"SPACE AND FUNCTION ANALYSIS" "A Computer System for the generation of functional layouts in the S.A.R. Methodology". by Alfonso Govela B. Arch. Universidad Iberoamericana, Mexico D.F., 1974 submitted in partial fulfillment of the requirements for the degree of Master of Architecture in Advanced Studies. at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY February 1977. Signature of Author Departlent of Architecture, February 1977 Certified by Prof. Nicholas Negroponte, As ociat Thesis Supervisor. - Accepted by Chairman, Departmental Committee for Graduate Students. Rotch APR 1 1977 rofessor 2 ACKNOWLEDGEMENTS I thank my Thesis Advisor, Associate Professor Nicholas Negroponte, for presenting, in a Su'mmer Session back in 1973, the possibilities of computer applications in Architecture, and for providing later on the opportunity and the support to work with him project. and to develop this I thank John N. Habraken, Head of the Department, for the patient sessions where he explained in simple terms the basic concepts of the S.A.R. Methodology that appear here in a convoluted way. I thank Mike Gerzso for the discussions we have had on Architecture and Computers, and for starting, through his work, my interest in Production Systems. issues, I thank Seth Steinberg for his constant help in conceptual matters. programming techniques and practical The work reported herein has been sponsored by CONACYT, Consejo Nacional de Ciencia y Tecnologia, Mexico, D.F., Mexico. Alfonso Govela February 1977. 3 TABLE OF CONTENTS Abstract .. ................................................... 5 0. Introduction ................................................. 6 1. Problem Definition and Assumptions ........................... 1.1. Background .............................................. 1.2. Problem Definition ...................................... 1.3. General Assumptions ..................................... 9 9 12 13 2. Methodological Basis and Enumeration Techniques .............. 2.1. S.A.R. Formulation of Design Problems ........... 2.1.1. Parts and Relations ......... ........... 2.1.2. Levels of Definition ........ ........... 2.1.3. Operations of Analysis ...... 2.2. Enumeration Techniques ............. ........... on 2.2.1. Problem-Solving, Representati 2.2.2. Representation .............. ........... 2.2.2.1. Basic Model ........ and.Search. 2.2.2.2. Graph Notation..... ........... 2.2.3. Main Types of Representations ........... 2.2.3.1. State-Space........ ........... 2.2.3.2. Problem-Reduction ........... 2.2.3.3. Theorem-Proving 2.2.4. Search ...................... 18 19 19 22 24 31 33 35 36 38 44 44 50 54 56 56 63 2.2.4.1. Basic Techniques ... 2.2.4.2. Backtracking ....... 3. Space and Function Analysis ............. ........... 3.1. Formulation ........................ ........... 3.1.1. Site ........................ . .. .. .. .. . . 3.1.1.1. Actual Site ........ ........... 3.1.1.2. Formal Site ........ ........... 3.1.2. Elements .................... ........... 3.1.2.1. Physical Units..... ........... 3.1.2.2. Use Space .......... ........... 3.1.2.3. Restrictions ....... 3.1.3. Relations between Elements .. 3.1.3.1. Simple and Compound Relations ........... 3.1.4. Position of Elements ............................. 3.1.4.1. Levels of Definition, Expansion of Position Rules............. 3.1.4.2. Position Rules in overlapping zones..... 3.2. Standards ................................ .............. 79 79 79 80 82 87 88 88 89 90 92 97 99 105 109 4 3.3. Enumeration ............................................ 111 3.3.1. Overview ........................................ 112 3.3.2. Description ..................................... 114 3.3.2.1. Truth Table ............................ 120 3.3.2.2. Evaluation of Position Rules ........... 123 3.3.2.3. Preclusion --............................. 126 3.3.2.4. Zone Dimensions ........................ 124 3.3.2.5. Sector Dimensions ...................... 130 3.3.2.6. Permutation of Elements ........ ........ 133 3.3.2.7. Margin Dimensions ........................ 146 3.3.2.8. Evaluation of Element Constraints ...... 147 3.3.2.9. Circulation ............................... 154 4. Computer System ---............................................. 4.1. Relational Data Base ....................- --- -- --- -----4.2. Spatial Representation ................................... 4.3. Query Language ......................................... 4.4. Search ................................................. 5. Conclusions -.--. ---.----.-.-.--.--.-.----.-Appendix 158 159 168 174 177 .......-.--.--. 181 .................................................... 185 185 5 Abstract "SPACE AND FUNCTION ANALYSIS" "A Computer System for the generation of functional layouts in the S.A.R. Methodology". Alfonso Govela Submitted to the Department of Architecture on February 1977 in partial fulfillment of the requirements for the degree of Master of Architecture in Advanced Studies. As part of the S.A.R. Methodology, a set of computer programs has been implemented to carry on the systematic generation the possible functional layouts for a given design criteria. of all They are intended to help the designer analyze and evaluate function, and display the the relationships between a space and its consequences of different design standards on different sizes and layouts of spaces. The main assumption is that a space can be analyzed functionally by looking at characteristic arrangements of furniture or equipment, that correspond to a certain function. A function can be defined in terms of the location, dimensions and relations between furniture elements and spaces. A set of design standards describe a spatial system and constraint a solution space where particular layouts can be effectively, and if necessary, exhaustively explored by a procedure that generates as many arrangements as desired. This generative capability is aimed to help in the development and evaluation of standards for spatial performance. By studying the different layouts that each set of standards permits, different evaluation techniques can be defined to compare, select and agree on the most adequate criteria for an actual situation or an hypothetical case. TSa Thesis Supervisor: Nicholas Negroponte Title: Associate Professor of Architecture. 6 0. Introduction: This T hesis architectural fu nctions it reflects this comes it hand design from an on their other the an from generat ive and tools within design and grown it has one On sections. in spatial Interest soatial Interests through several apolicatlons as a nalytical techniques, formulation of of different the computer in with the the analysis in t he methodolo gles, Interest and I t has origirated consequences. and concerned is of framework design. The main proble m that logical in S.A.R. Methodol ogy. characteristics, all the conceptual a participatory design alternative models that field of solution have become Artificial structured view of the which basis functional to mass spatial enumerates function an d come from two a space. different discipline origin ated as an housing standard practice Intelligence. problems in standards provide a an d present a procedure that a first " supports" to functions possible relations between Its fields, and attempts spatial way to define systematic It the is designing orocess of operation in the the it approaches are probl em s, in the The first design on the described, and the computer provided a basis while of the 7 "Problem-So lving" through second, A I thou gh around of first 'Tree-Searching' different generating the permit. represents new Ideas sometime. However, its application an In different in Idea that we might say, is an second, for is case the t he architecture, this way layouts that these standards functional has been a provided techniques, and in sense. Important Pr cb Iem-So lv Ing has been oriented to the representation of procedures find of range that solutions po ssi bilitles procedure tha t " sense, beside s b eing far design, would by searching according to some given rules. designs* or finds design solutions mI ss the away in terms of issue of the prob lem and A in this our knowledge of in design. values implicit What matters is not only the solution, but the of a through definition the process to reach this result, as values are de fin ed constantly In each decision taken. In design, it is more relevant the problem of Not how we ge t solutions but how do we evaluate consequences. To analyze decisions, we their consequences and through the we have enumeratior of have to enumerate results decision against another and select more evaluation. to their their evaluate results, we can com pare one the adequate. Searching and enumeration one we consider are in p rincip le 8 equivalents, and it fun ction At the scale of small a extend of par t oroblem of values, analysis, computer use the approach to as enumeration used In Prob lem-Solving are techniques of the in the context of Is is it and a room, method t hat can be used as its generative Thesis. no t another attempt to that reco gnizes the a methodology t o the scale of this but an designer, that provide a main analytical operation. In chapter first the assumptions. main presentation of the problem area Following it, is defined with there Is an the S.A.R. principles and a our general o utline of the Problem-Solvirng representations and search t echniaues. In the Artificial SPACE of third Inte chapter of the principles I ligence are applied and and FUNC T ION ANALYSIS, the computer system and in Is presented. both S.A.R. and extended the fourth an to the out Iine 9 DEFINITION AND ASSUMPTIONS. PROBLEM I.- Background: 1.1.- a have Spaces in human purpose, they Within them, activities. continuously identifying refer its to in in actions or call, think as, the "function* that has or carrying kinds are ways. The fact , our in names that language, with that can be reali zed within activities on a series beer proce ss created, or simply result later point from In time, can a Usi ng a space from a th e space is adaptation, spontaneous that to a function was neither considered in its creation, nor planed to be contained it. In depends of either case, the success of or fallure In the relations between different the space, such as shaoe, proportions we space, or resul t the time when at what is made of a assigned to it. of activities planning a constitute *use* that careful at humzn limits. This series of the of in shows everyday the action or different multitude purpose, their them, a for containers actions of performed of Importance as exist frequently must environments, this by "use" characteristics or dimensions, to 10 name a to few, it. the kind of Spatial sometimes certain and characteristics in a definite strong the process to spatial types initial, if spatial solutions. set at up not often form the standards These global, implications to a of the the of of the the range seldom understand what solutions might in the of possible a in form of spaces for framework existence, or a for the definition is made. By we sometimes the two, the impact to norms or personal characteristics rationale, be on different exist or provide comprehend the relations between their and in the design. These their of always roughly very end, according acceptable their one assumptions already when a change in of Is process circumstances function. the very unaware to assignment taken in the generation understanding the reasons behind their fall prevent, of the uses, preferences delimit oreferences, set being or performance types are almost the time, of that but or until cultural corresponding design, beginning more of of activities fluctuates Must space, way, the first step the definition the permit assigned functions. In other, function that can be of uses. fall to and consequently different spatial 11 might be considered a tri vial increasin g *performa nce* anal ysis of buildings cost of actual more "sof ter" life per ioc. las interest spaces out years, last the being shifted fcr The economic from to tte the building construction, in the builoing use over more the Its whole costs between the Initial ( the maintenance bills X) more substantial, understanding how the importance of are used.