Title: Time-Based Tree Graphs for Stabilized Force Structure Representations

Transcription

1 Paper for the 8 th International Command & Control Researh & Tehnology Symposium Title: Time-Based Tree Graphs for Stabilized Fore Struture Representations Submitted by: Sam Chamberlain U.S. Army Researh Laboratory ATTN: AMSRL-CI-CT APG, MD Fax: 2934 / 9223 wildman@arl.army.mil and Chris Leeds Chief, Combat Servie Support MTOE Division U.S. Army Fore Management Support Ageny ATTN: MOFI-FMA-SA 9900 Belvoir Road, Suite 120 Fort Belvoir, VA

2 Time-Based Tree Graphs for Stabilized Fore Struture Representations Dr. Sam Chamberlain Mr. Chris Leeds U.S. Army Researh Laboratory Chief, Combat Servie Support MTOE Division ATTN: AMSRL-CI-CT U.S. Army Fore Management Support Ageny Aberdeen Proving Ground, MD ATTN: MOFI-FMA-SA Belvoir Road, Suite 120 Fort Belvoir, VA Abstrat A basi, and often key, omponent of any battlefield representation is the fores involved. Obtaining high-resolution fore struture data has always been a major task. This is true whether the representation is for simulated or atual operations. However, the problem extends far beyond simply obtaining a single fore struture snapshot. The real hallenge is maintaining the data, espeially when numerous other programs are reating and linking their data to the fore struture. This paper desribes an approah for maintaining onsisteny in a high-resolution database of Army units that is undergoing ontinual hange due to fore modifiation. The use of time-based tree graphs is proposed as a tehnique for providing stability and maximizing the retention of existing entities to minimize the effet to systems that use the data. In a network-entri ontext, an easily aessible repository alled the Army Organization Server (AOS) is under development that will ontain the evolving, default fore struture of the Army. 1. Introdution In June 1998, pursuant to a disussion with the Diretor of Researh and Strategi Planning, OASD(C3I), a study was onduted to answer the question: With all this great tehnology, why an t our system interoperate? 1 The resulting report 2 ited three voids that perpetuate the inability to integrate systems; they are: The lak of a ommon naming onvention, The absene of a entral theme for data integration, and No designated authoritative soures of information. The proposed solution has evolved into three omponents. 1 2 OASD(C3I): Offie of the Assistant Seretary of Defense for Command, Control, Communiations, and Intelligene. Conversation with Dr. Dave Alberts, Diretor, Researh & Strategi Planning, June Chamberlain, Sam; Default Operational Representations of Military Organizations, Army Researh Laboratory Tehnial Report: ARL-TR-2172; February 2000; see: 1

3 The first is the adoption of a ommon naming onvention, alled enterprise identifiers, or EIDs, to uniquely tag data aross the enterprise. 3 The seond is the assertion that the entral theme through whih all battle ommand proesses onverge is the fundamental onept of fore struture. Consequently, a formal representation of fore struture is required as a foundation to integrate other battle ommand onepts. Finally, it is the atual fore struture data, not its theory or a model, whih is required by the battle ommand system users. Historially, in the absene of data, those in need reate it themselves. Fore struture data is no different. As a result, there are several different soures of fore struture data available throughout the Army, in various forms and different levels of detail. Clearly, they are not synhronized and require extensive effort to keep urrent. A reommended solution is to have the offiial domain experts (those who design and develop fore struture as a profession) provide and maintain this data for the user. However, in doing so, the requirements of the user must be inluded in the design of a repository and the proesses that are used to maintain it. 2. Bakground Tree Graphs A onvenient, mathematial tool for desribing fore struture is graphs, or more speifially, graph theory. A ommon manifestation of a graph is an organization hart. Beause of the basi priniples of military ommand, military organizations are onveniently represented via hierarhial organization harts that desribe the aggregation and omposition of lusters of people and equipment. A graph is omposed of a set of nodes onneted by a set of links. In mathematial vernaular, the nodes are alled verties and the links are alled edges. An example of a graph is illustrated in Figure 1 where a graph alled G is omposed of a set of verties, V, where V = {A,B,C,D,E}, onneted by a set of edges, E, where E = { (A,B),(A,C),(A,D),(C,E),(C,F) }. There are many ways to onnet the nodes listed in V, and the struture provided by E is just one. A tree is a speial type of graph that is fully onneted (i.e., every node is linked to at least one other node) and there are no yles (i.e., when links are traversed, only one path exists between any two nodes). When these two riteria are met, the graph must be a tree. Org harts are trees. Normally, one node is seleted as the beginning, or top, of the tree and is named the root node. Figure 1 summarizes several tree graph terms and illustrates how they are easily exploited to denote hierarhial organization harts. 3 For explanations of this solution, see: Also, the following papers are available: Chamberlain, Sam, Implementation of an Enterprise Identifier Seed Server for Joint and Coalition System, Proeedings of the 7th International Command and Control Researh and Tehnology Symposium; Quebe City, Quebe, Canada; September 2002; and Chamberlain, Sam; An Enterprise Identifier Strategy for Global Naming Aross Arbitrary C4I Systems, Proeedings of the 6th International Command and Control Researh and Tehnology Symposium; US Naval Aademy, Annapolis, MD; June 2001; Presented 19 June

