Calhoun: The NPS Institutional Archive DSpace Repository Theses and Dissertations 1. Thesis and Dissertation Collection, all items 1976-06 The estimation of Lanchester attrition-rate coefficients for an aggregated combat model Lee, Yin-chu; Pi, Yu-kung http://hdl.handle.net/10945/17749 Copyright is reserved by the copyright owner Downloaded from NPS Archive: Calhoun THE ESTIMATION OF LANCHESTER ATTRITION RATE COEFFICIENTS FOR AN AGGREGATED COMBAT MODEL Lee, Yin-chu •&ST NAVAL POSTGRADUATE SCHOOL Monterey, California THESIS THE ESTIMATION OF LANCHESTER ATTRITION RATE COEFFICIENTS FOR AN AGGREGATED COMBAT MODEL by Lee, Yin-chu and Pi, Yu-kung June 1976 Thesis Advisor James G. Taylor Approved for public release; distribution unlimited. T174013 SECURITY CLASSIFICATION OF THIS PACE (When Dote Entered) READ INSTRUCTIONS BEFORE COMPLETING FORM REPORT DOCUMENTATION PAGE HEPORT NUMBER t 4. TITLE (m\<i 2. OOVT ACCESSION NO. Submit) 3. RECIPIENT'S CATALOG NUMBER 5. TYRE OF REPORT The Estimation of Lan Chester Attrition- Rate Coeffi cients for an Aggregated Combat Model 7. AUTHORfaJ ft PERIOO COVERED Master's Thesis June 1976 « PERFORMING ORG. REPORT NUMBER ft. CONTRACT OR GRANT NUMBER^.) Lee Yin-chu Pi Yu-kung , , PERFORMING ORGANIZATION NAME ANO AOORESS i 10. PROGRAM ELEMENT. PROJECT, TASK AREA WORK UNIT NUMBERS ft Naval Postgraduate School Monterey, California 93940 II. CONTROLLING OFFICE NAME ANO AOORESS 12. Naval Postgraduate School Monterey, California 93940 14. MONITORING AGENCY NAME ft REPORT DATE June 1976 13. ADDRESS*"!/ different from Controlling Office) IS. Naval Postgraduate School Monterey, California 93940 NUMBER OF PAGES 58 SECURITY CLASS, (ot thlm report) Unclassified Mm. DECLASSIFICATION/ DOWNGRADING SCHEDULE 16. DISTRIBUTION STATEMENT (of thlm Report) Approved for public release; distribution unlimited, 17. DISTRIBUTION STATEMENT IS. SUPPLEMENTARY NOTES 19. KEY WORDS <' (ot the mbmtrmct entered In Continue on rewerme mlde It Block 30, It different horn Report) necemmmry mnd Identity by block number) Lanchester-type combat models Markov- dependent fire Estimation of Lanchester attrition-rate coefficients Theater- level combat models 20. ABSTRACT (Continue on reverme aide II necmmmmrr end Identity by block number) This thesis oonsiders the problem of estimating Lanchester attrition rate coefficients for an aggregated Lanchester type theater level combat model, BALFRAM, which has been used for various high level defense planning purposes. Several alternative coefficient estimation methodologies are examined, with their strengths, weaknesses, and problems of implementation in BALFRAM being discussed. Data requirements for coefficient estimation and approaches to aggregation are also discussed. DO .SS, (Page 1) 1473 EDITION OF NOV «• S/N 010I-014-«601 1 | IS OBSOLETE SECURITY CLASSIFICATION OF THIS PAOE (When Dmtm Entered) THE ESTIMATION OB LANCHESTEH ATTRITION-BATE COEFFICIENTS FOB AN AGGBE6ATED COMBAT MODEL by Lee- 1£in-chu Col, Bep. of China Air Force M.E.A., Tamkang College, China, 1974 Pi, Yu-kung Lt Col- Bep. of China Arm? M.E.A., Tamkang College, China, 1974 Submitted in partial fulfillment of the reguirements for the degree cf BASTES OF SCIENCE IN OPEBATIONS BESEABCH from the NAVAL POSTGBADOATE SCHOOL June 1976 LIBRARY DUDLEY KNOX PO NAVAL MONT ABSTRACT Tbis Lanchestei thesis considers attrition — the problem of estimating rate coefficients for an aggregated lanchester-type theater- level combat model, level EAIFBAK, lihich has been used for various high defense planning purposes. Several alternative coefficiect- estimation methodologies are examined, weaknesses, and problems of nith their strengths, Data implementation in BALFRAM being discussed. reguirements for coefficient estimation and approaches to aggregation are also discussed. TABLE OF CONTENTS I. II. III. INIfiOEDCTION A. RBI MODEL COMBAT B. OVIBVIER OF THESIS 7 8 . MEIHCECICGIES FOR THE ANALYSIS CF CCBEAT 9 A. 11B GAMES E. SIMULATION 10 C. ANALYTIC MODELS 11 9 EilFfiAM 13 A. GENERAL DESCRIPTION 13 E. SYSTEM APPLICATION 14 C. D. 1. System Organization 2. Systea Operation.. - ... .. 16 .... 16 Lanclieste^s Classic Combat Models. 16 2. Differential Fight Laws 19 a. Square Law.... b. Linear Lav.... - 19 20 EALEBAM PBOCESS 21 1. Scenario Geography 23 2. Force Characteristics 23 3. Contingency Logic - • 24 a. Battle Logic 25 b. Movement Logic 25 Inputs and Outputs NODH Program a. b.. DCSF Program 26 Capabilities. 27 UTILIZATION OF BALFRAM IN REP. OF CHINA ESTIMATION CF ATTRITICN-RATE COEFFICIENTS A. 14 1. 5. E. ..» HMBEMATICAL BACKGROUND 4. IV. 7 MABKOV EEPENDENT FIRE 26 26 30 33 33 ¥. ¥1. E. ECNDEB'S METHOD 34 C. CliBK'S MODEL 37 DATA BECUIREMENT FOB COEFFICIENT ESTIMATION 40 A. HISTORICAL DATA 40 E. LCGISTICS DATA 42 C. WEAPON DATA 43 D. QUALITATIVE DATA 44 AGGBECATICN OFFOBCES 45 A. HCTIONAI UNIT B. FIBEPCWEB SCOBES C. TEE OSE OF SATELLITE MODELS D. TEE APPLICATION OF EONDEB«S APPROACH IN 45 , 46 - 47 EAIFBAM. 48 VII. FISAI EEMABKS LIS! OF EIGUBES LIST OF EEFEBENCES 50 , 52 , • 53 INTRODUCTION I. A. SHY KCDEI CCMEJST came war A has been a military whatever means, cf more cppcsirg forces, defined (8) as "a simulati.cn, by operation conducted involving using tfco data rules, or and prccedures designed to depict an actual or assumed real life It is a systematic method of studying military situation." prctlems and can provide identifying errors or a means gaining of shortcomings, experience, and improving skills without paying the penalties of the real world. One the significant values of war gaming is ttat it cf can provide an impelling stimulus to innovation, and creativeness . It challenges and motivates different kinds establishes a an responsible mcti^aticn environment that participant. Thus cf war games have been used extensively to train officers in military forces throughout the world. Ancther advantage of war gaming is that a war garce can be played on any hypothetical terrain which may not te areas ccctrolled by the nation sponsoring the game, as must te the case in field tests and maneuvers that employ real fcrces. Any military nation may employ war gaming tc assess or test contingencies military requirements military fcrces must be anle tc cope to ensure the security and and survival cf the nation. the with which B. OVEBvIJ* CE THESIS paper is composed of seven chapters. In chapter This war gaming is Chapter 2 gives combat. These are mcdels. Chapter discussed three its different values methods simulation games, nar 3 and introduces a 1, described. are modelling for analytical and computerized analytical model EALEBAfi and discusses its analysis use in the Bepublic of Chapter China. means of introduces Markov-dependent fire and 4 calculating coefficient. Chapter the 5 attrition — rate Lanchester discusses the historical, logistic, weapon and qualitative data which are required for input simulation oi war gaming models. Chapter methodologies such as notional unit, the 6 examines several firepower scores and of satellite models for the aggregation of fcrces, use and the prchlem area to apply Eonder*s approach in Chapter to 7 gives seme final remarks. EAIFEAM. METHODOLOGIES fOR THE ANALYSIS OF CO ME AT II. A. WAR GAMES war game, as defined by Ecnder A frcm (2) reality of a field experiment or the removed field exercise a representing wherein crly teams of players step is a , commanding the officers and their staffs are included. Chinese people invented a game called "Kei Chi" and it is still played tcday on a stjlized map board with black and white colored stones and wen ty the Ahcut 3QC0 B.C., the flayer succeeded *ho in outflanking his opponent. Perhaps that was the