(1) At other there has been an where sav ings can become pointed has exploration is needed. however, The propor tion of indicates areas and of design experience different kinds. construct ion expenses ( X ) and for any need matter level, of studies functions without d uring interest to spaces problem further systemat ization or At a practical "sol id" a simply or explanati ons, where no rel ating of The problem a deeper, hand, values and becomes there more has assumptions aparent that significant level, been an increasing questioning behind design desigr problems with solutions, are clearcut not situations difficult problems which solutions represent of values, and which values must technical be agreed implementation. as the of it we Il defined solutions, technical attempting any on but implicit set upon before 12 (2), As Rittel critique increasing the the challenging their with preocupation lob, professional's pointed out consequ ences seen fact that "...the by eroded seems to be support renewed equity. The as solving an to be definable, is being confronted now of Is beirg .... "...There values...". a growing realization professional's by seeming consensus differentiation ... stems from for ap peared that understandab le and consensual...", with the work ef ficiency, for correctly, quite on prof essional "....once problems of assortment tests has that a weak strut in the where li es at the juncture system goal-formulation, problem-definition, Issues equi ty and meet...". How to provide a basis for spatial and evaluation of level, -of analysis, are course- the and how to coordinate standa rds main two they were not the or values first kind of the formulation at this second issues which motivated, although solved, In the work done this thesis. 1.2.- Problem definition: The idea behind SPACE to develop, along the lines of analysis was S.A.R. methodology (3), and FUNCTION the 13 a systematic way of -figuring out the relations between a space and its function or its set of possible functions. Its main a makes space objectives for adequate to were understand a certain use, when can we realistically assign activities to a given soace, the are what or what consequences of changing spatial characteristics or functional requirements. How to formulate functions, describe spaces, and analyze the relations between of parts the thesis function, one or all in two, these are the main work. How to find out, for a given the spaces where it can be contained satisfactory way; or for a given soace, how to find a out, one or all the functiors ti-at can be contained by it, are the particular questions that we attempt to development of would answer. 1.3.- Assumotions: A design problem In itself, this the SPACE and FUNCTION analysis has been approached with certain assumptions In mind: - Independently of defending, explaining where personal it was considered more rationalizing or values for a function come from, relevant to find out how can 14 in a way that functions be expressed in terms a function can be It was assumed that of pieces of the fur niture or a space, of that shou ld positions we can define be in contained and space To talk about then the series of by together it, the poss defined equipment that are needed to carry on a certair a ctivit Y. use understand implicatlors. their spatial - us hel Ps the el ements with their ible relations beween these terms, different pieces. - a set of define An Functions or standaras a range of arrangement "acceptable' of for one of elements In - To arrangements that are by what of their elements the have to a space can have. correspond a *functional to layout*, distributions variations implications, Functional c onseouences lo ok implicitly we can analyze, - that that I mplicitly which possibl e valid the evaluate we possible, u ses represent this and of space. standards, knowing in systems value definition is called functional it stanas formulated at of all the permited by In functiona I understand or design values in definitions different possible them. Only layouts are refutate , in terms design s olutions. are, there fore, not 15 attempted as definite statements open to evaluated through or laws refutation, the patterns, but rather as enumeration are implications of their as possible consequences. The formulatior of enumeration assumptions of behind methodological the next their basis chapter, Implementation of four. its functional standards and possible SPACE and layouts FUNCTION are the the main analysis. The for these two parts are explored final form In chapter in three, and the a generative computer system in chapter 16 One, Chapter No test -----(1) "Life-Cycle costing in the Public BulIdings Servi ce", G.S.A. General Services Administration, Public Building Service, for Inc. & Hamilton Study made by Booz-Allen G.S.A. ,P.B.S. Theory General a in "0ilemmas (2) RITTEL HORST, PlannIng", Working Paper 194, University of California, Berkeley, Cal. Nov. 1972. BOEKHOLT (3) HABRAKEN N.J. J.TH., THIJSSEN A.P., 0INJENS "Variations: The Systematic Design of forthcomming M.I.T. Press. of P.J.M., Supports", Cambrldge Mass. 1976. -7 18 2.- METHODOLOGICAL BASIS AND ENUME RATION TECHNIQUES or norms Standaros that the permits particular so obvious, they at least, layouts; should permits the systematic be be permit if understanding transparent Architecten be problem scale analysis; reducedand Problem-Solving as a two Group basis that, at the to of the second, and and not a way to point and can formulation must the but that defined possible from ma in of principles Research> allohs layouts standpoint is it not of so implemen ted. this chapter general way layouts. these Tha t their clear methodological rather, in is a clear implications , their a that be s ufficiently layouts, how it can be In the might in important, exaustive enumeration o f all necessary, First way but, more in different ways. the a stru ctured must the evaluation of solved in enumeration of That standards permit formulated their conseauences. They evaluation of should be described, testing be must (1), the S.A.R. having solve d of scale the can be aoplled in the as a principles actual -or and functlon Enumeration techni cues are framework that can be used <St1tching this formulation s pace general revised. presen ted are housing, are Id eas of presented generation 19 of layouts. 2.1.- S.A.R. formulation of Design Problems$ 2.1.1.- Parts and Relations: In the S.A.R. formulated be in terms of "...an that have to be placed in that is a site, always problems methooology, design environment environment and elements environment..."(2). or context where in terms of that space, fro m define in design solution can always be expressed site, Depending on vary There different elements can be positioned, and the stanoards that a set of values can its and their the scale we the spaces in an d accordingly, elements city block, the elements work, positions. the site to one area in that are can room positioned 20 them In can range from perform a function to dwell in an already and they pieces as dimensions, well needed can appear is that in design or contexts that are external problems an oeople constrains given that exist to the designer actions, urban blcck, as might where be situations, and or they stand instances The elements or They can the case of infrastructure, size, and surrounding buildings are established defined; for several to as be defined during the design process. can site furniture the support structures where represent one specific situation, a of urban environment. Sites situations, the can be used to then as general and describe multiple or schemes models of similar sites. that are positioned In a site can 21 be defined by the designer possible options. that relate In either elements and or selected case, to formulate elements site, described with sufficient exactitude. elements, from some range shape and dimensions, of standards have to be Tyoe and number of are the basic information -. A I IzzIIzzzzz~ t 71 1 J- 777f t 0 7t 22 that them. description of for any be required aoula Among these elements there are certain relations have that should in other each being Elements different ways, that that requirements solution, as ard In present to their from, seDarated result relations Their etc.. by, relate can is in terms of other, each to near adjacent, contained also be to be defined. from program the design should be part of these requirements they should aescribed precisely. A is methodology, designed defined" "well then formed by the problem the in following parts and relations: environment. 1. A descriotion of the 2. A defined set of elements that could be used in the environment. 3. Data about the Iocation of the Iocation ot elements relative to one another. 4. the Data about the elements In envir6nment. 2.1.2.- Levels of d ef inItion: The formulation of Positions, can be applied different scales of Site, at Elements, different the design problem. Relations and Ievel s and at As said before, 23 L--J the Site can be the elements an urban can be block, or a s pace in a room, struc tures, housing or and furniture pieces. It the Site at is characteristic one scale, in this formulatIon, becomes an Element at that the next 24 level up, and its elements f or next the con sequently, formulation one level stand can positions used be in their s ite, the defin itio n of in layout with one possible functional site. an element elements A dwelling w ith all In the 2.1.3.- Operations of The evaluation layouts of is performed by f or a given site We structure , start explore next up. and together a for element and become buildings level up. the selected to pop up as elements we reach To continue among the the possible room level the of possible la yer an d in t1e if satisfled, ns as elements genera te aga in all wh1ch, in t ur n, - layouts of in the fi na the the are some to the nex t el em ent example this level. smaller In there we a ny al I layout variatio variations u ntil Analysis: given possible the level one its rooms becomes then alternatives its for above be an generating with select some basic levels, standards standards at at a agree to the site, building that elements the urban block. in hierarchy , the together with furnitur e layouts, for A room can be the Site dwelling one of Sites The variations down. of el ement layouts at one scal e become then, their as in the I layer. IlIustrations layouts, or the uses that can be 25 1en I - 1 26 e;2'C7R A4AAs:15 L4 /A45,C J4ZI4771N5 : 4::, 27 All 00;/B0g <//E gq/Bl N/ A/ E 401/C VA/477V6 grAr/<! V M/4 o,*,y<! some would of m would of 3.60 x 4.10 And the variations floor, X/ </' CoA%'; assigned to a room apartment AV/ a// 0-3 of room be: ( arrangements be genericallyt4fiSc which could result in an V4A%9/74WS> in the oossible floor p I a n s ,(i!