4 Graph G: B C D Child E Parent A F NODES (or verties): set V = { A, B, C, D, E, F } LINKS (edges): set E = { (A,B), (A,C), (A,D), (C,E), C,F) } GRAPH: olletion of verties and edges: G(V,E) A Tree struture is a onneted graph with no yles, i.e., every node has at least one link to another node and only one path exists between any two nodes. Via a link, a node an be a parent or a hild of another node. A node without a hild is alled a terminal or leaf node (e.g., the nodes at the bottom of the tree: B, D, E, and F) A node with hildren is a non-terminal or internal node (e.g., A and C); The root node is a speial internal node with no parent (e.g., A). Organization Charts are Trees (w/ boxes instead of irles) ( Often the name of the tree is inherited from the name of the root node - e.g., A ): B A C D B Figure 1: Tree Graph Definitions and Terms Terms like unit or struture are ambiguous in isolation; therefore, a more formal definition is required. Using the tree graph formalism, a node of a tree an be named an organization (e.g., node A in Figure 1 is alled Organization A.). The term assoiation an be used to refer to a link of a tree (e.g., the line onneting nodes A and C, denoted by (A,C), in Figure 1 an be alled assoiation AC, or more speifially, organization-assoiation AC ). An organization hart is a tree graph omposed of a set of nodes and a set of links, or in this new vernaular, a set of organizations and a set of organization-assoiations. For onveniene, the graph an be alled a unit. Thus, a unit (a graph) is omposed of organizations and organization-assoiations. The ation of moving from node to node along the links of a graph is alled traversing the graph and there are numerous, well-known algorithms for doing this. The links and nodes of a graph may inlude additional attributes to allow them to be filtered (i.e., seleted or deseleted) during the traversal proess. This allows different paths to be followed by applying parameter onstraints during the traversal proess, as is illustrated in Figure 2. The left-most graph, marked Base, shows all the nodes and links with the addition of a label a, b, or. To traverse this tree, one provides a set of permissible labels to be used during the traversal proess. The middle graph illustrates the ase in whih only nodes and links with a label of a or are inluded. The right tree illustrates the ase in whih only nodes and links with a label of b or are inluded. One an inlude as many different labels as neessary to desribe the different path ombinations. A C D A B C D 3

5 Base A T x A T y A a b a b B C D a b B a C C D b b a a b E F G b a E G a E F b T x T y a T 1 T x < T 2 T y < T 3 b T 1 T 2 T 3 Figure 2: Time-Tagged Nodes and Links To simplify the annotation and seletion proess, a sequene of always inreasing (or dereasing) numbers may be used. 4 A onvenient set of numbers that meets this riterion is time. Consider a graph where every node and link has a time interval inluded in its definition (i.e., a start time and an end time). The lower part of Figure 2 inludes a timeline that denotes three time intervals using the three times T 1, T 2, and T 3. The time period from T 1 to T 2 is given the label a, the period from T 2 to T 3 is labeled b, and the period from T 1 to T 3 (the onatenation of periods a and b) is labeled. One an now apply the time-based meaning of these labels to the same labels in the graphs of Figure 2 and use any time on the time line as a value to seletively traverse or filter the nodes and links of the graph. Any node or link whose time interval inludes the provided time is inluded in the traversal proess. The middle graph shows the result of seleting time T x (from the timeline in Figure 2), whih is inluded by both time intervals a and. The right graph shows the result of seleting time T y, whih is inluded by both time intervals b and. This tehnique provides a simple mehanism for building seletable graphs using a single parameter (i.e., time) even though there may be many different intervals (i.e., labels) assoiated with the nodes and links of the graph. Notie that this tehnique may be used with any sequene of always inreasing (or dereasing) numbers. Time just happens to be a very familiar, and natural, hoie beause many proesses are based upon it. 4 In mathematis this is alled a monotoni funtion. In this ase, it is a monotoni inreasing funtion. See: 4

6 3. The Current Proess and Data Modeling In the urrent U.S. Army fore struture doumentation proess, douments alled Modifiation Tables of Organization and Equipment, or MTOEs, desribe the authorized struture of military fores. There are about 4900 MTOEs that over the ative, reserve and National Guard omponents of the Army. These MTOEs may hange several times per year. Fore struture data is an integral part of automated battlefield information systems. Managing the fore struture data within these systems is a hallenging task that is urrently aomplished via manual means. A major objetive is to move towards an automated system to manage this inreasingly voluminous data that is beoming higher in resolution due to the impending requirement for soldier level data. For this reason, the primary impetus for the design and implementation of a data representation is the ease and simpliity of automated maintenane. To be a viable system, fore struture updates must our in a manner that is automated and transparent to the battle ommand system user. The task of maintaining this data an be signifiantly simplified by seleting a representation that redues the effets of modifiation by isolating and minimizing hanges. To ahieve this objetive, a new approah for representing fore struture is proposed that uses timed tree graphs, like those of Figure 2. This allows one to exeute tree traversal algorithms to arbitrarily move up and down the tree seleting and deseleting nodes and links based upon time. However, a major task is the onversion of the urrent, disrete, doument based system into one that is ontinuous, timed-based, and uses formally speified tree graphs. To understand how the urrent system an be transformed to meet these new requirements, one must first understand how the urrent doumentation system is onfigured. An interesting feature of the MTOE is its dual personality; it is used to desribe both generi fore struture, like its ounterpart the TOE, and to desribe real units. The TOE is a requirements doument that desribes the model ase and is used as the starting state for building an MTOE. Assoiated with every resoure in a TOE is a value that indiates the number of assets required. To develop the MTOE, the model (TOE) is analyzed and adjustments may be made to the requirements, to inlude the addition of qualifying information, or in some ases, adding or deleting requirements for, or the amounts of, personnel and/or equipment. The final authorizations are based upon many other variables and onstraints, suh as budget, unit priority, and resoure availability. In most ases, the authorized amounts are equal to the required amounts, but sometimes authorizations are redued below the requirement. So the name Modifiation TOE is quite appropriate. The TOE and MTOE have nearly idential strutures, most notably, the use of multipliers in the doument. There will be multipliers for both the number of resoures required and the number authorized. There may be 18 Automati Riflemen (AR) required and 18 authorized. This approah is adequate for desribing generi strutures; but in real units, where eah of the 18 ARs must be traked individually, there must be 18 separate individual ARs. Clearly, there is a differene between the definition of the position named AR and the 18 separate entities that are the atual instanes of the definition. This is a ommon trait of data models for whih a thorough desription is beyond the sope of this paper. Suffie it to say that a real unit is an instane of a definition; and in this ase, the instanes are alled organizations (niknamed org) and the desriptions are alled organization-types (niknamed org-type). Therefore, there will be two 5