origination of war games the seventeenth development military Christopher Heikhmamn the "King's Game." as chess century came of to [22] it was in games reflecting the like a Eut . new age. In 1644, 01m developed a war chess called It is said to have been highly an aid in military training. regarded Since then various kinds of military chess were invented in many different countries. In 1824, a war came was played before General Vcn Muffling, the Chief cf General Staff of players rather ecldly Prussia. He first, but at expanded en the map and move by move the out their had as received the operations combatants worked plans, the old general's face lit up and at last he broke cut with enthusiasm: "It's not a game at all, training the fcr it's war. I shall recommend it enthusiastically to the whole army £2) ." This was the beginning of the war game as a serious military tursuit which was to spread to alacst every country with military pretensions. But what is the usefulness of war games? but the It is net any information from them, to combatants of tactical and present they test acquired strategical knowledge. "The only difference from actual war is the absence of danger, of fatigue, of responsibility and of the friction involved in maintaining discipline and these factors 1S37) all important in war." said Wilkinson are (1653 to who was a British military reformer. , When playing called so free regarding the effects of combat games, assessments war decisions other and were subjectively by a control team of experienced military made officers. So different a high variance of the results was expected decision makers were used. And because it takes a long time tc develop and to play a single game, it is not feasible if a mechanism for analyzing a broad spectrum of system alternatives in responsive manner to meet a planning cycle requirements B. SIHOIATICN War games can te simulated also by using To this kind of model, the military process is studied develop and computers^ microscopically activities, which decomposed are into basic events and to be ordered in a logical seguence Befpre being able to compute, the computer and programmed . must with data and parameters such as firing rate, be fed kill probabilities, etc. When those the start key is pressed, events and activities of the different combat process are essentially followed in a specified decisions based upon predetermined rules. seguence The final outcome will be trinted cut at the end of simulation. 10 Baking processes ccmbat Since protabilistic events distributions number large a activities, and probability reguire contain simulation many for of models of the input variables and generate the probability distributions fcr the output variables results. or statistical If saapling techniques involving the generation of pseudo random numbers employed, are simulation the called is Monte a Monte Carlo simulation models are simulation. Carlo employed in military plaining studies. Hcnte Carlo simulations tend to be games war tut still are much more abstract concrete more than than, for exanple, the Lanchester- type combat models discussed below. instance, two Marine Corps colonels fighting a war game For over Cuba are a much subtler model of real campaign than a trying tc simulate the same thing. This is because computer the human brain can still perform a much processes than the variety wider cannot it qualitative make of elaborate electronic computer. most computer can handle much more data with accuracy and but a judgements A speed, or deal with intangible factors such as leadership and morale. C. ANALI1IC MCDEIS Like siirulaticns, To develop a model involvement. study the events models also have no player analytic activities. and mathematically. Finally then He one integrates of the we first process. By describe them these events and assumed mathematical consistent mathematical activity descriptions into an overall structure kind, and decompose it into its basic process ccmbat this of operations, solutions can be obtained which will indicate the relationship between independent and dependent variables of combat effectiveness. When such a 11 relationship can be developed, obviously it sensitivity analysis interpreting contained since can not mathematical techniques, but ottairable. often an conduct increased ease the of in combat dynamics are the examined readily solutions analytical provides and results, the in simplifies Sometimes equations. obtained by appropriate be approximations numerical are provides substantial reductions in This cost and time for the conduct of military analysis. Analytic models can either be In the deterministic case, prctabilistic. deterministic or the same of set values always produce the same set of output results; input while in the probabilistic case, seme of the input variables have probability distributions and produce different results Eeplications ever the cutjut variables. desirable are in prcbabalistic case. Combat is a" measurement. process that does not readily lend itself to operational The effectiveness combat of systems are often times edicted by military personnel. experimentally not are Thus they should not be verified. used as an evaluation mechanism to provide predictions estimate accurate, effectiveness combat They should decision lakers. purposes of used be They only point for use by analysis for as to have a better understaning of the system so number dynamics. For this purpose, a variations variables of the model large paranetric of required. is As a consequence, for such parametric studies analytic models are preferred tc simulations and war games. developed Among many analytical models States, BAIFRAM Methodology) China Force (Balanced Requirement is the one which has been used by military personnel as a ccntingency flans. BALFRAM will be chapter. 12 the in tool tc discussed United Analysis Republic analyze in the of their next BALFRAM Ill- A. GENEE4L EESCEIETICN EALFBAM (Balanced Force Requirement Analysis Methodology) is a computer war gaming model. It consists of scae FORTBAfl 10,000 primarily statements. The capabilities force requirements and can be analyzed and the effectiveness of force levels and fcrce mixes can be evaluated. The entire program subprograms, namely, the NODH program and two of The NODH program, when provided with the DCSF program. scenario is an analytical bookkeeping device which provides a framework within which problems of consists program tasic the abstracted from a hypothesized campaign geccraphy environment in terms of nodal points and lines access of between nodes, will compute the matrices of minimum distance between nodes and of next distance. characteristics force on the of path minimum program, when provided with inputs of DCSF Ihe nodes and contingency logic (tactical decision rules) will move units over the scenario geography, enabling them computing the to according engage outcome of the to the engagements scenario according and to appropriate fight laws designated by the user. The program is It is very flexible and may be used to user oriented. examine In a wide range of military applications. order to use BALFRAH, scenario geography must first be abstracted into nodes representing geographical at which tattles may congruous EALFRAM terms. take place and locations translated into Then fcrce characteristics and the 13 decision tactical rules must be formulated and described through tte use of BALFRAM descriptors. EAIFBAM is not a In certain complete war gaming model until descriptors en fcrce characteristics and sense, a set of contingency legic has been prepared and assembled to compose a scenario. B. SXSTEH AEEIICAIION !