-V)e45W. in an This formulation structures a design problem In ano which could be Positioned, similarly urban block a way that as: 6/-,A r7ststrE} permits the systematic enumeration of In chapter 3 these general the scale principles will of SPACE and FUNCTION ANALYSIS, shown through an.example how is the be layo uts. applied to and it wil I be enumeration car -red 28 $3 9Z as X/ 'V) '/ &/9-Y477Nw - 29 4 1 "'1o ii j~J /w A4P14 iii iii~i ate 111 1[~ ' UAwAV A I gfW4* 7/55r 30 ~~LT3 on. In from techniques drawn from the fol Iowing other then the enumeration. sections, fields basic will some relevant be described as we aporoach for our method have of 31 Enumeration Techniques: 2.2.- the beginning of At Combinatorial furniture Theory were explored as it arrangements layouts enumeration of are both problems These whole booy particular represented problem, of among of where elements furniture as set, A furniture a layout SYSTEM of for for of repeated) a in on In a our could the site it, and we of be a set see each pieces, set, positioned one in for all our factors, of our problem, piece a and considering the of the sets case, like size, this distinct we have. assignment fitting in site, relations to other elements, that cannot be formulation of REPRESENTATIVES pieces as a 'system representatives' several to use in selection the spatial Unfortunately, depends nature. are positioned respectively in each collection of positioned (not therefore the examole, each space can be positioned that assigning atractive DISTINCT the spaces. Considering then each *representative' of that could be brought configuration as each that and the existence of configurations we define that was thought of a strong combinatorial theory case. from the site, and techniques presented the of techniques constitute a problem to parts in pieces furniture the Thesis, defined in the as our problem consist In fact 32 In finding out at aone first, and all, correspond what of Artificial and and as of were techniques functional of evolved explore, other among for frameworks studies of the development of these heated an for our use for the expanding the in Artificial computers, during the the last 20 formulation From years of of psychological techniques that per.mit approaches, (3), it has a Whether the existance of a tasks, procedurzi produced, series of to our besides principles and problem. these techniques show Intel ligence in the person or system any degree uses them, or whether there can be to general proceed in solving particular are relevant of of can interest things, polemics techniques that models standards. *problem-solving". how people definition we from methods explored as that application Intelligence has general be that layouts 'tree-searching' Generated from quite can "problem-solving" alternative, Intelligence, evaluation to the enumerating then an representations areas assignments to them. As problem these such a thing that as a problem solver, are questions not only beyond the Interests of this thesis, but questions that tend to 33 the relevance that these methods from us distract have In themselves. The problem-solving of outline basic follows tree-searching methods representations and general, in Problem-Solving 2.2.1.- on. Representation and Sea rch: We can say t tat Is and situation a the this actio n, and bring the actual our in A problem, recogniti on and from we think we ought case, terms, it that we to t ave. desired in the situations can take us to the conditions that have, for discussion on go about producing plans of be of (4) is enough room might the consists that that we actually our these a plan of conditions reduce situatlon. actions There for to different of find to that can of discre pancles between the p roposal the ideal forced it be. should characteristics the prese nt situation as near as possible character istics of it we are or the col lection of actions, dif ference as a situation that discrecance, this with Confronte d think we as solved is a problem to be perceive a d iscrepancy between we when there sufficient how action, to say that it is but effective 34 of must actions, the time, result from previous thought anO evaluation, and such thought and evaluation arise from an understanding, through an internal model, of the in an Problem structure that we are confronted with. A model of internalization of the structure this consists main characteristics of the problem the set of possible operations that we can perform to and bridge the gap between present and desired conditions. To build models or representations of a problem, a process of urderstanding, we engage in have to demand certain conditions in we build. reality, produce, the with we situation, that results or otherwise of the processes our formulation in of problem that to reach its solutions. structure of the problem practical the descriptions that the want to *real' solutions can Descriptions should lend themselves to expression expression of attempts the actual results our not be of any use. practical we we wcrk with a representation instead of since if situation might so should not contradict aspects of Descriptions dealing directly with correlate and to do a can and be permit used in the our We want to describe the consistent way, with a its Information and a relevant representation of the processes involved in changing old 35 conditions into new, more desirable ones.(5) and how to manipulate the two main then conceptual methods, corresponding in problem for solutions, are looking in issues the for to build such descriptions How *problem-solving* to REPRESENTATION and SEARCH. 2.2.2.- Representation: problem that specify how initial actions that defined by the a set of As problem is structures symbol problem be used in among laws, we are Newell which actual or situati on, legal of set the oroblem are and the combination of specify there the might extent of the exist looking for. and Simon present to designate (1) a TEST generate it a "'...To for the problem), structures (potential to The solving and transformations, (solutions of is transformation the describe situation. transformation situations solution that laws of an intermedlate or partial or *goal' can both conditions of conditions situation, desired with to change one condition into another. Problem and the together conditions describe representations "Problem-solving* a a class of sym bol and (2) solutions). structure, state a To using GENERA TOR solve (2) a t hat 36 satisfies the test of (6) (i)..." Basic Model, 2.2.2.1.- Post Production Systems: The basic principle behind these representations computation mechanism presented goes back to a general Emil Post in 1943. Post systems logical finite alphabet, that rules transformed logical some string how of other string systems as of symbols". words in backwards, that in in some "*sets symbols may of be (7) for example, the that represents, "palindromes*, and forwards symbols written and analyze A simple model structure of of strings as tell into proposed to analyze expressions by read identically a Post's Production System, would bet Alphabet. a,b,c Axioms or initial situationst a,b,c,aabbcc Productions: -- > a$a (P1) $--> b$b (P2) $ c$c (P3) $ -- > where, can use Initial in constructing rew strings. situations, the the alphabet represents or the strings that symbols The axioms we take as we are our given, 37 not derived that string, any string, from other that is either from has been generated to productions ordered *$' any and symbol a right side such as on the right. That As can will system is *$* into be seen from only on the given *c*, already, and elements the fact each that the of 2;- c c A- aca I- b |acb an initial problem of that will left the string Into r ules, the and alI tte the letters rule mirrors the same a previous word. *bacacab* would bet peda1ron (p;) a - Cac (P/) cac . _ dt'4e ( and the *Problem* of out of as e,4'g/ron .Mi/A - such production The generation of the word /, an the axioms are " oa lindromes' production in both sides rep resent 'a$a*. possible *palindromes* compose d from or appilcation of left side, "palindromes* generate string which Indicat e possible " a$*, the string of transformations with a a or productions strings pair of for a stands an ax lom the success ive The axioms. "* source. situation finding out take us from 'a' ca4' Into -4CQCW generating a goal would be the sequence *a' c4 -+a of *bacacab* equivalent to the production *bacacab". rules 38 the both then, solutions. The structure of palindromes expressed as a yield resul ts external that are ex ternal "f... short) of of " ... In as of transformations these to the system nor by enriched I aws, transformation its of Interp lay and collection mere a the structure is preserved or Involve law s! the properties, their and elements not and system test for is understood of transformati ons 'system much as It is a and the for generation capability i tse If has built within problem descr iption Our which never elements employ to it..." (Its) three key transformation, notion of structure is the Ideas: and the comprised (In Idea of wholeness, idea of self-regu lation..." the idea (8). 2.2.2.2.- of this Graph Notation: Before going basic model, principles notational problem-solving methods we should used in look the that are relevant Problem-solving representations being systems mathematical of notion of variations Into the description of transformations, first at some of description to this share, the use thesis. besides of a GRAPH as a common notation. the (9) 39 G = A "GRAPH": (NE), of 'nodes' used to represent a set of nonempty exist set between elements we want corresponds to these elements by represented lines that and a set "N', elements The set them. consists of and nodes "'N to talk about, and of of a finite, "edges' relations stands for the set of edges "E", that the "E" the relations that exist between pairs of in the dots Graphically, set. circles, and or link related nodes. A edges "GRAPH" nodes are are shown as would be, for instance: 10 G = (NE) N = (1,m,n) E = ((I,.m),(I,n),(m,n)) If we reoresenta tions as think, being before, of pro blem descriptions of pro blem did we general situations related by transformations, we can begin to see the corres pondance between the notion of a graoh and the notion of problem representation. At elements, a first conditions level, or nodes can characteristics edges can correspond to desired relations correspond in a to prob Iem, among them, and 40 a final problem situation. the s et graph standing then for we descriptio n of their can it which set a of above in as in used to reoresen t bot h the p rob lem the of generation nodes were not listed one by one, as simply or edges were defined as between or it can be f igure(2.1), situations edges, general addi t ion of e xamp Ie valid of words palindrome" but combi nati ons; or as the letters , where Initial as defined produc tion rules, schemes I mpl i cit ly, described and where relations strings, denoted b y * $', new conf i gurations that contain the previous an the and situations, described expli citly, as In the be can th e and p arts According on how w e defi ne the nodes and graph "a", "b* or *c*. scheme It situations and production