7 different strutures of trees: org-type trees that ontain multipliers and org trees that do not. The org tree ontains instanes of the org-type tree nodes that are expanded based on the multipliers in the links of the org-type tree. These strutures are illustrated in Figure 3 using the example of the platoons authorized within a ompany. 5 From the org tree and org-type tree strutures, one an understand why org-trees are neessary to desribe real units. It is not possible to use multipliers to represent individual units. On the battlefield, one annot trak 3 X Rifle Platoon that is represented as a single link and node in the org-type tree. Instead, this must be expanded into three separate rifle platoons named (in this example) 1 st Platoon, 2 nd Platoon, and 3 rd Platoon. This apability has been a primary impetus for the development of an Army Organization Server (AOS) that will provide automated battle management systems with a soure of fore struture data that meets their operational requirements. From Figure 3, one an see that an org is an instane of an org-type. This is expliitly denoted by the horizontal, dashed line, alled an IS-A link, that assoiates eah org with the orresponding org-type from whih it was established; for example, 1 st Platoon IS-A Rifle Platoon, as are 2 nd and 3 rd Platoons. There may be many (e.g., hundreds) of instanes that refer to a ommon org- Org-Type Tree Rifle Co Inf Bn (Meh) Org Tree A CO Co HQ nn IS-A A CO HQ nn nnn 3 1st Plt Rifle Platoon nnn 2nd Plt nnn nnn 3rd Plt Figure 3: Org and Org-Type Trees 5 A popular data model using this terminology and semantis is the Land Command and Control Information Exhange Data Model, or LC2IEDM, that is a proposed NATO Stanag. See: Army Knowledge Online aount required). 6

8 type, and this allows information ommon to all the instanes to be stored in a single plae via the org-type tree. Semantially, the org tree represents real units while the org-type tree ontains the desriptions of those units. This auses onfusion when omparisons are made with the urrent MTOE implementation beause of the MTOE s dual nature. An MTOE is used to desribe both generi fore struture and real units. In other words, it ontains faets of both orgs and org-types: an MTOE is used to represents a real unit, but it does so using multipliers. This is the quandary; semantially an MTOE maps to an org tree, but struturally it maps to an org-type tree (with multipliers). This irular ontradition appears to ause a lash in mapping between the urrent proess and future model-based shemes. However, this problem is easily fixed. The dual nature of MTOEs is exemplified by the fat that there are four primary identifiers assoiates with an MTOE: a DOCNO (doument number), a CCNUM (ommand and ontrol number, whih is analogous to a version number), a UIC (unit identifiation number), and an EDATE (effetive date the date the unit s status will be ompared to that MTOE). A minimum of two of these identifiers is required to identify an MTOE. From the perspetive that an MTOE is simply a modified TOE, the DOCNO/CCNUM ombination is the natural identifiation sheme. A DOCNO is derived from the TOE identifiation number (alled a standard requirements ode, or SRC) and identifies a type of unit (e.g., an INF BN MECH (FXXI), a Fore XXI strutured Mehanized Infantry Battalion). The CCNUM identifies the version of the doument and has four digits: a two-digit version number followed by a fisal year. So a CCNUM of 0103 would be version 01 in FY03. From the perspetive that an MTOE represents a real unit, the UIC/EDATE ombination is the natural representation. A UIC identifies a military unit (typially a battalion for an MTOE) and the EDATE indiates the date for whih the unit must meet the speifiation of the MTOE. So if one queries an MTOE database with either an UIC/EDATE or a DOCNO/CCNUM ombination, a single result will be returned. Figure 4 illustrates this dual nature of an MTOE. It is a omplex diagram that shows the relationship between DOCNO, CCNUM, UIC, and EDATE. There are nine boxes representing nine different MTOE douments that ontain authorization details for one or more unit. At the top of eah box is the CCNUM for that MTOE; and the boxes are positioned, left to right, in inreasing CCNUM order. Four of the boxes represent the older L-edition struture (initially designed in the late 1980s) and are distinguished by a CCNUM with underlined (green) letters. The other five boxes represent the newer F-edition (Fore 21) struture and are distinguished by a (purple) CCNUM without an underline. The nine boxes reflet the perspetive that a DOCNO and CCNUM define an MTOE that an be represented by an org-type tree. In this ase, one would state that there are nine MTOE douments present, identified with different CCNUMs. To the left of the boxes is a list of the four UICs of the real units that are established by this series of MTOE douments. 6 Dots are plaed in the boxes for the MTOEs that are used by the units 6 WAGL, WAGN, WEZE, and WEZK are the UICs of the four mehanized infantry battalions in the 1 st Cavalry Division: 1 st -5 th, 2 nd -5 th, 2 nd -7 th, and the 1-9 th Cavalry Battalions, respetively. 7