• System Organization A in figure simplified system organization cf EALFRAM is shown 1 as a tree type diagram. 14 SYSTEM SUPERVISOR W STOP J Descriptor Processor Surface Supervisor A T Surface Matrix Output BATTLE SUPERVISOR Next Event Determinator Figure 1 - Survivor Contingency Determinator Plan Processor SYSTEM ORGANIZATION 15 It scenario shows the steps in control ficm time the inputs and execution submitted are of a the to ccmputaticn center until the time final outputs are produced alcng functional supervisors each lines. charge in idea The of a of specific having step three in the execution piccess is veil conceived. 2- $1§ ten Operation In order use to scenario and translate it First, scenario BALFRAM, the user must develop a into geography NODH serves EALFRAM inputs. is submitted to the computation center in natrix fcrm as inputs from congruous to NODH program. Outputs geographical inputs to DCSF program. as Then ether ircuts (force characteristics, movement logic and tattle logic) are submitted to the computation center as Outputs from DCSF program are the results of the scenario with respect to specific sets inputs tc DCSE program. final of parameters. BALFRAM maintains no and NODH tasis. DCS* files data except base the are on a scenario to scenario which All outputs from both NODH and program DCSF are returned to the users. C. MATHEMATICAL EiCKGROOND '• Lancaster's Classic Combat Models The iiathematical foundation of BALEBAM rests en the extention and enrichnent of Lanchester's classic combat 1946) between two was an English opposing forces. aeronautical 16 theory Lanchester engineer, who of (1668 to believed that "one cf the great questions at the root of all strategy is that cf concentration; resources cf a concentration the of field of operations made a strength of naval or military, at cne point in the whether force, whole belligerent on a single purpose or object, and concurrently the concentration of the main his the To prove this $5} ." point, Lanchester simplified jnathematical analysis of the relation of oppcsing fcrces in battle, He wrote the equations" in (1) T9"14 famous "Lanchester under the assumption that: Iwo oppcsing forces each capable of inflicting casualties en the ether are engaged in combat, (2) Each unit engaged in battle are within the firing range cf all pther enemy units, and (3) Ihen Ecth sides use aimed fire. the combat between the two oppcsing forces was modelled by dx/dt =-ay w£th x (t = 0) =xo (3. 1) dy/dt =-bx with y(t=0)=y where x (t) =the numbers of X force at time y(t)=the numbers of a=atrition rate of b=attriticn rate of and t= Y force at time t t force X Y force denotes the time at which the battle begins. 17 from Lanchester deduced his classical square (3.1) lav t(x 2 2 =a(£-y2 (t)) -x (t)) which implies that if one (3.2) side committed more forces to tattle at the very beginning, his casualties will be reduced significantly. He by histcry pf force level, x and y(t) (t) were given 18 Cosh/abt-y /a/bSinh/abt i (t) =2o 0( (3.3) Cosh/abt Xo/FTaSinh/abt J (t) ~lo If assumptipn is changed to be area fire, the (3) model beccmes dx/dt =-axy with x(t = 0)=x o (3.4; dj/dt =-bxy with y(t=0)=yo The state eguation is given by t (Xo-x) =a the flod (3.5) (y.-y) time histcry of, for example, the X force level is given by 1 I bx ~ } asio <?xp[(ay ~bx for bx °n yi ay ° T) tJ J x(t) =- ———-— X I -r b xn for t 18 bx 0=d y (3-6) Cifferential Fig ht 2. L§ws. Fight laws define the number of surviving components of tasic differential fight law. They are linear BAIFBAM contains two as a function of time. side each a square the laws: law and the modified version of the Lanchester's eguations. a. Sguare Law The sguare law has the form dx/dt=-ay-c with x(t = 0)=x C3.7; cy/dt=-bx-e with y(t=Q)=y„ where the x(t), v (t) excgenccs incremental , a and b are defined as in parameter firepower capability of y that (3.1) and c represents is the to inflict attrition Or virtue of the exogenous firepower available to y, and x e by is defined sinilarly. The state equation is given by (fbxo* e//b)-(/bx+e//b) 2 = ( /ay. + c//aj*- (/iy + c//i) 2 Analytic sulutions are always available and (3.3) are given by j (t) = ( (/"cx +e//b) Cosh/a~bt-(/ay e +c//a) Sinh /abt)//~b-e C3.9) I <t) = ( ( /ay +c//a) Cosh 19 /a~bt- (/"Exo +e//b) Sinh /ab t)//a-c Ihe differential fight laws contain the implicit assumption area) that fire wfcich exogenous fire is aimed (rather than is subject to discussion. this linear Lav t. The linear law has the form cx/dt=-axy-c with x (t=0) =x (3.10J dy/dt=-bxy-e with where x,j f t,c,and attrition of surviving side components is dependent both of assumption cf area fire: receiving =y are as previously defined. e each y (t=0) sides. the more distributed uniformly on The rate of number the of This is because the components in area an fire, the more casualties incurred; and the mere components firing intc the area, the mere casualties thay will inflict. Ihe state equation is given by t (x -x) =a (y -y) for tc=ae analytic suluticns In case e=c=0, (3.10) (3.11,) for equation eiist. (a) x bx« =ay (t)=x./(1+btXo ) (3. 12) J(t)=y./{1+atya (t) ) bx *ay 20 i <t) = /a7br h/(exp ( /abkt)-r) (3.13) j <t)=/E7a€xp (/abkt)/(exp (/abkt)- fcfaere r=bxa/ay and k= /a7"by Vt/ax o When c * or both c, e * 0, no analytic e * 0, r) solution exists and it is necessary to perform solution of eguaticn Kcte: ether numerical a (3.10). Shen one side uses aimed fire area fire, one side suffers attrition at uses proportional to only the number of firers, while at a proportional rate the to called "nixed the a the the rate other product of the number of firers and the number of targets. The resulting is anc combat law BALFRAM can also accommodate law." this situation. D. EALFBAS IICCISS Although EALFRAM war gaming model, language. to form a it is essentially a order has been prepared and (descriptors) hypothesized scenario. consist of three categories. force high a computer conpiler can be regarded as model only when a ccnplete It set of inputs usually referred to as is characteristics, decision rules) . Figure They are essential imputs scenario contingency and 2 The logic shows the entire EiLFRAM 21 assembled geography, (tactical process. / Q \ u / H >\ UI h O thfe o. < 2 2 O Q I lil / H < \ ( < £ ^ > \ a 2 / \ \ to Ui M u \\ „E // \ "2 1 .-• o J i as Y I i < I 4 Q uj t— ui 2 O — 30tt ^ < 2 2 a O O 2 UJ . CC z> h < H CM a' M H 22 Scenario Geography 1 • abstracted from a projected campaign envircnment into nodes and lines of access between geography Scenario represent ncdes The then. specific locations. even represented fcy a defined areas or They can be located on land, on sea or air" tie "in is scenario the as geographic reguires. ncde can range in size from reguired that an infantry sguad to that reguired for a battle between fcr corps or field armies. associated their and The lines of distances access sides must move. between represent cottmunicaticn ever which the forces and nodes Given a can be logistics both of The network as detailed as reguired by the scenario. netwerk cf nodes, the NODH program will compute distance network. Geographical swamps and of The movement of combat units during the shcrtest risers aodes lines the course of battle most also follow these routes. of The area route between irregularities also can be twe any nodes such as in the the mcuntains, input to the network for assessing the effects of force deployment and mobility on battle cutccne. Fcrce Charac teri stics 2- characteristics refer to the number, type and nature of units each side can commit to battle, the nobility of rate at which each unit can move over the