and plus is said that the graph Is defined implicitly, because by means initial the general problem transformations. structure of a or Being the graph a n abs tract be oarticular of s ituaticns am 0ng a structure, tha t is rel ations, structure and edges rules, production of solution. a for search can one situation in t o a new diffe rent one; from changes in themselv es, la ws 9 transformatlor represent nodes level , second situati on s problem now become a At or a a nartial an initial, for then stand would graph of rules, we can always 41 have a way of the set of set of links, each and exo I icitty everyone of of of is a member to define both sets. of one portion of the would be: graph, "palindrome* having the members representation A graphical Impl icit Instead of a member a string Is a transformation or whether nodes, of the finding out whether 4040 "Ccc" AaCa94=iv where we we can select each of can continue each apply so doing "palindromes" further too we at the where of as everytime the long the have an empty string, the possible axioms, possible as to which productions, we want, graph from grows and generating new further and down. If this graph notation is going to be used as a 42 besides purpose, we what is for concepts that are relevant important to define several this then it for problem-solving representations, convention about have already said nodes and eages. have a seouence of When we edges of the form: (nign2),(n2,n3),(n3,n4),....,(nnnn-1) corresponds to this sequence at node the where the end each of the node at the beginning of the Is called a "path". A linearly it can also be represented next edge, *oath' goes along a connected nodes and sequence of edge for this reason ast n1,n2,n3,......,nn P1'z and be said to have a "path length* of n-1, that 43 the in edges included to the number of equal length Is a sequence. with s ay that and las t nodes another, a sev e ral oaths. the graph, t hen graph "Trees* are is is a Initial node, bottom of the root the set has n o cycl e at each of from of has nodes first node, In to some because the in sequences of nodes from an final nodes, down at tte Initia I situation p lus the rules a 'tree' that might whose branches situation, transformations to constitute a solution; finding a solution becomes then equivalent to sequen ce all, the other t he the possible paths from tha t given situations m * I* fro called a *tree*. the graoh. The of t o one path' to ever y node transfor mation, can then ' grow' are all equal one cycle, has Important graphs to us, , first "cycle'. and p aths they show distinct their the th ere the *root', called edge, a re the like the second one , a graph, When When path* figure(2.1) one entering have onl and last node, n ode which no edge enters, first of first and the a nd the graph In figure(2.2) *i*, but distinct, *simple'. Is this path a in graph the 'simple a of then we call The to the 'path* are 'path' a in of possible exception then we nodes all the When finding through a certain path 44 that can node down desired to a root the tree take us from its branches. Everytime in the down. The set of at in turn becomes the of say we expand the node one the end of these expanded a level edges the expanded node, and of 'successors* the called these nodes out edges grow a series of we tree, node are we to the newly paths, are basic *predecessors* created nodes. Nodes and edges, of concepts expressive methods theoretical on chapter(3)). 2.2.3.There representations not but possibilities, to other (ref. and Theory which are used as notation for Graph problem-solving trees only of because because they give notions that will Main types of have been three be general kinds of (10) STATE-SPACE REPRESENTATIONS. 2.- PROBLEM-REDUCTION REPRESENTATIONS. 3.- THEOREM State-Space: later Representations: 1.- 2.2.3.1.- us access explained in problem-solving methods: PROVING. their problem 45 In conditions or problem of of characteristics STATE-SPACE representations, problem situations which DESCRIPTORS" "STATE in case the are described by certain represent the problem solution at a certain poirt and situations are time. Initial final expressed respectively as existing and desired characteristics which not necessarely have but that strings, approplate problem "palindromes*, for a form words *a' and the for the problem . appear Legal as the same fashion description back combinations what In of more In the case of tte In hand. and of transformations in In Post DESCRIPTOR" all is this Production 'bacacab". representation our the very much how can the be reached by initial state "'STATE-SPACE", that example, contain the to DESCRIPTOR". into a new "STATE 'pallndrome' words that in Systems, situations that called the state descriptors final generating the expression as such 'bacacab* would have been aoplication of these operators to constitute format of instance, strings would have been The set of still the "'OPERATORS" or rules that soecify, change a "STATE the of form descriptor and the state Initial be restricted to can take any the for to letter all *a' Is, the in the 46 middle and or ec" whose letters repeat on each side of where Nilsson(1971), a is strings, a model for a 8-puzzle. sliding-block In agains from representation has neither state-descriptors, nor operators described as located "b" it. An example, taken STATE-SPACE or alternatively "a' in this a 9, puzzle there 3 by 3, cell the empty cell are space, 8 numbered block which can be slide to form certain configurations such as: 1,5 1 71 187 Operators in this and possible movements of to another, as case correspond to the empty cell from blocks are slided to occupy olace, and an example of the operators the one valid location its previous 'rules" wculd be figure(2.3). Supposing that the initial and the final, situation Is: desired configuration is: then one sequence of transformations that can 47 RYl, R,- 417, gA4 2~3: 48 g4eWOT= final produced the configuration would bet STATE-SPACE practical expressions have a sequential be explored from zi- representations of problems characteristic. previous situations. lend themselves with structures Different At to that situations can any point in time 49 we can analyze the state we are existing difference to our in, final to goal, find what is and the process reaching a solution can be composea of a concatenation the of of 50 one operations, after the other, Prcblem-Reduction: 2.2.3.2.- representations, In PROBLEM-REDUCTIOIN to taken problem can be larger or reduced the into decomposed methods simpler concerned are with that can be pursued to reach a solution. These strategies are oriented to decompose an a set of smaller components original problem which solutions might be to obtain. The elements are representation called steps one. strategies easier different which solutions imply the solution of Problem reduction into Rather explore how an original we problem, problems" "primitive the a solve of descriptions with than working deal we the problem itself. the structure of with a state. state into a new instead continuously modifying "PROBLEM situations, goal situations as was mode of of problems, consisting of Initial and operators that transform presented representation. this representation as being with in this descriotions therefore DESCRIPTORS", one into the other, "state-space' that we deal We might, one level in the in fact, higher previous think from cf the 51 previous one, and contain Droblem in a transformations in this situations. legal The decompositions descriptor specify how one problem these operato rs, of application subproblems and among we those descriptors that If the "state-space' applications of problem-reduction more easily satisf y. of a for operators generates r eached be can that can chose of problem repr esents the o perators, we Iook combinations state-space strategies by the appl ication of se t the before from, of all exp Ior ing state-space des criptor. any particular One example of this *decision tree", following Protection Association protectionsystems general can generat e a set we can its possible situations all possible transformed be might of possible subproblem descriptors. Through the into a set related possible its into problem one of are case, They are accomplished through "OPERATORS" that components. set of of hierarchy components of descriptors problem *state-space' representations as statement protection", but for reduction method c an be p roposed by the Nat iona I Fire for the analysis buildi ngs.(11) of the problem the the design of might fire In be here of fire the most simply 'fire protective measures 52 desired cost taken implies a fiqure(2.7). or performance. The elements are cifferent protective measures, as subset Subproblems of in of get for "primitives" be to have represent problems solved. different and so on until every ki nd of imp l ementation. different altern a tives of For different in Tne measures. our problem descriptor nodes. Either itself our problem, OR alternative avallable, second two, can be alternatives descriptor nodes. solve the problem, first graph rotation, as of alternatives OR w hathever 'AND* are All of tnem can th em we say then tha t as with are "OR' successors of and selec ted its There might be used only i n combination which can or alternatives described, themselves can a at ternatives of combinations co nstitute a sol ution. be alternatives which by other collection eac h into solve can that we action. The operators b r eak a protecti on measure col l ection be can and their Implementation turn, their acti ons with smaller subset some basic actions that that problems which, strategies, protection and the desired the provides that one the valid different the selection among implies othe r solve by one, alternative alternative is a win ning strategy. The succes sors have of to be satisfied problem in Ln Decision Tree jF5ZPO 7 54 to have our order to al ternative that alternati ve proceed problem solved, and therefore we say four, AND five, in t he picture then consists strategy AND*/"OR' nodes, through in has structure a this and when the stated as a when described can be soluticn its its different reaching a solution can process of be sol utions. Theorem-Proving: using predicate final be can we want to model order among hierarchi cal In THEOREM-PROVING and that synthesis of related "primitive" 2.2.3.3.- "first-order found. f ashion, similar a 'primitive representations the problems when decompose d initial to paths, of tha t reaches a set of Problem-reduction attempted (figure(2.7)), combination the of problems" which solution can be are satisfied i n our solution. As can be seen parts, and selected be must alternative AND three, a representations, situations formalism, f or example logical calculus", conditions of as a language a problem can be in which expressed 55 as valid sentences, and performed in order to find The elements the case alohabet. strings which with or in this Post symbols relevance, be "well their besides from in particular, by statements. *syntax' or the formulas" can formulas" or out as assertions be Its that or symbols the formulas" to the domain 'true' 'false* Although two of on 'rules or of Inference' to regulates how out of "we first, Il other "well in the alphabet; and that relates "well of Interest, the deduced be include part all can be made about an main parts constructed of set as new statements can part 'semantics" or formulas"; formulation, can the statements that the application of previous the legitimate formed legal formalism represents and meaningful manipulated into that of Interest. The area as to a given operations assembled called belong, Systems, always be decided by interpreting them valid or proofs representation, Production can expressions, some domain analysis can be implications, out Their combination result how dictate logical statements of our problem. deductions about in where by assigning the formed formed second, formed them a value. this representation offers the 56 advantages of the power of logical always generality, uniformity of representation and remains express our difficult is that formalization for making deductions, techniques to reach and demanded, knowledge of specific the level difficult problems in it of also to logical formalisms as the predicate calculus. 