9 DOCNO: 07245LFC FFC10 CCNUM: WAGL WAGN WEZE WEZK Aug Apr Apr 03 Earliest E-Date 16 Ot 98 FY99 16 Ot 99 FY00 16 Ot 00 FY01 17 Aug 01 FY01 17 Ot 01 FY02 16 De 01 FY02 19 Aug 02 FY02 17 Jan 03 FY03 16 Ot 03 FY04 Time Figure 4: Time-Ordered List of MTOE Douments learly, every unit does not use every MTOE. Below eah box is the earliest EDATE of the units that use that MTOE. Other EDATEs are listed inside the box next to the dot of the units that has that EDATE. When no other EDATEs are listed, all the dots in the box share that same EDATE. From this perspetive, one would state that Figure 4 denotes 23 MTOE douments. This identifiation tehnique is implemented by plaing a list of the UICs and their EDATE in the header of a doument speified with a DOCNO/CCNUM. In this example, a set of dots loated in the same box represents the units (eah with a UIC) that share a ommon struture and authorization speifiation, although eah unit may have a different date (EDATE) to meet that speifiation. The defining question is: are there nine or 23 MTOE douments displayed in Figure 4? The problem is that the answer is yes. The solution is straightforward: an MTOE, as it is urrently defined, annot be represented by a single tree. The boxes are represented by org-type trees and the dots are represented with org trees. The result is that an MTOE and a TOE have nearly idential strutures that are represented using org-type trees with multipliers. One the generi struture is deided, an org tree is reated for eah real unit; and this proess inludes expanding the multipliers into individual, trakable entities. Finally, every node of an org tree is linked to the orresponding node in the org-type tree from whih it is established. Authorization information (e.g., personnel and equipment) is assoiated with the nodes of the org-type tree, so that this information is maintained in a entral loation that is shared by all the real units represented by org trees. One this design feature is in agreement, the issues assoiated with implementing timed trees an be addressed. 4. Time in Org and Org-Type Trees Figure 3 illustrates three strutures: the org-type tree, the org tree, and the assoiation between them (i.e., the IS-A links). Now onsider the ase in whih every node and link has a time 8

10 interval assoiated with it. The interval is denoted using a start value (alled s_date) and a termination value (alled t_date). Although all the nodes and links use the same format to define this time period, the use of this interval is different depending on the struture. In an org tree, whih represents a real unit, the time is real time. In this struture the term effetive date is appropriate as the time periods assoiated with eah node or link do indeed represent real time. If one wants to see the default struture of a unit on 15 Ot 2002, then that date is used during the traversal proess to selet only those nodes and links with time periods that enompass that date. Any node or link whose time period does not inlude the given date is exluded. Figure 5 illustrates a situation in whih a unit hanges strutures on a given date. In this example, the org tree has six nodes and five links. All the nodes and links, exept the right most, have the same assoiated time period that extends bak and forward in time. However, the right most node and link have a different assoiated time period. In this ase, the right most node and link are no longer viable as of 17 Aug Therefore, if one does a tree traversal using a date on or before 16 Aug 2001, the right most node (D Company) will be inluded; if a date on or after 17 Aug 2001 is used as the riteria, the right most node will not be inluded. In other words, this is how one represents the fat that on 17 Aug 2001 this unit hanges struture and the urrent use of the term EDATE ontinues to be appropriate. In the urrent vernaular, it is orret to state that 17 Aug 2001 is the EDATE for the transition of this unit from one struture to another; other units may have the same or different EDATE. Now onsider the org-type tree. Reall that this tree ontains the generi template (with multipliers) that defines the struture and assoiated desription about personnel and equipment that are used by the org trees. Several org trees may refer bak to a single org-type tree. It is important to understand how the different links that are assoiated with the org-type nodes of the tree differ in their use. First, within the org-type tree, there are the vertial links that denote the parent-hild relationships, with multipliers, among the nodes. These are illustrated in Figure 6 with some of the assoiated time periods annotated. These links are used by the fore development ommunity to traverse up and down the tree as the sequene of modifiation is developed and speified. Org Tree: 2 nd -5 th Cav Bn (WAGNAA) s_date: 1 Jan 1990 t_date: 31 De 2010 s_date: 1 Jan 1990 t_date: 16 Aug 2001 HHC A B C D Transition from L-Series to F-Series Struture on 17 Aug 2001 Figure 5: Example of a Time Based Org Tree 9