scenario Perce geography, the combat effectiveness in terms of the atility to inflict casualties on the enemy units, and the breakpoint in term cf the number of casualties each unit before being defeated. cne. It retaining can its consist cwn can sustain In BALFRAM, the unit is a conceptual of several units characteristics or 23 a with unit each fraction of a unit which possess the same attrition breakpoint the as mobility capability, and Units of different sizes can be unit. input into EiLJBAM as force equivalents of the standard unit which is designated They can be icpct into the same prirciple applies. either as single ship a As to ships and airplanes, the user. i>y EilffiAM or as naval (or airplane) (cr air units. force) Ccntinqency Logic 3- Ccntingency logic refers to the sequential decision rule tactical the way forces are to be employed during or example typical initial the course cf deployment would be what proportion of ground forces are to deployed be piopcrticn campaign. front to combat line are forces cf A of positions and what to be held as reserves, and the relative pcsition cr locations the front line units and reserves are to occupy. An example of fcrce utilization would be what proportion of tactical air csed close for combat. air the fcrce are to be aix support and what proportion for air to initial After deployment, may it become necessarj to change deployment policy or mission allocations These events specific contingent upon some activities contingent might which provide the occur. operational priorities and novement logic which govern the way in a conducts unit which its operation as the scetario progresses. The general form cf contingent logic input statement is some condition true, the specified units do the In developing logical operations, the users are following." cautioned is "if to be consistent. Inconsistencies may cause one logical step tc be negated by another. types of contingency logic inputs, namely, battle logic inputs and movement logic inputs: They There are two can be summarized as follows. 24 a. Eattle Logic Eescrite rattles are tc forces involved and nodes at which cccur. attrition Specify factors and criteria for tattle teminaticn orders Eermit of battle of several units tc be merged. Specify allocation ef fectivecess and of supporting weapons. Ercportionally assign and redistribute fcrcss. Eescrite logistic pipelines and interdiction effects. Vary parameters such as order of battle, firepower, ncrility. ^Ffly principles of concentration. t. ecvement Logic Cove units from node to node contingent on arrival events. Eelocate units contingent on defeat events or at specified tine. Eermit withdrawal if force ratic is unfavorable. 25 Cause one force to chase another. Establish a sequential link up cf forces. Eedeploy units after battle is wen. PEBA Trace ffovement of tattle area) (forward of the . Inputs and Outputs **• HCEH Program a. represent Inputs to this program abstracted geography in matrix form representicg direct distance between nodes. (1) edge the scenario with elements The outputs are matrix cf direct distance between node pairs, of minimuu distance between node pairs, and (2) matrix matrix (3) of next nodes ct the path of minimum distance. LCSF Program t. Eeside the preprocessed scenario geography inputs from KCEH program, additional inputs must be prcvided to construct a EALPRAM scenario. These include force levels, motility, indices cf combat effectiveness, defeat and attrition rates. also be developed and logic inputs. the "end cf criteria, The initial concept cf operation must formulated as battle and aovement The outputs include the "battle histcry" and campaign summary". The former provides a chronological record of the scenario showing the location objectives of forces, location and status of anc next 26 tattles, step. and size the surviving latter, The casualties forces involved at each tattle of resulting both on sides, from exogenous fire, and the duration An output of of the battle. forces sensitivity analysis results can also te cbtained at the option of the user. 5 . Capabilities forces heterogeneous as homogeneous, EAIPEAM provides an aggregated approach to the simulation of Treating conventicnal conflicts between two opposing forces composed of ground, air and naval units. It can be used to assess the capabilities cf the component ground, air and naval elements in their coordinated support of a military operation cr to ccapare soundness of tactical decisions by examining the tfce effects cf alternative courses of action upon the outccme of a conflict. theoretic sclutions tc sensitivity perform program can also generate game software The of force analysis to guantify between input and output, thus analyzing tradeoffs the allocation problems providing between a and relationship framework for forces and component deriving optimal force level objectives. simulation model for a It can be used to model combat at the unified high command. theater level. There is no explicit statement on the level however, the it can handle; cf force cr number of units EALIBkH is primarily inputs on force primary a and contingency characteristics logic in the form cf data cards usually do not exceed 300 per scenario. & analysis typical E2LFRAM example can of the perform kind would of te analysis superiority, parity and I-force superiority en the variations of Y- force 27 levels sensitivity the X — force as shown based in the following figure. 28 o r-i P •P CO U O > rt CO 4-< O > 3 o •H CO -H 4J Q 1.0 1.5 2.C 2.5 Y-Force Level Multiplier Figure 3 - SENSITIVITY ANALYSIS 29 3.0 i. 0TILI2A1ICN OF EALFRAM IS THE REPUELIC CI CHINA EA1FBAM was first introduced to the Republic of China in ISVO's. early Since ccnjuncticn kith manual war games well as tattle simultaneously has Because games. used been computer in war manual war games can cnly represent the as reality field BALFRAM then, using seme to extent, same the playing scenario them hopefully would elininate seme of the weaknesses of both. A lajor difficulty encountered in using EALFRAM was the determination of attrition rate coefficients types of (e.g. Army) different for EA1FRAM treats units from the same Service units. as being homogeneous regardles cf their weapon characteristics and units frcm different Services (e.g. Army and Air fcrce) units wherein EA1JRAM in supposedly brought normalizing The use of notional as heterogeneous forces. to units from the same Service are footing equal by pooling resources by the standard unit is their estimation in the right direction toward the and a step attrition of However, it falls short cf achieving its rate coefficients. goal because cf the definition of homogeneity. An example may help tc illustrate this point. For example, if an infantry standard unit. urit division is used divisions notional units. capability cf is and This means an the given a value of 1.0, i.e. one notional and An armcred division is found to be eguivalent infantry as armored tc two given a value cf 2.0, i.e. two that the casualty inflicting division is twice as that cf an infantry division. However, in terms of other contr itutiens, an division may be worth more than two infantry armored divisions. The use of armored 30 divisions fcr the exploitation victories is a possible case. The defeat of France by Germany in the Second Wcrld War is an example cf such an instance. On the other hand, an infantry tattle of division may te equivalent to armored an division under certain ccmfcat sitcations. Due to its aggregate respect this to BALFRAM's nature, approach with problem was to leave the determination of attrition-rate coefficients suggest that