2.2. 4.- Search: For all the previous representations, our in their ved is have formulated Issue that remains problem be to sequence of operators transform a state descriptor, so or descrip t or break a problem a new statement out nference of an old control paths transformations the bounds, are still find the rules, that can anoth er that state or deduce ne.(12) to growth of branches that 0 the second to how in t o its components, Which alternatives several into terms, we once select when can be apolied, there are and how in our tree to a number can be explored within reasonable to of tiffe the main problems of search. 2.2.4.1.For the Basic Techniquest first problem, that is which 57 to alternatives next, select conventions two be can the paths established on how to explore systematically all in the solution space of a given problem. Depending on how generate at each alternatives to or in what orded we decide we proceed exploring nodes, of the graph, we can level move along the breadth or along the the paths the successor nodes depth of that extend out of the tree root. If we decide to explore all at a point, before continuing to expand them into given other levels further down, we say we conduct the search in a BREADTH-FIRST manner as shown in figure(2.8). If we decide to explore only one node at each level of the tree, and proceed doing so, each successive nodes until we reach a terminal until have explored all we for branch, or of the possible paths, we say that we conduct the search then in a DEPTH-FIRST way, see figure(2.9). BREADTH-FIRST on conventions how to DEPTH-FIRST or solution space, and depending problem, each disadvantages. through the of them has When solutions levels of the of each visit on the tree, the are nodes in a structure particular are searches of the advantages unevenly or distributed as in figure(2.8), a 58 Sl=( 1,2,) S2*( 2,3,4,) S3x( 4,) S4-( 4,5,6,) S5=( 5,) SS=( 6,1,2,) S7s( 1,8,) SO-( 1,2,9,) SOa( 1) S1=C(1112) SIe (13) Sl=( 1,2,3,) S2-( 2,3,4,) 53( 3,4,1,) 54u( 4,1,2,) PAWM&57 Z- 59 depth-first beyond alternatives have on But, Instead. unnecessary we had the other would Independent oprators in of Even small quite rapidly. the breadth-fir st use solution muc h mo re tend incre ase to genera t ors alternative 8-puzzle presented in the of positionfour -at three that produces the and paths. For operators, four sides- four -at possible paths, into extending out that a their size , like the example, -at when state-space of our initial of other into In the center applied, produce each of the cornerscan combine only two alternatives; that contain 14 moves, one alternatives different alternatives each, and generate of kind for d escr iption large number possibilities faster. trees convention s, these representations, combine with e ach case, at of lot a 2.9) state-space this search hand, when solu t ion s exist find the situations problem-solving path might would be done by breadth- firs t searches work when depth-first a shorter where tree,(figure the of levels similar level if found been the exploring time longer spend may search three that in sequences of 1,497,792 configuration 60 in figure(2.4). and with Problems with operators oossibilities can that these terms, numb er larger Iarg er expand quite reduce a easy exponential of this *ever explosion of search" of the problem structure. understanding r epre sentation of a (6), We can, problem in have to the less we know about of demands a knowledge fact, measure our ope ration, and a problem, the more we A policy for is next successive nodes suppress expansion against their altogether rules of of the a pr oblem space, the process of alternative s. branch selection can ta ke advantage characteristics of the in knowledge in branch selection and generation of of particular or a reduct ion that have to be explored can only come from embedded certain In terms of rep resentations In *pruning" of branches, combinations or of search for its solution. A which threat present which processes reduce search to a minimal that number to a non-operative alternative. The control say o f operators, the the problem, and to by proceed, *promise* (10) to decide ranking succeed; exploration of branches, by thumb , 'heuristic called Information* (10). Simple heuri stic information can reduce H to 62 substan tially all figure( 2.6), state, problem problem a the 8-puzzle tree of In space. for recognizing that without the state we had the to situation every partial tree of n ode path a nd can be evaluated the 'EVALUA TION of number FUNCTION in the figure(2.10). If besides ttese reductions, each that before its applica tion. Preventing their application results smaller each type of movement a operati on -centersidecorner- there is reverse s to blindly operators were applied the in terms of misolaced the *promise* of length of the blocks, then the (10)2 f (n)=g(n)+W(n) where number nodes in the g(n) is the path misplaced blocks, a "EST-FIRST" manner of tre e in figure (2.11). length, and can W(n) is the help us to select the and reduce the search to 63 AV4t47T )1 1 eV10 1iaer F,// 2.2.4.2.- Backtrackingl One of the exhaustive techniques for searching 44 the set of all a given problem explores systematically the solution space of It at problem, by partially expanding solutions an element or retracing its steps all It reaches a where no further elements can be aaded, or whenever the components have beer, added to form a valid result. a combinatorial technique have sequential expansion of several several the. parts, clearly In established restrictions or together - from the definition set their that taking one of of of parts is values and the stipulate what structure consists oft X2, each of values formed result. A set of ( X1, general with all "constraints' constitutes a valid Their them capable of definition this that permits tte their solutions. These are each of this by solved structure depending on some values, problem. being to amenable Problems by a previous decision to the state of try another possibility, whenever in order to point fashion, and by "backtracking" in a depth-first time, the Is (13) 'BACKTRACKING*. a possible solutions to parts, or .. **.** which .,, "selection spaces" Xn ) represents a set of possible where a particular decision or selection can 65 be made according to a of constraints Criteria - ( 00000 xi, x2, 09 ) xn In order t o expand a - Solution represented as a "vector* of ( x1, x2,. .... elemert *x1" where every ) *x2* *xn* correspond to the set of parts * X1' or to a vaid selection f rom 'X2* or respectively. *Xn* An exhaustive search values compared with 'selection spaces* until product of the all the elements in alI and, possibilities that 'Xr* out the set of all with all the set 'X2' for vectors this for *Xi have again st i ts of all the ele ments has all t hese sets, and so o n until to be explored that of t he set represents satisfy al I the simi larly result val ues the elements in have In their possible ........ x Xr, X1 x X2 x so I utions or another , The 'Cartesian Product* or restrictions the all been explored. i.e.: *X2' for possible way one considered, be for possible val ues the structure means that all to xn 000.0 length n the "XI' times set "Xn*, numbe r of in order to find the constraint valid solutions. In a BRUTE FORCE approach, what we would do, Is 66 proceed to construct each of test combinations and once constructed, these from resulting the possible complete vectors we If order to find out them against our criteria in have a valid solution. in the spaces' , *selection proceeding By point whatever the vector being succeeding in Looking at always an whether tell expanded. of the criteria At solution, being valid by whether or solution can not be expanded any point can not provide explicitly, and a valid solution is partially be extended anymore. We can not without without a a know a stopping rule that our solution space, more. generation of during the always formed, but we can solution time our these, that direction any in not guarantee that can we in of none having guarantee continuous advance towards a of can select from where we the next set, can we constraints element, contains a candidate for a valid extension of the vector, can for chances the are what are, we in sequentially we can always test the selection of values for a solution, at the values explicit consideration of all the unnecessary makes however, works backtracking way The having when solution, excludes having to to but we large sectiors explore them wait for a complete 67 vector in order to test for its validity. Backtracking can be better understood using our Stop(11) Se( 2122) S3nC 3132) S4( 4142) S5u( 51) Sea( 61) F/6e4rME graph notation: - fZI figure(2.12). At the beginring of the tree of possibilities, solution, or a node that represents our we have an initial have initial vector of length zero, as no decisions been made yet. - From this starting point, we can tree by representing construct a each of the possible selections as branches that grow out of this root. Each of the values in *Xi' would appear as a node at the end of and stand for a possible these branches selection to be added to our vector. - From each of these nodes, we can select now 68 one of into the a values In "X2' of level second in its turn, which, branches; nodes we continue branching on until from the and expand would resulting have included we all the possible selections of "Xn'. - our path towards pushing until Constructing we reach a terminal a consists solution, further then in levels down in the tree, branch, or until hit we dead a end. By convention, we can select node, out the always in a left-to-right retrace our we left branch when we and down directions, visiting the importance however, branch selection that 'stopping nodes that have criteria, and node, nodes in the following the 2,I3) The or a can move systematically across all we paths in the tree, tests available two these left-to-right, mechanism, a steos to return to a previous node. With way: {{/,4/A left to exit the next to return of manner, such that we always pick in the first branch branches out to everytime of relies backtracking In a be we a search stronger criteria for incorporates, as rules' as described before, to help reduce the number of explicitly explored. advance from With such a given node to its 69 SI-t 1,2,) S2--( 6,5,) it S3- 3,4, S4( 5,4,) we successors, that not do a%- look 10% for a subset first lef tmost member. we find we advance there. subset of *X2* are both At this as a Possible point the we extension and took now for the vaid and find out that the nodes *x21* and *x22* selections Invaid constitutes a vatid doing *xt1' shows Starting from the root, this procedure. the node first figure(2.14) In A graphics example of choices vaid the problem constraInts, and violate any of from this subset we pick out Its consequences of %~ and only the node *x23* possiblity of extending our path By so, we can see now, how a whole region of the tree, extending below consideration, the since invalid nodes, is ruled out of all the paths that go through these 70 SIS( 11) S2=C212223) S3*( 313233) S4=( 414244) h IWNALID MEES FA*M4 u . 