11 Org-Type Tree: INF BN MECH (FXXI) s_date: 1 Jan 1990 t_date: 31 De 2010 s_date: 1 Jan 1990 t_date: 30 Sep 2001 X 3 X 4 HHC MECH INF (XXI) s_date: 1 Ot 2001 t_date: 31 De 2010 RFL CO INF BN (MECH) s_date: 1 Jan 1990 t_date: 31 De 2010 Modernization from L-Series to F-Series Struture Figure 6: Example of a Time Based Org-Type Tree Figure 6 portrays the generi ase for the situation in Figure 5. There are two links between the top and right-hand nodes with mutually exlusive time periods. The link with the multiplier of 4 is valid until a date of 30 Sep 2001, at whih time the link with the multiplier of 3 beomes valid. So if the org-type tree is traversed using a date on or before 30 Sep 2001, the battalion is authorized four rifle ompanies. If the tree is traversed using a date on or after 1 Ot 2001, then three rifle ompanies will be authorized. Ultimately, it is antiipated that org trees an be built and maintained automatially via the exploitation of these links. For org-type trees, the atual value of time used to define the time periods is irrelevant. Reall that the org-type tree represents the nine boxes of Figure 4. These boxes indiate a relative state of evolution, and so it is with the org-type tree. Beause there may be many org trees referring to a single org-type tree for their authorization data, a single, real date, like an EDATE, is not meaningful. All that is required of a date is that it reflet a state of evolution, and that a later date indiates a later state of evolution (i.e., a point farther in the future). In other words, the time periods assoiated with the nodes and links of the org-type tree refer to an independent timeline representing evolution time. This is in ontrast to the time periods of the org tree that reflet real time. Instead of an EDATE, an org-type tree reflets a modifiation date, or MDATE, that is simply an indiator of relative time. Therefore, the MDATE and EDATE do not have to orrespond in any relationship other than a relative one. When one traverses an org-type tree, an MDATE must be provided to exeute the node and link seletion proess. This is analogous to stating: show me the generi template for a typial Mehanized Infantry Battalion with a modifiation state of 1 Ot This is the fundamental reasoning behind the proess of evolution for the fore developer. Seond, there are the horizontal links between the nodes of the org tree and their orresponding node in the org-type tree; these are the IS-A links of Figure 3. To obtain the information about what personnel and equipment are authorized for the partiular real organization, the IS-A link is traversed from the org tree node to the orresponding node in the org-type tree. End users and battle ommand systems will rarely, if ever, traverse up and down the vertial links of the org-type tree. They are interested in the real, not generi, unit struture, so they traverse up and down the org tree and only refer to the nodes of the org-type tree (likely unknowingly) to obtain the assoiated authorization data. From the perspetive of users outside the fore development 10

12 E-3 11B10 s: 1 Jan 1990 t: 31 De 2010 s: 1 Jan 1990 t: 31 De 2010 E-4 11B10 s: 1 Jan 1990 t: 15 Nov 2001 s: 16 Nov 2001 t: 31 De 2010 Grenadier Person-Type : Org-Type M2A2 s: 1 Jan 1990 t: 31 Aug 2008 s: 1 Apr 2002 t: 31 De 2010 M2A3 s: 1 Jan 1990 t: 15 Ot 2003 s: 16 Ot 2003 t: 31 De 2010 M2 IFV Crew Materiel-Type : Org-Type Figure 7: Authorization Information in the Org-Type Tree ommunity, the vertial links of the org-type tree are of minimal use and the purpose of org-type nodes is to store ommon authorization data. Analogous to the vertial links between the nodes of the org-type tree, there are links, with multipliers, between the org-type nodes and personnel and materiel type entities as are illustrated in Figure 7. Although the details of these entities will not be disussed, the links to them follow the same pattern as those between org-types. (In Figure 7, the multipliers on the links are all equal to 1 and are not shown to simplify the diagrams.) Like every node and link in the fore struture graph, those used to assoiate personnel and materiel data with an org-type node have an embedded time period that indiates a relative modifiation state. Therefore, as the fore developer traverses the org-type tree, personnel and materiel information is filtered based upon the MDATE provided to the traversal proess. The left side of Figure 7 illustrates the link with personnel data, and on the right is a materiel example. Both examples inlude two mutually exlusive links. The personnel example denotes a transition point (beginning on 16 Nov 2001) at whih time the pay grade requirement for a grenadier position hanges from an E-3 to an E-4. Similarly, the materiel example denotes a transition point (beginning on 16 Ot 2003) at whih time the authorized vehile for the M2 infantry fighting vehile rew is swithed from an M2A2 to an M2A3 variant. Just as before, the MDATE hosen for the tree traversal proess will determine whih of these links are seleted for traversal. Both of these examples show mutually exlusive ases: one annot be both an E-3 and an E-4, nor is one authorized both an M2A2 and an M2A3. In both ases either one or the other is seleted. However, this does not have to be the ase. A typial example is when a new piee of equipment is added without replaing another. 7 As previously explained, there are two ways in whih one traverses to an org-type node: one is from another org-type node while moving up and down the org-type tree ehelons, as in Figure 6; and the other is from an org node via the IS-A link, as in Figure 3. In both ases, an MDATE must be provided so that the orret personnel and equipment authorization information is obtained. In the first ase, one already has an MDATE, as it is required to traverse the org-type tree. However, an MDATE must be speifially provided when aessing the org-type node via 7 The hallenge of synhronizing the many MDATES is an interesting one, but beyond the sope of this paper. 11