this wuold user. the to increase seemed It to flexibility the and applicability of BAIFBAM in that the user was free to choose the attrition coefficients and simulate many types of Military operationswhich In view of the objectives admirably. by command of at was intended to be used, it had achieved its BilFEAM relatively level However, for more effective use at lcwer levels of command, BAIFRAM can be improved employing existing other methodologies estimate to attrition-rate coefficients for different types of units and under different sets of circumstances. ICE was the calculation of encountered (index cf combat effectiveness). Another status the difficulty or etc.) , the was large a reflects ICE condition sustaining experience, of a unit capability, of ICE is primarily a matter of determination judgement and to problem effectiveness ccmfcat (training, motivation, Eecause extent a subjective one. The further complicated when information regarding the enemy troops concerning these factors is incomplete can net be relied upon. and Under these circumstances, the results cf the simulation could easily be tempered to please would be Th£ net result of all this cne*s superior. tantamount to the negation of the purpose of the entire effort. One way cf attacking this problem would methodology for estimating the 31 ICE^ fce to develop a of various types of units with respect to quantitative and which have a qualitative factors direct bearing on the effectiveness of to carry en combat. It is true that (measure MCE a unit a of effectiveness) for such factors is hard to decide on because it is difficult to devise a acceptable ffesaurement of these factors iihich will provide a reasonable approximation to the BAIIBAM took ICE. a passive approach to this problem as the case cf attrition- rate coefficients. in An active approach would be to develop a methodology and incorporate it in the program for the estimation of ICE of various types cf units. This is nest desirable on the part of the users. EALEEAM, In ICE is a Multiplicative factor acting upon the attrition-rate coefficient. because the attrition— rate It may net coefficient diainishing marginal return with respect to ICE. A be should realistic have increasing a the curve such as an exponential might be mere suitable to descrite this relationship. 32 ESTIMATION OF ATTRITION-RATE COEFFICIENTS IV. In utilization the essential fart is of Lanchester — type model, the a determine to numerical values for attrition— rate coefficients frcm weapon system performance data. Two significant developments in this field appeared during the 1960's, namely, (1) the development of Bethcdolccy fcr th€ prediction of Lanchester attrition — rate coefficients frcm weapon system performance data by S. the development of Bender (3) and C. Earfoot (l) and (2) methcdclcgy fcr the estimation of such coefficients from will Clark (6} We Monte Carlo sinulaticn output by G. discuss these methodologies in this chapter. Lanchester . A. MABKCV EEEENDEKT FIRE The purpose of firing a gun is to destroy a target. the distribution hit is interest of in weapon So system analysis work. If we could assume that the bit probability the is same for all rounds and each of them is independent then of the others, the Binomial distribution would be suitable for the hit distribution. Let E be a random variable denotes the number of rounds ttrcet with each round has hit probability p, then for firing n rounds, the probability that at least one round hit the hit the target is given by Pr< B 1 1 ) = 1-Pr( H = ) 33 = 1-( 1-p f (4.1; many circumstances this model is inadequate because fire can be adjusted and aim pcints for rounds are net statistically independent. Eut in Bender assumed that the firing process is a Karkov process that is the weapon system performance depending only en the cutccne of the last round fired. = Erob ( hit on first round g = Erct ( iiiss p then eguaticn Pr E ( > 1 paper This = ) on round|miss en -previous rcund (4.1) ) Define ) may be revised as 1-( 1-p ) <f~', for n > (4-2) 1 deal with this type of firing process will and called it Markov-dependent fire. E. ECNDEE»£ HETHCE Becall the Lanchester- type equations fcr combat between two homogeneous forces dx/dt = -ay with x (t=0) =x<> (4-3) dy/dt =-bj with In this model, y (t=0) a =y and b are called the Lanchester represents the a attrtion-rat e coefficient. Fcr example, rate at which a single Y weapon system destroys X targets. It has the diuensicn of (X casualties) /( (number of X units) 3± • (unit time)) Ecnder has defined A, random variable, as the rate a of destruction cf X target and its value is given by A = where to (4-4) l/l is a randcm variable denotes the time for T kill an destruction target. X would "average" The rate a Y firer target of be the expected value of A, denoted ty a and i = E< A = E( ) However, in a 1/T (A-S) ) paper published in 1969, out the average attrition of equation Earfoot (4.5) pointed is inadequate as well as net ffathematically tractable. He suggested defining the average kill rate as the reciprocal of the expected time to destroy a = a target. 1/E(T) (4-6) The reason is that in distribution function Bender's represents model the the fractions or targets killed fcr which each rate is used, then the of the rates should be used. probability harmonic If the probability distribution function represents the fraction of the time that each is used, arithmatic the then mean mean of a rate set of attrition rates will be appropriate. Eguaticn a, (4.6) agrees with the intuitive definition of the average attrition rate is the ratio cf the number targets killed in a number of battles to the time large interval over which the targets were killed. are killed and i-1 and i t of If n targets is the time interval between which targets were killed, then 35 n n S*' In a later paper E( 1) 1 1 n i n . (4.7) l (£/>! (u) , Bender suggested a way to calculate as B(T)=t*+trth *<th *tf)/PK +({tm+t f )/p). (<1-u)/P K +u-p, ) C4.8) where time tc acguire targets td = ti time tc fire the first round = time to fire tj,= a round, given the .preceeding rcund was hit time tc fire a round, given the preceeding rcund was tm = a a aiss projectile flight time tj = p, first round hit probability = the conditional probability of kill, given a hit pK = conditional probability of a hit, preceeding rcund fired hit the target, and given the conditional probability of a preceeding rcund fired missed the target. given the u p = = Prof. Tajjlcr stated that if 36 the hit, following assumptions can te justified, | (D t = (2) ti =th=t m =t = 1/v, where (3) p = C u = p, = ph , where denotes the rate of fire, v ph denotes the single shot hit probability, and U- <<*) C Then equation E(T) where C. (4.8) can be simplified as follows l/(v.p) = (4 g) # p = ph pK CLARK'S MCEIL G. Clark [6) probability, has added another factor, target Lanchester- type into combat development cf the COMAN (COMbat ANalysis) The model is funcanental a ccnbinaticns. models in his model. concept used in constructing the CCMAN rate kill acquisition These for specified rates kiJl firer/ target —type are estimated from simulation data and they provide insights as to the relative effectiveness of various weapon types without resorting to nuaerous siiulation runs. Clark points out that for modelling purposes a target is acguired by a firer when fire can be directed towards the target's position. Acquisition can be accomplished visually detecting the target so that its position is 37 by known then directing fire at the position where the target is and actually located. Ecwever used. are This is the case when direct fire weapons indirect fire weapons such as artillery can be fired ever hill masses that obstruct a line of sight between the firer and acquired target position. So knowledge of the exact target position can be acquired by firing at area for targets to be located until fire happens to likely be directed at an actual target position by chance. effects The of using the prctability that are introduced by acquisition target target a unacquired is as a parameter. Define = p the probability unacquired by an individual g = that a specific X target is firer, and Y the corresponding probability for a I target X firer ccntinaticn, then dx/dt = -a5 (1-p x ) dy/dt=-bx (1-q y ) where a>C ,b>C,0$p< 1 and 0^q<1. It can be seen that if equation (4. 