JAYCOR, Structures Division, 1201 Cherry Street, Vicksburg,MS 39180

9. SPONSORING/ MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/ MONITORINGAGENCY REPORT NUMBER

U.S. Army Corps of EngineersWashington, DC 20314-1000

11. SUPPLEMENTARY NOTESThis report is available from the National Technical Information Service, 5285 Port Royal Road, Springfield,VA 22161.

12a. DISTRIBUTION /AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE

Approved for public release; distribution is unlimited.

13. ABSTRACT (Maximum 200 words)

This report is the second in a series that documents the conversion of the NATO Reference Mobility Modelfrom a deterministic code to a stochastic one. The first report described basic concepts and procedures and dealtwith relatively small quantities of data as methods were "bootstrapped" into existence. This report describes theextension of the procedures to realistic quantities of data.

In addition, the methods are applied to two historical mobility assessments that were influential, years ago, inthe procurement of some Army vehicles now in the inventory. When the studies were originally done, NRMM wasused in a completely deterministic way and no consideration was given to the perturbing effects of uncertain dataand of scatter associated with field-data-based algorithms. The analyses were repeated (albeit with greatly limitedscope) using stochastic mobility forecasting methods as a means of introducing the NRMM user community to thesemethods in an inherently interesting way.

14. SUBJECT TERMS 15. NUMBER OF PAGES,i,,ility modeling 156

Stochastic modeling 16. PRICE CODE

17. SECURITY CLASSIFICATION 18 SECURITY CLASSIFICATION 19 StCURITY CLASSIFICATION 20. LIMITATION OF ABSTRACT

OF REPORT OF THIS PAGE OF ABSTRACT

UNCLASSIFIED) UNCLASSIFIED UNCLASSIFIEDNSN 7540 0 1.280-5500 Staoddrd ) iryi 298 (Rev 2 89)

Stochastic Vehicle Mobility Forecasts Using the … · Mobility Model Report 2 Extension ......

Documents

Transcript of Stochastic Vehicle Mobility Forecasts Using the … · Mobility Model Report 2 Extension ......