13 Org-Type Tree: INF BN MECH (FXXI) s_date: 1 Jan 1990; t_date: 16 Aug 2001; m_date: 30 Sep 2001; s_date: 17 Aug 2001; t_date: 31 De 2010; m_date: 1 Ot 2001; Org Tree: 2 nd -5 th Cav Bn (WAGNAA) RFL CO INF BN (MECH) s_date: 1 Jan 1990; t_date: 31 De 2010; m_date: 1 Jan 2002; A-C Rifle Co (WAGNA/B/C0) AN/ PVS-14 E-3 11B10 E-4 11B10 s_date: 1 Apr 2001; t_date: 30 Sep 2001; s_date: 1 Jan 2002; t_date: 31 De 2010; s_date: 1 Ot 2000; t_date: 30 Sep 2000; s_date: 1 Jan 1990; t_date: 30 Sep 2000; s_date: 1 Ot 2001; t_date: 21 De 2010; Grenadier s_date: 1 Jan 1990; t_date: 17 Aug 2002; m_date: 1 Jan 2002; s_date: 16 Ot 1999; t_date: 16 Aug 2001; m_date: 1 Ot 1999; s_date: 17 Aug 2001; t_date: 15 De 2002; m_date: 1 Ot 2001; s_date: 16 De 2002; t_date: 16 Apr 2003; m_date: 1 Jan 2002; s_date: 17 Apr 2003; t_date: 31 De 2010; m_date: 1 Ot 2003; D Rifle Co (WAGND0) Grenadier 1 Figure 8: Combining Org and Org-Type Trees s_date: 1 Jan 1990; t_date: 16 Aug 2001; 54 in 2-5 th Cav Bn an IS-A link. This requires that the IS-A links have an assoiated MDATE in addition to the time period indiators (s_date and t_date). This attribute is alled the m_date. The situation beomes interesting when the features of Figure 5 and Figure 6 are ombined to produe Figure 8. This is a subtly omplex diagram that ontains several interesting results. Clearly, these are not omplete org or org-type trees, but only seleted slies to illustrate speifi features. Also, although every node and link has an assoiated time period embedded in its struture, only a few are shown to simplify the diagram. The parent-hild links of the org-type tree are also not shown for the same reason. Note that the m_date attribute is inluded only with the IS-A links that assoiate the nodes of the org tree that represent real units, with their orresponding node in the org-type tree that represents the generi ase and ontain the authorization data. Starting at the bottom of the figure and working up, one an see that many grenadier billets (in the org tree) refer to the single grenadier position in the org-type tree. 8 When an EDATE is hosen to 8 A distintion is made between a billet and a position. A billet is a real organization (an org) to whih a person an be assigned. In this example, there are 54 grenadier billets in the battalion. A position is the desription of a billet; it is an org-type and ontains the authorization information about the billet. So tehnially, an MTOE Line Number denotes a position, and not a billet. 12

14 traverse an org tree, different IS-A links are seleted based upon that date and the time periods speified on the links. The four links between billet Grenadier 1 and position Grenadier denote that the billet Grenadier1 has four distint modifiation states during four different time periods: Modifiation level (MDATE) 1 Ot 1998 FROM 16 Ot 1999 TO 16 Aug 2001 (EDATE); Modifiation level (MDATE) 1 Ot 2001 FROM 17 Aug 2001 TO 15 De 2002 (EDATE); Modifiation level (MDATE) 1 Jan 2002 FROM 16 De 2002 TO 16 Apr 2003 (EDATE); and Modifiation level (MDATE) 1 OCT 2003 FROM 17 Apr 2003 TO 31 De 2010 (EDATE). The ontinuous and mutually exlusive harateristis of these links are lear. The MDATE speified in eah IS-A link is used by the traversal proess to selets the appropriate personnel and equipment authorization links that are assoiated with the position (i.e., an org-type node), based upon the embedded time period of those links. This is the idential approah that is used to traverse any tree. The result is to relate the EDATE of a real unit with the MDATE of the generi unit. In this example, the result is that the soldier oupying the billet Grenadier1 is authorized to be of pay grade E-4 ontinuously through all its EDATES; and beginning on 16 De 2002, he is authorized an AN/PVS-14 night vision goggle. These riteria an be applied to all 54 grenadier billets or the links an be tailored individually for different results. The four links reated for Grenadier1 orrespond to the right most four dots of unit WAGN in Figure 4. However, in this example, two of the four links have no affet on the result; these two links are denoted with dashed lines. Beause of this, they an be removed and the same results will be produed. This is a ommon ourrene when mapping one-for-one between MTOEs and timed tree graphs. It is aused by an MTOE property known as the parent unit in the urrent doumentation proess. The ramifiations of this will be disussed shortly, but for the moment it should be realized that unneessary IS-A links an be produed and should be avoided. Continuing up the diagram, one should reognize the pattern of the top set of nodes from Figure 5 and Figure 6. This is the ase in whih the battalion (Bn) loses a ompany (Co) as of 17 Aug This is speified by the t_date embedded in the link to (and the node) D Co from the battalion node and auses both the node and link to expire for dates after 16 Aug The same is true for the t_date embedded in the IS-A link to its assoiated org-type node, whih is no longer useful after 17 Aug However, notie that the IS-A link from the other ompanies (e.g., Co A) is singular. This is beause nothing has hanged at the org-type node during the ourse of the speified time period. The RFL CO INF BN (MECH) org-type node has no assoiated personnel or materiel links, only links to hildren org-types, as is illustrated in the left tree of Figure 3. Even though there have been many hanges elsewhere in the org-type tree of the battalion, the links to the hildren org-types (i.e., 1 X Co HQ and 3 X Rifle Platoon) have remained onstant. Therefore, there is no need for multiple links. There are two IS-A links between the top org and org-type nodes. However, this is atually optional. As previously explained, the purpose of the IS-A link is to provide ommon authorization information, maintained via the org-type nodes, to the org nodes. Like the ompany nodes below it, the INF BN MECH (FXII) org-type node has no personnel or materiel information diretly assoiated with it; it has links only to its hildren org-type nodes. Two IS-A links may be used to reflet the hange from four to three rifle ompanies, as is illustrated in Figure 6 for the generi battalion. However, it is highly unlikely that an end-user would require knowledge about the generi ase when this information is already refleted in the org tree for the 13