10) targets are readily becomes dx/dt=-ay dy/dt = -t* which is the familiar Lancheste^s square law. 38 acquired, When individual targets become increasingly difficult to acguire that is the probability of a target being unacguired assume values close to one, the combat situation represented by eguaticn by (4.10) Lanchester is equivalent to the situation envisioned when he formulated relationship is shown by expressing function cf p and g the linear eguaticn law. This (4. 10) as a and expanded in a Taylor series about the points p=1#g=1 dx/dt=f (p) =-ay(1-p x ) =-ay(f (1)+f« (1) <p-1)+f"0)/2!) (p-1) *.--) =-ay <-x (p-1) ) =-axy (1-p) similarly (4.12) dy/dt=-fciy (1-g) 39 1- £ATA BE^UIREMENT 121 COEFFICIENT ESTIMATION running simulation or war gaming model, a wide range cf factors must be considered. Though simplifying parity assumptions can be made regarding common factors, Id a guantified values must be provided to describe those factors for which differences exist between the two opposing forces. This gives rise to the problem of input data requirements, which is crucial tc the estimation of attrition coefficients for different types of under units different combat situations. The data required for input to simulaticn or war models for the evaluation military cf problems gaming can be classified into four categories; A. HISTCEICil EAT* Since tte formulation of the linear and square laws by Lanchester in 1914, substantial amount of research work has the area of mathematical modeling of ccmfcat. been done in Efforts in this respect were primarily centered on the enrichment of lanchester's theory of ccmtat, which has teen validated to be adequate to represent the dynamic process of classic combat by: extention and Helmtcld using data on twenty seven battles which occurred in the United States between 1759 to 1945 (also (1) several subseguent studies (1Q) (14) 40 ) , using Engel (2) Seccmd Wcrld Imo the Jima engagement data cf the fcar, iillard using data of the land battles of the years (3) 1SC8, and 1618 to Weiss using Anerican Civil War data. (4) However, parameters historical Lanchester— type using cf data the model must be very combat - estimate to careful because: data from different sources usually are Historical (1) net consistent, It (2) difficult is to decide hew tc count reserves, reinf orcenent and saneuvering elements, Casualties may in some instances have been estimated (3) by subtracting "stenjgth after battle" from battle". It is calculated value a "strenth from before two inaccurate numbers hence large errors may by expecte.d, and Ihe duration of engagement is unreliable, (4) usually it is estimated by the author. these along Furthermcre, discussed th€ uselessness future types ccmbat of developed outcome singulation same historical of predictions. and war Helmbold lines gaming for data (11) has aaking Meanwhile, different have models since the end of the Second World War. been While war gaming technigues are being energetically extended tc fields where in little or no previous experience with the technigues sophisticated efforts military have fprm exists, essentially no parallel been made to compile and analyze data on past engagements. This may 41 be attributed to the following reasons: Shen nations were at war with each ether, they would (1) engulfed so te in it that they could not afford to divert their effort tc data collection, Ihough much can be and have been learned by studying (2) the successes and failures in past wars, the chance to fight had a a- nations few have war in the same general situation second tise, and Even (3) actual if cemtat data available, it was doubtful whether significart they would be impact military on doctrine operations have been developed. be of any value to military analysis because of the rapid advances in science and technology. New weapons great wars were past on fought concept the have of The kind of war which will weapons will equipped forces ty and which these with definitely be different from past wars. view In United usefulness This has sinulaticn validation the resides only in led of models; it has only in predicting the outcome of future wars. the to war and value of historical data the above, the cf present gaming to trend of using ccnputer generate data for military analysis E. L0GIS1ICS EATA particularly susceptitle to guantificaticn. Appropriate data on practically every thing that can be procured, transported, used and consumed exist Logistics imput data in seme tangible form. food, gasoline, etc., are Items such as equipment, ammunition, fall into this category. 42 Data such as distances, means cf transportation, volume and weight to be transported, tine in transit, etc., are readily available can be used as and inputs. Furthermore, logistics reguirement for each type of units can be established and sithin placed preparaticn reasonable bounds. This processing for use. and is cct kncwn hew these logistic facilitate will However, currently it factors influence combat capability. In particular, there is no commenly accepted (or used) prccedure for modifying combat capability due to logistic shortfalls. C. UEAPCK ZilA Weapon data include range, rate or fire, lethality, etc. Using characteristics, different types of weapens or these weapcn systems can be converted into input data by firepower sccre information ef f evaluated. weapons, of problem the in characteristics one's of conventional Hcwever, and the Thus cover tactical of is complicated. In not the weapons can be compared and case the protective significant factors. cases, is forces. own areas significant, but troop density time, most In obtained based en known data and are characteristics ectiveress methods. weapon enemy en Estimates available. weapon appropriate ether cr Net only are lethal the in or addition, nuclear area expesure at the also are contamination effects which deny the area to both friendly and enemy use must also be considered. Since tactical nuclear weapens have never been used in the past, their psychological effects in actual ccmbat situation with respect to the fight of trccps involved is not known. morale 43 will to Presumably estimates can be made, but how close these estimates real effects is not ascertainable. and can be tc the QOALIIAIITJE DA1A D. The term qualitative data refers to factors such as discipline, motivation, courage, morale, will to fight, etc. By prcfessioral military judgement, these factors can be quantified values to represent different levels or degrees using scaling method. Qualitatve standards such as outstanding, superior, good, etc., can be assigned numerical assigned values, shich can be degrade used expected tte multipliers as performance upgrade to But factors of a unit. such as the effects of shock and fatigue en or personnel; the resourcefulness, initiative, and leadership cf the relative opposing ccananders; the results of communication or command control failure not are susceptible to guantif ication in that each war is urique in its own right. impact The of factcrs on *ar outcome is immense and very difficult, these if net impossible, to estimate. Probably BALfBAM is the enly computer *ar gaming model which allows for the consideration Though cf some cf these factors. factcrs in a unfortunately, inclusion of these simulation or war gaming is essential tc the representation adeguate the of ultimate their battlefield impact is reality, beyond the imagination cf the human mind. The prctlem cf data complex and complicated. have been established requirements for war gaming is Except where definite objectives and quantitative data exist in some tangible fcrm, the problem appears to be magnitude and deserves a separate treatment. 