2-N nodes would by definition be wrong. This process is at formulate the search By discovering a dead end the tree, we *preclude* or exclude from further considerations, points. during regions called *preclusion*. certain level of a out cutting paths the all below such To preclude large regions of the tree we have to our constraints analysis sequential in a way that makes such possible, and structure our solution space in a way that brings these forward violations as soon as possible. One way of doing this is to sort the spaces* by increasing different levels of with the smaller number of the tree, so that choices we have *selection along the the sets number of elements at the beginning or 71 and the larger violations occur near the root, As tree. the of certain combinations in having the fewest choices at the beginning elements, produce to tend bottom sets at the will If than preclusions of paths larger of It were done otherwise. of techniques as such to help are used 'branch merging', bound' and amount other some ordering', *Branch and Bound' incorporates to for branch selection, considerations Is against and preference, or maximize solution. modified as encounter new pro ceed minimize Bou nded our criteria for we out by to by the criteria the lower of along we can rank It now upper scale of trying to predefined selection in preferences solutions a respectively, limits is the branches of choices that can be made. increasing or branches overall acc eptable move a out in the we can decide whether or not and reduce for preference values option, Invalid an ot hers acc ording the and successors. Besides knowing if a successor or valid a *branch for solutions. work spent searching among different "branch and *preclusion' with Together continuously the Looking tree and at them we can Improve our situation, decreasing our previous bounds, drop that extend beyond out limits, effectively 72 reducing the regions of nodes that have to be considered. the that fact solution space, of elements, but a spend might that in many cases what redunaancy As solution. solutions diff erent only might t hat of them transfor mations of out properties are said or transformations, merging al I ow of regiors same pr oblems whose to us c onstruct *isomo rphi c* as tries paths merge to of into seque nces reduce our sol ution expand the results or collapse to under it loks is also called, of *clusters' space, but allow us the such to one an other. Branch non-isomorphic all new tha t share these for these equiva Ience s either before or during the and We a e quivalen t be to In So lutions *isomorph r ejetion*, or paths. rot a ti ons or t ran slations, all one S * old versio n a combinatins In definition of case the be share symmetr i es, solutions in the size of time considering diff erent lot of constitute increases the only the explotion is not recognizes hand, the other 'Branch merging", on search equivalent solutions which nevertheless to variations If possib le desired. 2.3. Conclusions: As backtracking a combination of all these provides an organized approach to techniques, exhaustive 73 even with more some cases, of that implications a that importance in the same grounds. find out Having to construct solutions in orded to exist at theoretical it bad, might not be a graceful all, but sense, Is the only choice lacks structure stlil we a more have powerful In our particular case, to Indication a larger problems explanatory not but rather using tree-searching methods, for good or theory. as It need that demands for whose from us stop take a of design the generation configurations, this criticism should an for they if way in or elegant the time, must of generating in has it frequently critiziced is solutions, enumeration be solved otherwise. which could not The for method valid stli]4 resources, memory on has the with expansion, sequertial this use a clever But in take vs. explicit enumeration of Implicit backtracking makes of problems than anyone could wait. the and solutions solution its and computers, can the use of digital the preclusion, spaces* can still "selection explosions, combinatorial bring back time, size of Increase in searches. it is, future and related development. And these realize processes that is only of the "...computerization secondary importance. of The main 74 Issues are still spatial the better understanding of configurations manipulating them. investigations..." Here (14). and seem of to our the theory of reasoning In lie the significance of 75 Chapter Two, Notest A.P., DINJENS P.J.M., J.T., THIJSSEN (1)(2) BOEKHOLT HABRAKEN N.J. "Variations: The Systematic Design of Supports", forthcomming M.I.T. Press. Cambridge Mass. 1976. (3) WEIZENBAUM J., "Computer Power and Human Reason". W.H. Freeman, San Francisco 1976. --- DREYFUS H.L., "What computers can't do". Harper & Row, N.Y. 1972. --- PAPERT S., "The (4) ARCHER, Artificial B.L., "An Overview of Intelligence the of Structure H.Dreyfus"... the of (esign Process". in "Emerging Methods In Ervi ronmental Design and M.I.T. Press, Planning", (Ed. G.T. Moore), Cambridge 1970, pp.285-307. --- RITTEL, H., for the design of an principles "Some educational system for design". part I, DMG-Newsletter Vol. 4 No. 12, Dec. 1970 pp.3-10. part II, DMG-Newsletter Vol. 5 No. 1, Jan. 1971 pp.2-11. (5) KAPLAN A., "The Conduct of Inquiry", Chandler Pub.Co., San Francisco 1964. --- McCARTHY J., "Some Philosophical Problems from the Standpoint of Artificial Inteligence". Machine Intelligence 4/1969, (Ed. Meltzer Michle). American Elsevier, N.Y., 1969 pp.463-502. and 76 A., & SIMON H., "Comouter Science as Symbols and Search". Vol. A.C.M. Comm. pp. 113-126. (6) NEWELL 19, 3, No. March 1976, M., and Infinite Machines". "Computation: Finite N.J., 1967. Prentice-Hall, Ergleicod Cliffs, (7) MINSKY (8) PIAGET J., "S truc tura i sm", Harper Torchbooks, N.Y., (9) Inculryl Empirical 1971. HARARY, NILSSON N., Methods "Problem-Solving Intelligence". McGraw-Hill, N.Y. 1971. (10) in Artificial Committee on Systems Concepts for Fire Protection FIRE NATIO NAL 1974, Nov. Structures, in PROTECTION ASSOCIATION, Boston Mass.. (11) (12) SEARCH (13) GOLOMB S. & BAUMERT L. "Backtrack Programming", J.A.C.M. Vol.12 No. 4, Oct.65, ---- pp.516-524. BITNER J.R., & REINGOLD E.M., "Backtrack Programming Tecniques", C.A.C.M. Vol.18 No. 11, Nov.75, po.651-656. ---- KNUTH 0., of Efficiency the "Estimating Programs". Vol. Computation, Mathematics of Backtrack 29, No.129, Jan*75, pp.121-136 ---- SWIFT J.D., "Isomorph Rejection in exhaustive search 77 technIques". Amer. Math. (1960), Op.195-200. Soc. Proc. Symp. Math. Appl. ---- WALKER R.L., "sAn enumerative combinatorial problems". Math. Soc. Amer. techniaue Proc. for Symp. a class ADOI. Math. 10 of 10 (1960), pp.91-94 (14) RITTEL, H., in corfigurations" "Theories of cell and Planning, Oesign Environmental in Methods Moore, M.I.T. Press, Cambri dge, 1970. Emergirg ed. G.T. -9, 79 3.- SPACE AND FUNCTION ANALYSIS. This chaoter exoands the S.A.R. methodology to formulation of their general principles of the description of spaces, the standards functional and analysis the of relations. in It describes about of the can be functions elements, their detail defined in relatlons proceeds then to present a process along alternatives, possible "state-space* and generating and how a site, a set terms of their standards positions; and that enumerates all the lines the the *problem-reduction* of for methods alternatives. 3.1.- Formulation 3.1.1.- of Standards: Site: The site constitutes place elements according the environment analysis, the or a set of spaces that define a where a environment to certain rules. At functional room first, is where the scale formed by room or an area we of a space within a function can be performed. As container for this function, standards aboLt 80 the of can be defined at two main site is called the what terms and another in ACTUAL SITE, of SITE. the FORMAL 3.1.1.1.- a of proposed as part defined be represents site of terms in or a space being of Characteristics design. under area the space existing given a consideration, can ACTUAL SITEt actual The site In terms one levels: this or MATERIAL its SPATIAL ELEMENTS. SHAPE, define their or figure (3.1a, of height these parts through 3.1b). all of spatial descriptior the adjacencles, Shaoes RELATIONS. include the components. length, the total overlappings and done is site between relations width of how The representation parts. in to rectangular rectangular Dimensions of each of them for and restricted together to form fit the and been have combinatior any figure space, total DIMENSIONS application, this and the form that areas set the SPATIAL ELEMENTS we can describe As them, such as shown containments In figure 3.1.c. As physical like MATERIAL contrapart blank walls, ELEMENTS that we can describe delimits the spatial access walls, windows, etc., thte elements, and define 81 OI~(A4~ AP.7445I A1'A//e ;/, again for F/A2t' each of them, besides its their SHAPE, TYPE, 5-.Z DIMENSIONS and RELATIONS. The actual site figure 3.2. might be thought as the bul I t 82 total space that represented, correspond in can the dimensions, shapes stands for the spatial spaces the two material and their between function. kinds of elements. FORMAL The formal site, or areas that It as a set of can be nodes that elements together and or material 3.1.1.2.- the sense of a graph notation, to spatial and contain a links that set of relations with that exist figure 3.3. SITEt on the other we might say, hand, do not exist represents at alI built space, but are conventions used in for the formulation of standards. As situations in the described case of rooms In the housing with the S.A.R. Methodology, we can 83 several functional layouts variety, furniture pieces look at their the of tend they are centered or MARGINS. elements possible arrangements. furnitu lets say a kitchen where the along group represented as the site As an actual dimensions formal site the wall No and particular made a general 'site* has statements general for example, where a of Its walls, to would be connections, we can describe elements (figure been statement its spatial 3.4a). As a a MARGIN ad]acent where kitchen furniture may layout where site, ire and appliances tend we can specify a ZONE and "W1' MARGINS figur -e 3.4. site, physical and only on one necessary in and positioned. A room, accomplished be can function ZONES make can we Through them about or can be, are, of actual the in areas can we ZONES scale, (1) "functional those functional housing the at done A system of represents in spaces, describe such schemes of agrupation, Is as in layouts. most equipment talk, along aligned space, surrounded by circulation or operation To among grouped in to be time, the spaces, living most In wall most They are, ways. certain that and notice be defined yet, about to nositioned. but we have possible layouts 84 along the wall "W1'. A zone can help POSITIONED, define what have. margin are the we If agree, statement a represent lengths be zones in possible fit such Margins can be a must would layouts, on the can system of in a in figure 3.4., the site might pieces that elements that positioned about can the zone can help us time a restriction represent at the same of kitchen then to instance, that for elements in margins, end adjacent different margins, zones and always a us define where elements and DIMENSIONS site. (figure 3.5). Zones conventions and about used to the POSITION and DIMENSIONS of represent elements 85 In one direction. A system of zones and margins can have different widths, the it in across its whole length or width, or run along one boundary of the space, cross extend middle, (figure 3.6). simply cover one part of our actual site. define To dimensions In S.A.R. concept. which length conventions on and positions the opposite direction, we can again use a A SECTOR Is a part of a can be zone and margin specified. If for instance, in our kitchen example, the wall *W1' has a window and we want to say that some of the furniture pieces must always be In front of the window, then we can break a zone into several parts, one of which has a length that corresponds to the 86 10=76202~~ length of .