15 real unit. Therefore, to be tehnially omplete, two IS-A links are shown. But this an be redued to a single link, as was done at the ompany ehelon diretly below it, if the end users deide that it is not required to distinguish between the strutural hanges in the generi ase (i.e., in the org-type tree). Note that the full org-type tree still remains, but the links between the orgtype nodes are not seletable when aessed via the IS-A link from the org tree. If only a single IS-A link is used (with a single m_date attribute), then any results will be based on the seletion riteria of the single m_date attribute. 5. The Retention Problem The primary impetus for this study was the development of a sheme to maximize the stability of fore struture data as it progresses through the evolution proess. In other words, the objetive was to eliminate, prevent, or minimize unneessary modifiations that might ause aggravation to systems already using the data. With the urrent MTOE system this is not the ase, based in part on the onept of a parent unit. 9 In tree graph terms, a parent unit orresponds to the designation of a partiular org-type node as an ad ho to form an org-type sub-tree. Typially, the parent unit for an MTOE doument is a battalion. By definition, a hange to any of the hundreds of element of the battalion is onsidered a hange to the MTOE and requires that a new MTOE doument be produed. An extreme example of this is between the MTOE douments labeled 1502 and 903 in Figure 4. In this ase, the two MTOEs are idential exept for a six-word sentene in the narrative of the header. Yet, a new doument was published based upon this hange even though the orgtype trees produed by these two douments are idential. The primary notion behind the approah presented in this paper is that as time progresses, every entity is assumed to be unhanged unless stated otherwise. Beause every node and link has an embedded time period, at any given time the fore struture graph will have an earliest and latest time within its omponents that defines the urrent epoh of the graph. As time progresses, the epoh should slide towards the future as obsolete entities are pruned from the past and new ones are added to the future. The time horizon of the fore struture graph is defined by the date furthest in the future that is embedded in any node or link. The fore developer ontrols the time horizon as the evolution proess progresses into the future. The reader may have realized by now that the key feature that enables this time-based approah to funtion so elegantly is that the lifetime of an entity an be extended by simply inreasing the value of the termination date. If this is done uniformly to all the nodes and links, then they all ontinue to be inluded in the graph. This is the default ase. Every node and link is retained, with its node or link identifier, as the time horizon is inreased and the epoh slides forward. Thus, stability is the norm. Only when hanges are required are the start, termination, and modifiation dates adjusted. This is what maximizes entity retention. If only one item hanges 9 Parent Unit: from AR 71-32, Fore Development and Doumentation - Consolidated Poliies, 3 Mar 97. a. A parent unit is an MTOE numbered unit of battalion or equivalent level, or a numbered ompany, battery, troop, platoon, detahment or team, that is not an organi element of a battalion. The 5th and 6th positions of a UIC that end in AA identify an organization as a parent unit. b. TDA units organized under a unique TDA number assigned by HQDA. See: 14

16 from one MTOE to the next, then only those hanges within the large tree struture are exeuted. Everything else remains the same, beause the only modifiation to the vast majority of entities is the extension of their termination dates to form a new time horizon. Figure 9 is a modified version of Figure 4. The basi differene is that the doument boxes have been reordered and replaed with vertial lines. 10 The CCNUMs have been retained at the top of the diagram for referene and a modifiation timeline has been applied to the bottom with MDATES assigned, in relative order, to the douments. 11 In the bottom half of the diagram, three examples of authorization hanges are provided. First is the hange of the battalion s organizational struture from four to three rifle ompanies. Seond is a personnel hange for a grenadier billet between pay grade E-4 and E-3. Third is the addition of a new item of materiel, the AN/PVS-14 night vision goggle. Next to eah of these examples is a line that represents the period of time that these features are in effet, based upon their presene in the MTOE douments. The dots for the four units whose UIC are listed on the left side of the diagram retain the same DOCNO: CCNUM: WAGL WAGN WEZE WEZK # RFL Co X 4 X 3 Grenadier E-4 E-3 Equipment AN/PVS LFC FFC Ot Ot Ot Aug De Aug Apr Jan Apr Apr Apr Apr 03 MDATE 1 Ot 99 FY00 1 Ot 00 FY01 1 Apr 01 FY01 1 Ot 01 FY02 1 Jan 02 FY02 1 Jul 02 FY02 1 Ot 03 FY04 Figure 9: Time-Based MTOE System The 1099 doument is not shown for simpliity. The 0903 doument has been removed beause it is idential to the 1502 doument, exept for a sentene in the header data.. Modifiation time is a monotonially inreasing funtion using relative, not atual, time. 15