44 cf considerable AGGREGATION OF FORCES VI. It was feinted out in the preceeding validation cf chapter that Lanchester* s theory of combat had led enccurageuent of development of simulation and war the to the gaming models based en the enrichment and extension of that theory. Such develotnent was clearly in response to the growing need models such for field the in of military analysis and decision caking. New the problem which remains to be is the acgregaticn of solved forces when the forces involved on either or tcth sides are composed of more than one unit above division level. In the case of homogeneous forces are (i.e. of and forces composed of identical units), forces is not a problem; however, the in aggregation the case of nenhomogeneecs forces, the problem is complicated and lather difficult. Several methodologies used to deal with the problem in All of these problem a the been propesed and nonhemogeneous represent methodologies direction, but none provide have case. steps in the right satisfactory solution tc the because each methodology has its strengths and none lacks weaknesses. There is nc commonly accepted methodology in existence. A. N0TICSA1 ONIT The concept cf notional unit (16) is one cf the several methodologies being used to address the problem of aggregaticn cf forces. The major strength cf the approach existing that it takes into account all the resources (personnel, Though there are ether eguipment and weapons) of a unit. is 45 to t€ considered, factors the approach appears to provide a reasonable approximation to the contributing factors concerned. assigned tc cf far unit is what values should be of a at approximation reasonable a unit? Since the effectiveness of a depends en the type of target against which it is contribution the major as different weapons or pieces of equipment in tfce arrive capability as capability the to Eowever, the problem is: tc order problem of a the weapon used and of equipment depends en the piece a tc ervironment in which it is employed, the cencept of notional units must re applied with special care and emphasis en the type of cemfcat in which a unit is to engaged be and the envirenment in which the unit will be fighting. E. FIBEECWEE SCORES The firepewer scores approach appeals tc (see Stockfisch and (17) further fcr a further references.) discussion relative firepower The to be based cc actual or experimental data ccrtribution the of certain a are supposed with respect to elusive. seems However, the problem standard weaken. firepower of scores tc be assigned to each type of weapons can military because cf its simplicity and ease of application. analysts scores most type a How of weapen to a Bilitary cperation be isolated and singled cut from the many weapen involved? systems Given this firepower sccres should be assigned to can a be weapon dene, what considering the different types of targets against which a weapen nay be used? tank Fcr example, consider the firepower scores for gun and a M14 rifle. a 90MM If the M14 rifle is chesen as a standard weapen and assigned a value of 1 , what value should be assigned to the 90MM tank gun? If lethality, rate of fire, motility, protection, type of targets are seme basis cf cemparison considered, exists and a value for the 90MM 46 tank gun may be derived. Military experience expertise and familiarity with the weapons may provide a professional judgement as to the reasonableness of the estimation, but and this nay C. as far as one can go. fce THE QSE CF SATELLITE MODELS increased In view cf the techrigues gaming use simulation of CCHANEX combat) model (CCMiN cf and battalion sized used be resolution of COMANEX is a satellite conjunction with CABMONETTE, in simulation combat Carlo units in crcund lower or Monte (a Data model. a high relating to characteristics, combat environment, missicn, etc., weapons particular fcr a CARMONETTE Extended) are examples of such models. tc problem forces lies in the use of satellite mcdels. cf sinulaticn war in the analysis of military problens, it appears tc the authors that the solution to the aggregation and mix performs CAEMCNETTE CAEMONETIE. fcrces opposing of a input are to prespecified number of replications cf the replication, time- sequenced casualty history identifying a the time at which the killer type. CCMAHEX which a It battle. outputs, then for each casualty occurred, the casualty type and output This massages the in turn, input to the is, data outputs and a set of Lanchester-t jpe parameters which represent, essentially, the kill rates fcr each firer/target combination in the battle. paraneters are then used in DBM (Division Battle Kodel) approach ground ccmbat assessments. The advantage of this The traditional over scores is developed that by methodology CAHMONETTE environment, including frcii and weapon the based unit reflect weapon firepower performance measures on variations in the ccmbat synergistic effects resulting the entlcyment of various combinations cf weapons. 47 D. THE APPLICATION CF BONDEB»S APPROACH IN EALFRAM Though EAIFRAC heterogeneous forces reality it is cnly has the as well capability homogeneous as homogeneous a handling of forces, in because model of the aggregated fashicn in which it handles hetercgeneous forces. Heterogeneous force in BALFRAM ccoponent different refers Services frco cf the armed forces, the Army, Aii Force and Navy. types units to Units namely, classified are the into grcund, air or naval according tc the Services to as which they telcng. Units which belong to Service same the treated as homogeneous though they may be equipped with are weapons cf different characteristics such as lethality, rate of range, fire, BALFRAM's definition of protection, etc. homogeneous and heterogeneous forces is different frcm is generally assumed ailitary analysts. for force to be firepower, researchers and in his mathematical For example, hcacgeneous identical operation most between two homogeneous forces, Prof. combat defined by Gordon units. a force force to of J. acdels Taylor composed force a Clark M. composed mobility, be what of defined hetercgeneous weapons with protection, and different detection characteristics in the general context of land combat. Bonder extended Lancheste^s original formulation cf the linear and scuare laws to incorporate target acquisition time, rate cf fire, and conditional kill probability in the calculation cf Lanchester-type attrition-rate coefficients. His definition of homogeneous and heterogeneous forces was also in the general context of land combat and agrees with these of Ercf. modeled the ccmbat J. Taylor process Gordon and in a M. detailed Clark. fashion Ecnder while Eonder initially used the EALFRAM in ar aggregated fashicn. arithmatic mean of the time required to destroy a target as the estimate of the Lanchester attrition-rate coefficient. 