--.Z :, the window, and stands Ir-;;E for the sector where -.J1 OP V. I - those ecs 7,e those pieces can be positioned. (figure 3.7). A zone, as a formal construct, has only one 87 dimension: The width. length and width. A sectors can have both dimensionsi zone in an actual length from a combination of actual dimensions of the space such case, we can say that a site take can sectors, or take it where it zone from the is positioned. with its its length In and =V'k1 744'1 f 74C.3 width defined, has only one sector equal A system of a ZONE DISTRIBUTION. to itse lf. zones, sectors and margins It stands for a set of well Is called defined areas within a room, which can be used to describe general statements element on how elements layouts are acceptable. 3.1.2.- Elements: are positioned, (figure 3.8). and what 88 The elements that equipment said in the defined in assumptions or 3.1.2.1.- material we can units the of of elements used of terms their for each of and these or equipment are constitute RELATIONS any combination of can their in furniture them. have 3.1.2.2.- USE The use-space or around a physical can be the space is For each As or several one its among them. the shape of limited to rectangular them. ways, but length, that exist pieces, site element an assembled In different to define be UNITS: of the restriction elements can describe the SHAPE and DIMENSIONS of to 3.9 SPACE, pieces that Similar figures or function or As was function. layouts of defined PHYSICAL together with units a 1, furniture might be described. pieces, physical be their USE The physical element perform a of chapter can certain RESTRICTIONS actual of it. Elements the pieces of the oossible terms UNITS the needed to are while performing PHYSICAL are shown In figure physical units, for each of them we have width and height. SPACE: is the space that Is needed unit to be able to use we need around a table In it. order along Use-space to sit 89 down, the free soace needed free open to arawers, for equipment parts Use-space several The lines for rectangles themselves to in several form physical or the space that should be left to move around. be described also Physical certain 3.9b pieces. complex as one or delimit use-soace They can be related use-spaces, or among related to or adjacencies. RESTRICTIONS: units and ways. For a of its physical figure units through position, 3.1.2.3.- in can the space rectangular shapes with DIMENSIONS and RELATIONS. dotted specific needed to swing doors open, use-spaces can be given piece of restricted furniture, certain parts or use-soaces could be overlaped by 90 parts different of furniture how Independent on Is elements through stated (i.e. definition of standards elements), we pieces. can specify at relation the the next 3.1.3.the level between two In the issue the piece can or cannot be overlapped physical units or use-spaces of other elements. with this restriction minimal access side is furniture of each piece, if we can define between Relation for each either by Together piece If a required and by how much. 3.1.3.- Relations between elementst Elements can be in several located related to one another ways. Relations between elements, called here 'element 91 constraints*, relative positions in theory that three variations of one- - ADJACENCY. - OVERLAPPING. - CONTAINMENT. express and satisfied In a example, specification can be described certain functional whether *element-2*, *element-2* *element-3*, or and that can only three -In relate two elements conditions that layout. They can "element-1* whether about be fact, are supported: relations Element constraints time any this application, however, In tested. are should should be *element-1* at should regulate, be adjacent contained should the be for to by overlap 92 "element-4*. be expressed Relations can we can of the RelatIonst Simple and Compound Element 3.1.3.1.- statements constraints in 'simple' one in in two ways, list a series of form 'ELEMENT-RELATION-ELEMENT*, as was done in the last paragraph. This clear is quit e straightforwards and and simple. formulation of all important, But, relat ions to or "Compound' it restricts the equally of lists long the we way we t tis option, can be relations logical connectives denied by exoressedt are AND' simple 'OR*, With them 'NOT'. there relations about think is a *compound' option. second way in constraints. linked by constraints and one select might anothe r if we could have the To include which relations to the way elements, constraint over hand, the other havin g to be satisfied, constraints which rarely corresponds between on in this sense of capable we can formulate being relatlors such as: SI.. "*element-1* shou Id never be adjacent be adjacent to cr *element-2* *element-4*, to but "element-1', or 'element-3* *element-2* should neither should 93 element-3* be contained by "element-4'". Compound relations have of Instead (1), notation to obscure their advantage. bisarre sufficiently perhaps a special saying: element-1 relation element-2 as where, inside the relation and by it. element-2) element-1 (relation related are statements now as: expressed first simple before, said we So that a pair of then the parenthesis list of we elements define that are the statement: ***element-1* should be adjacent to *element-2*" is turned into the statement (should be-adjacentto *element-i* 'element-2*) or for simplicity e2) The reasons for although clearer, user, if (adjacent el perhaps we think that themselves other this not as statements connected by *and". elements: statement-i and say: "'statement-1' and justifiable might can the case would be The relation "and' statement-2. So if 'statement-2'", become in terms of in a relation elements relations, inversion we say be a in for two has we want two to 94 (AND statement-I statement-2) format with the same relations we if be each statement and, simple substitute as (adjacent el e2) (adjacent e2 e3) then we have of "AND', a them relation binary nested binar the two elements, with is a relation that by formed relatLor or a compound relation e2 e3)) ei e2)(adjacent (adjacent (AND 'ADJACENT', each with tte format: ei e2)) (relationi (relation2 et e2)(relation3 (relationi elementi element2) element Compound blocks, the relatlons, hierarchy of can complx relations nested binary other together with the be unary like: at building "CONTAINS' binary the relations relation as use "OVERLAP* and 'ADJACENT", which relations as bottom to a and 'OR', "AND* "NOT*, of form more (52) Which corresponds to our previous statement 51. A can be relation graphical visualyzed representation in on each node, terms and for of of a each these relatiors tree, which shows a node two or one 95 (C4U*x-WNA=' [R 52:(AN E54-) ev e )) (40- 14sWr A3 J4)) (coornavrp as elements successors respectively. , would The statements look help explain the Or show them as of like the or binary in (5-1) unary , and following figure 3.12, the system of nature of if we don't break our a Jj) E0 j64 line statement, in 04Z) which can oarenthesis. blocks, building it would relations look but like the tree In figure 3.13. Simple relations several satisfy, to that alternatives combination of compound Both list one constraints of represent can be accepted depending on tthe connectives that Statements satisfled. represent constraints iInked by we use. *AND' have to be both relation-cne and relation-two, represent 96 F/#A(Awf,/z 11 ,-6(=44:23 -"r constraints that we want 5W4) (A#sii satisfied in a 54} functional layout. 97 Our list of a to then equivalent simple of formulation first statements statements be would linked only by AND's. can them have In present the same at time. by preceded Statements their then t he other cannot be the two satisfied, of one we if imply that Exclusive OR'S layout. the final two for that either or t he two, or both of we can satisfy statementts, impl y OR's Inclusive *exclusive-or". or *Inclusive-or* as ways: two in considered be connectives, 0R' linker by Statements that are "NOT * simply reverse relations. final For our example as considered alternatives accepted as if then, there (figure 3.14) which combinations valid 3. 11, OR's inclusive possible figure in If appear the OROs w ould are be 7 be would in the final arrangement. the OR's If three 3.1.4.The between then only the first would be considered acceptable. lists through were exclusive, location of *POSITION an Position of RULES'. elements in elements in the the si te. site Is defined These rules specify a relation element and a site where it can be positioned. 98 I-- Z - (Amiqret'r (r"as0/r 4:- (r (*4M £1 El4) 54 Z 54) r (gm~mW 6,-(9 A2~egur 46*) F3 51 5i ff4) £14) (4404rrrr-5 Position rules can relate elements at different levels. Elements can be the space that forms the actual site, to located *room-x', the site simply In they can be located In zones within the room, *zone-1*, o r they can be positioned in sectors within the zones, Similar pos ition rules to can rul es, or 'compound* the be relations expressed pcsition *sect o between rules. elements, *simple r as r-a". The simple position re lation bet ween an element and its site would be: elementi is positioned in sitel, or (PUTON elementi sitel) In the same notation that we used fo r element 99 constraints. The compound position rules wii11 use connectives or element, several locations alternative rules that define several the to form nested position and *NOTO 'OR* 'AND* again for an alternative elements that a space in zoredw MAPerr.,0a ;E4 4zZxwV)) 7 ~ (F//~4r('{/7 2/6/4 3/F the site can contaln, like the rule in ano its tree representation 3.1.4.1.- figure 3. 15 (figure 3 .16). Definition and Levels of Exp ansion of Position Rules: rules Even though both different from are express relations, element constraints position In several ways: - relating First, elements they relate to elements. elements to site, vs. 10/ 4W --- Sw 4/%(p/;'s za#a463M ) , 4/S (PrQ/ o - Second, zones and def ine relations three I evels. these blocks*, (4 4Z element constraints, To correspond position rule this 6o' can use any using the dlginr with can (P.R.) these of combination same terms as in z1=17) to describe a position standard as: generate to as a space between elements and site at any of A compound P.R. 