17 meaning as before, although in a slightly different ontext. Extending from the sets of dots are vertial dashed lines that indiate a point on the modifiation timeline (an MDATE) for whih the unit has a synhronization point. Next to the dot is the EDATE to indiate the real date at whih time that unit is expeted to omply with the harateristis assoiated with the MDATE. If only a single EDATE is provided, then all the units aligned at that point share the same EDATE. If the vertial dashed line rosses a horizontal time period of the three examples, it means that at that MDATE those features are in effet. For example, the left most set of dots indiates that all four units must be modernized to the state indiated by MDATE 1 Ot 99 by 16 Ot 99, and at that MDATE, they have four rifle ompanies and the grenadiers are E-4s. As explained before, Figure 9 an be represented using a single org-type tree and four org trees (one for eah real unit). With the modifiation timeline present, the CCNUMs an be removed; and all that is required is an identifier for the single org-type tree (to replae the DOCNOs), beause the MDATE alone is suffiient to define the modifiation harateristis. There is an issue as to just what modifiation order means; that is, why were the MTOE douments reordered to the sequene presented in Figure 9? A preedene of modifiation is proposed that plaes modifiations in sequenes based first on hanges to the organization tree strutures, then on hanges to personnel requirements, and finally on hanges to materiel. The douments in Figure 9 are ordered based on this preedene. There is no mathematially orret modifiation order, but there are reasons that this sequene was seleted. The organizational struture domain is plaed first beause of the tenet that fore struture (represented by the org and org-type trees) forms the foundation by whih all other battlefield objets are assoiated. Changes to the struture of the org-type tree an affet all the hanges that our elsewhere, espeially the numbers of entities authorized. Therefore, hanges to the struture of the tree are taken into aount before any hanges to the objets that are attahed to the tree (e.g., personnel and materiel authorizations). As an example, there are signifiant strutural hanges aused by the transition from an L-series to an F-series organization (e.g., as is exemplified by the hange from four rifle ompanies to three, hanges in fire team struture, and the movement of ombat support personnel out of the HHC). Beause the highest preedene is given to organizational struture, one ould argue that any F-series organization is more modern than any L-series. In Figure 9, this results in reversing the order of the sequene of CCNUMs 0202 and 0601, beause 0202 was the last L-series doument and 0601 was the first F-series doument. The personnel domain was plaed seond beause it losely follows hanges to the organizational struture. When representing fore struture down to the billet level, all personnel hanges are assoiated with org-type nodes that are leaves of the tree (i.e., loated at the bottom of the tree). Finally, materiel hanges are onsidered. The information in Figure 9 is ordered, from left to right, in modifiation sequene as defined by the proposed preedene of modifiation. Fititious MDATEs have been applied to the timeline to provide synhronization points; these dates are merely relative, and any set of inreasing dates an be used. In the examples, the time periods for organizational strutural (four versus three ompanies) are lean and ontinuous beause these features were onsidered first when ordering the doument information. In the personnel domain, the grade of the grenadier has a short toggle between E-4 and E-3; but this only affets two of the four units. If the information were left in the original MTOE doument order, there would be two toggles. For materiel, there is now a gap in 16

18 the time period when there was none before. However, this is easily handled and a synhronization point with an MDATE has been reated to allow the real units to speify this set of features. If the three examples presented were the only features inluded in the trees (and in reality, they are only three of hundreds), then two of the synhronization points are unneessary. In this diagram, only the first five MDATES define unique ombinations of the example features. The last two MDATES (1 Jul 02 and 1 Ot 03) have the same ombination as the fifth MDATE ( 1 Jan 02), and therefore, an be oalesed (i.e., all four dots would line up with a single MDATE). This returns one to the topi of parent unit. Hundreds of harateristis (organizational, personnel, and equipment) an hange from MTOE doument to doument, but in reality only a small perentage atually hanged. For the end user, whose fous is the org tree of their own unit, and not the org-type tree, the onept of parent unit is unimportant. At the tehnial level, all the end user ares about is that for a given EDATE (used to traverse the org tree), an MDATE is provided in eah IS-A link that selets the orret personnel and materiel information from the orresponding node in the org-type tree. For the org tree users perspetive, the MDATE an be different in every IS-A link; this is of no onern. However, for the sanity of the fore developers, whose task requires them to traverse up and down the org-type tree, MDATES must be synhronized to some higher level. To date, this is done through onsolidation at the parent unit (typially battalion) ehelon. In terms of the time-based trees, this means that an MDATE represents the same modifiation point (i.e., state) for all the nodes and links below the node designated as the parent unit. Formally, one an say that the parent unit is the highest org-type node in a tree for whih all the nodes and links below it are synhronized to the same MDATE. Pragmatially, this means that the onept of parent unit an remain even when the fore struture proess transitions from one that is doument based to one that is time-based. The beauty is that this is simply another option for the fore developer. As the situation ditates, and the org-type tree morphs into new forms, the ehelon at whih MDATES are synhronized an hange to aommodate the needs of the development proess. 6. Summary This paper has presented a fore struture doumentation sheme that transforms the urrent fore development proess from one that is doument-based to one that is time-based. Fore struture an be represented as tree graphs where every node and link of the fore struture tree is assigned a time interval that indiates when that node or link is viable. The primary notion behind this approah is the assumption that, unless expliitly stated otherwise, every entity remains unhanged as time progresses. The key feature that enables this time-based approah to funtion so elegantly is the uniform extension of the time intervals assoiated with every element of the fore struture tree. Thus, stability is the norm. By using a time-based approah, the maintenane of numerous douments that depit snapshots of a speifi organizational entity at a given time disappears. Instead, there is an evolving fore struture tree with many different forms and options that are differentiated merely through the appliation, or filtering, of time. One an imagine a knob that, through turning, allows one to advane or retard time so that one an observe the evolution of an organization s struture, 17