48 However, this untractable, approach turned out to be mathematically an explicit expression fcr the Lanchester and attritioE-rate coefficient could not be obtained. Later Barfoot suggested using the harmonic mean of the time required tc destroy a target as the estimate of the average attrition- rate coefficient. Bender modified his formulation acccrding tc Earfcct*s suggestion. Eoth EAIFEAM and Bonder's model are based' extention and enrichment of Lanchester's theory of on the cemtat. underlying difference in the definition of hemegeneens and heterogeneous forces led to different basic However, the in the two models. As a result, the application assumptions of Ecnder's challenge apprcach in BALPRAM presented than was originally anticipated. in the levels cf details covered in the two ccnplicated the problem. The beyond the scope of this paper. H9 a greater The differences models further amount of work involved is VII. The study of war will despite become never efforts striving the PINAL REMARKS of scientists and military analysts. inability operation science researchers, reason The main is Another reason is vast number of variables present in a ccmbat situation. These variables do decree, not Heights or recur fixed in relative of fashion, importance. amount, Therefore, although varicus kinds of wars have been fought in the and the of man to predict how an individual will react in stressful and cangerpus combat situation. the exact an most past Hill continue to be fought in the future, likely man's understanding of the process of never will war be adequate and complete. In man's guest for insight into the process of war, both formulation of combat models and techniques of mathematical simulation ard war gaming have been the been area extensively used. mathematical models, substantial interest has of maintained since theory mathematcial Lanchester coefficients, this paper published first his Among the various methods of warfare. suggested for the estimation of Lanchester considered — type attrition the Markov-dependent fire and Ecnder's and Clark's models in particular. area In In the of simulation and war gaming, sophisticated techniques Now war gaming and increasingly employed. sinulaticn have become standard practices in the analysis of this paper Of all the existing models, military problems. have been discussed EAIFBAM feasibility estimation EAIfBAM. of in applying some detail Bonder's and explored methodology for the the Lanchester attrition— rate coefficients in It was found that BA1PRAM considers theater— level cf 50 ccabat in a very aggregated fashion, while Ecnder s approach Thus, applies tc fire fights in a more detailed manner. 1 Benders methodology is not applicable directly aggregated constat model such as BALFRAM, very heterogeneous example, Hence, a conglomeration division pr apparently a a corps different as which in models an the of forces found in, for a homogeneous approach must force. be used to estimate attrition- rate coefficients in such Lanchester-type force- planning models. This problem is a state-of-the-art problem in cemtat modelling, beyond the scope of this thesis. 51 and further discussion is LIST CF FIGURES 1. System Organization 15 2. EAIFEAM Erocess..... 22 3. Sensitivity Analysis 29 52 LIST OF REFERENCES 1. Lanchester Attrition - Rate Coefficient: Scjne Comments on Seth Bonders Paper and a Suggested Alternate Method", Cpns . Bes i# v. 17, p. Earfoot, C. 888 tc 6S4, 2. the 13th November, v. Ecnder, S. An Overview of Land Battle Modeling in • Proceedings OS", Sympcsiua, 3. 1969. Bender, S. The •• Coefficient", Qpns. , 15, v. 1974. Attrition— Bate Lanchester Res. Operations Arm y s_. 73 to 88, p. The " U. 221 tc 232, p. .1967. 4. Bender, S. Qpns. 5. fles_., v. 18, Mean p. Lanchester Attrition Eate", 179 to 181, The oretic Game Chaiken, P. The " 1970. Analysis Military of Ope rations Des cription of the B attlefie ld Mod els to 43, Stanfprd Research Institute, flenlo p. 25 Park, Ca., 1966. 6. Tie Clark, G. Combat Analysis Model, Ph. D. Dissertation, Tie Ohio State University, 1968. 7. J. Opnsj 8. A. Engel, " Res_., v. Verification 2, p. of 163 to 171, Lanchester's Law", 1954. Simulation in Warx Business, and McGraw Hill Book Coopany, 120, V enture fiausrath, Politics, p. A 1 to 1971. 9. R. Helmtcld, "Lanchester Parameters for Some Battles of the Last Iwo Bundred Years," Combat Operations Research 53 lechnical Group, Operations, Inc., - CORG SP — 122, Febrcary 1961. 10. 11. Belmtcld,"Bistorical Data and Lanchester's Thecry of Combat, Eart II," Combat Operations Research Group, Technical Operations, Inc., COBG-SP-190, August 1964. B. Helmbcld, B. Some " Observations Lanchester's Theory for Prediction", 12, 12. 778 tc 781, p. Helnbold, B. EquaticDS", 12. B. 14. R. 14, Belmfccld, " Res,., v. 13, of Bes^., v. Lanchester's cf 857 to 859, p. 1965. Attrition Model", 1 624 to 635, p. Cjcns^ 1966. "Decision in Battle: Breakpoint Hypotheses Engagement and Modification 'Universal A Opns. Use 1964. A Op^ns. Belntold, Res., v. " the en Termination Data," B-772-PB, The RAND Corporation, June 1971. 15. F. Lanchester, "Aircraft in Warfare: The Dawn cf the fi. Fourth "Engineering 2138 tc 2148 Mahtematics, 422 98, Principle The Bc.V., Arm (see also p. IV, Vpl. 423 to (reprinted on p. (1914) 2148 to 2157) Newman J. Concentration, of of The iorld of (Ed.) Siacn , and Schuster, New Yprk, 1956). 16. E. Means, C. Phillips and Young, S. BALFBAE Oser Manual for the Staff of the Commander in Chie f Pacific, p. 1—1 to B-15, 1974. Park, Ca., 17. J. Stanford Research Institute, Menlo Stockfisch, Models, Data, and Bar: Study cf Conventional Forces, Prepared for United States p. 6 A tc Critique cf The 27, A Seport Air Force Project Band B- 1526-PB, Santa Monica, Ca., 1975. 18. J. Taylcr Solution of and G. Varible Brown, " Canonical Methods in the Coefficient 54 Lanchester — Type 19. Eguaticcs cf 44 tc 4S, 1976. J. Hodern Warfare", "Mathematical Taylor, Cpns. res., v. Models of Combat," L€cture Notes for 0A4654, Naval Postgraduate School, Ca., 20. J. 1 S 1 Bes_., 21. H. v. A. 1 22, p. Heiss, Hilscn, Lanchester- Type Equations for Opns. with Variable Coefficients", Solving " Rarfare Civil War", 22. Monterey, 5- laylcr, •Modern 24, p. " 756 to 770, 1974. Combat Model and Historical Data: Ihe OS OjanSj. ijar Res^., v. Gaming, p. 14, p. 759 to 790, 13 to 29, Ltd,Harircndsworth, Middlesex, England, 55 1966. Penguin Ecoks 1970. INITIAL DISTRIBUTION LIST No. Copies Defense Documentation Center Cameron Station Alexandria, Virginia 22314 2 Library (Code 0212) Naval Postgraduate School Monterey, California 93940 2 Professor J. G. Taylor (Code 55Tw) Dept of Operations Research & Admin Sciences Naval Postgraduate School Monterey, California 93940 1 Professor J. B. Tysver (Code 55Ty) Dept of Operations Research & Admin Sciences Naval Postgraduate School Monterey, California 93940 1 Chairman, Code 55 Dept of Operations Research Naval Postgraduate School Monterey, California 93940 2 & Admin Sciences Col Lee, Yin-Chu P.O. Box 7022-5 Taipei Taiwan Republic of China 1 Lt Col Pi, Yu-Kung P.O. Box 7022-5 Taipei Taiwan Republic of China 1 Gen T. S. Yen P.O. Box 7022-5 1 , , Taipei Taiwan Republic of China , Professor Chien Lih 2 T-amkang College of Arts and Sciences Tamsui Taipei Taiwan Republic of China , , 1 Mr. E. H. Means Stanford Research Institute 333 Ravenswood Avenue Menlo Park, California 94025 56 138095 The estimation of Lanchester attritionrate coefficients for an aggregated combat 363 3 3 8 Thesis U435 Lee The estimation of Lanchester attritionrate coefficients for an aggregated combat model 3 6 u ise .5 thesL435 The estimation of Lanchester attntion-r 3 2768 002 12016 4 DUDLEY KNOX LIBRARY