'bulIoing structured simple a sectors, urtv a site for r440a) all rule, the possible however, layouts that we have to know the 101 *ell positions that detailed the site. of way In a sImilar constraints, element and enumeration, for to the case made compound there is a conflict between how much should we information For *zonel*. wittIn take *spacel*, and positions have to be defined at the most all level can *e2 the positions that enumeration, can actually have within the give way a in think we permit to standard positional about constraints. rule, For a position in the conflict this Each defined P.R. the possible at any p.r.'s level of the that can be Implies all relating its element to each of the parts level If, Z2, be solved following way: - has one can formed the that site by site down. for example, a site has two zones: Zi and then the p.r.: (PUTON el sitel) Implies both: (PUTON el Zi) (PUTON ei Z2) - vs. specifically, As the rule I.e. is defined generally, Z1 or Z2, i.e. site, then we can assume that 102 either of to link them to with an *exclusive-or*, proceed can the two positions is valid, and we form the new rule: ~V which simply says, if we want positioned in any *site1* then "e1" 'site1' zone, two in turn, has sector s, for example: Z1 s-/ //rdA (lg (r,--04/ S2, and Z2 has S3 and S4, in valid *S2*, or In "S3*, parts of each has Si and )) / $$9) 5f 54 ))) that or "ei' *e1* In turn or In *S1', *S4*, into: would all 'et* be positions. By general we *e1* X then the rule would which again assumes in S/) C/ 6C two of the as a valid solution. If accepted be would positioned *e1* can possible level avoid automatically expanding a rule from to its constituents in the following having to define one levels, each and everyone of the positions that an element can have. Our original 103 rule at the section, can be expanded then beginning of the (Q0tc( PCX (0 5/ 53)f E/ %3) (P-Q {o A position standard can be expressed also as: where if e* and *e2. e1* say: any place of positioned the It the site, as our reans: "put all as long in sector "S2"", we did before, all for all stands the expression element el site) (PUTON e2 site) lets the elements *e2* Is in not and by a simitar procedure as e+ Is first elements in the problem definition: (PUTON elements, expanded into 104 linked then we because *AND*, by want all sat isf led ( and we which sectors Si and have f or S2 #(/7:w Az 6/!) the resulting rules a site with two zones, ZI and Z2, in ZI, and only one sector S3 t'Pri~u f/ (w (Plr# S )) 6/5)) 5Z ,i)=) two In Z2, 105 intot would be converteo 3.1.4.2.- Position A third difference from results constraints Rules between the p.r.'s element and two or more having site both by shared Is sector one overlapping zones where Over lapping Zonest in areas. position rules were If defined at problem, as sectors, this would represent no of to we would have in each sector. levels, level the If we however, which element then always always simple and at rules compound allow elements that the list all we have to define go several a way to find out goes where. For a example, problem following the with def inition: Site: 21 with Si Elements: el and and Si, Z2 with S2 and S3, e2, (Qa (Pt!EQW (pc cm~ss Position Rule: can -be positioned does not in sector S2. EZ zz)1 '3 Zi/)) ) 3 ZZ ))) which elements If we think of P.R.'s as define MITLbraries Document Services Room 14-0551 77 Massachusetts Avenue Cambridge, MA 02139 Ph: 617.253.2800 Email: docs@mit.edu http://libraries.mit.edu/docs DISCLAIMER Page as been ommitted due to a pagination error by the author. 107 subsets of describing elements that can be positioned in a site then, Zi can have (el or e2 or e3) Z2 (e2 or e3) can have and if (AV we expand the (er X/ .5/1) (/f 4/'4 rule 41P (Ol its sector SO (Par S/) (ar £Z 52,) (m (pgtog(CX (P17 5 )/ (P4r [ 4Z (e into s6)) (4r5 Sz z (pr1 55 ))) defini tion: then the subsets Si different can have for each sector would (e1,e2,e3) S21 can have (e1,e2,e3) from ZI S2 can have (e2,e3) from Z2 S3 can have (e2,e3) where S2 subsets bet is of undefined because elements that can it has be positioned two in 108 it. these Having zones. If both zones Zi and Z2, defines the only be those that In S2 can elements can then the decide a elements in overlapping position to how on convention we however, subsets, Intersection of appear In S21 and S22 position rules for S2. S21 = (eie2,e3) = S22 (e2,e3) S2 = (e2,e3) (4' (#/T (4ire' ,E/ S/) ( (e~ (0/ 52 )) 50 5Z 5 (Pgen4W / 5 )) 3 )))) ,(parNOw(fss - S-4 2 3 (#2(PCgvx and the position rul e would bet To expand can proceed overlapping Intersection as then sector, of the rules of we we rules, di d have and Intersecting we 3.1.4.1. but for each in to zones, check select first for the those positions that 109 satisfy this test. system 3.2.- Standards: A set of of elements to the spatial The functional parts of both site relations that by defined a and relations. The elements correspond correspond that regulate relative rules standards is regulate and first, locations, furniture pieces. to element constraints and position second, absolute locations of an element in the site. In a more formal consists of manner, a set of standards a 4-tuplet S, E, R, P where: parts that 'S* form - *E' Is the set of - "R" Is the set of is the form *P* the set of compound, between elements - actual and formal, the environment. parts that use-space, spatial, spatial, physical-units and furniture pieces. desired relations, simple or in 'E*. Is the set of desired positions, simple or compound, that relate elements In "E* ~ with elements in 110 "s*. A set of stanoards L graphs , which are elements In *S* that to relations correspond the of in *R*, possible to one of correspond together with positions in *P*, in graphs which one of enumerated. a In or "P*, L, the by to edges or both. varlant, furniture the desired for a given contains all 'R' the where the desired combinations of To evaluate a set of L, linked and are or layout, A functional one formed by nodes that correspond or both; or "E", possible of set a implies is links relations combinations of S,E,R,P. standards, functional the layouts, subset L* has to of be 111 Enumeration: 3.3.- a do a have not layouts functional Is a valid or an layout a when membership. We we but have, members, of list have a are the poss ible what priori that a standard can Judging for instead a criteria but we have know we do not the set L* For for criteria invalid furniture arrangement. To enumerate L* then, we have which qualify layouts according to t his criteria. From the possible graphs, possible all graphs our "solution space*, the variations that To do solution-space might reject, we need into different equal ly Not analized that rules ( figure 3.17) th e points Importance, to the same comb inations conf iguration. and rules these having situation where all same where solutlors need we the partition as soon as po ssible, "chunks" that do not contain any solution at all. the to extract we have "chunks" important, of L set the that rules functional as L*. of are members this, and exist, L in all construct to of variations mean in the design therefore, and Ievel would detall, of this all having have criteria have checking criteria a to be all the on each 112 express Wh en these rules can be defined, we can We can state, necessary. a are systematical ly worth are layouts developed, we can construct needed more next, at. looking of L+ members look for the if know to order in satisfi ed of time, and check that some conditions the at one layouts of detail, levels At different becomes this terms of some conditions. its validity In layout, or be, the possibility them, through We can might case the whenever space this of port ions solution-space. whatever reconstruct construct or entire our the structure of the m through We can to evaluate a standard. figure 3.3.1.- 3.171 Overview: The S.A.R. relations, and parts formulation of standards, with Its provides a way for expressing these rules. sequentially by variants furniture of The generation can a solution or constructing carried on * graph", where we add one element to the site according the position satisfaction rules, of the positioned elements. and we element check at be each constraints step between to tte the 113 Pr/60/W 3,/7 At as a carried on of the most simple layouts 'depth-first* this generation can the search, and represented be can level, be enumeration as a 'State-Space* model where: - the graph being constructed represent our *state-descrlptor*, - the *state-operators*, position and rules constitute the 114 - the relations against criteria which between states elements are the tested, as shown in are .44 9, ' figure 3.18: figure 3.18: At a higher can be used to level, decompose subproblems which alternative combination position rules, with can make of the compound the problem the search Position can be considered as several subproblems descriptions, as shown figure 3.3.2.- in po i nto expressed sepa as figure 3.19: 3.19t Description of rules different s impler. rules a sition the process: Each in compound rate problem state-space 115 P01=4N 41 -C (AkP (ac (02 (Wf '-5- v A-1! ()0/r C Z/,9 zz)V (Q4' (c (A (41r 4 z) A/11 At 4/ 0ftz /l cZ P/f P 0,477M!W3 I (A!T~ %3 A (foPr z/0 (Rir C/) A76245 511V 4tC4r6 $~9 The process task of with description of different for levels this hierarchy will outline of in a search of be down of the smaller complexity. done, first The in a the problem reduction steos, and second, functional detailed example. will breaks exhaustive enumeration into a hierarchy problems quick for generating L* layouts presented through Generalizations and- definition of be made along the way as it a terms becomes necessary. 3.3.2.1. Problem reductiont The solution space of be constructed through the rulest - GENERATION RULES a functional application of standard, can two kind of 116 - TEST RULES partition the solution space into subsets that may contain TEST rules solutions. through check the existence of solutions *operators' *constraint and systematically expand and preclude regions of In and Generate GENERATION. by produced those partitions Test, or alternatives produce rules GENERATION criteria', the solution space. Generate corresponding to are rules of kinds, different two levels, simple and compound, the two that oosition rules can have: - TRUTH TABLE, and - PERMUTATION OF ELEMENTS For compound position rules, we can explore alternative position that are acceptable for different element (figure 3.19) through the construction of TABLE, as explained in 3.3.2.2.. rules, we can explore the different can the take within OF ELEMENTS, For an TRUTH a simple position locations an element its zone or sector through the PERMUTATICN or the variations in absolute positions. Test rules are also of - POSITI8NAL, - DIMENSIONAL and two different kinds: 117 These are operations that check DIMENSION of elements in site, a Positional Tests are - EVALUATION OF POSITION - EVALUATION OF ELEMENT EVALUATION tested by OF RULES, Tests - ZONE - SECTOR DIMENSION - MARGIN DIMENSION as elements include the size of a margir, tests fort an element to be against against the both dimensional CIRCULATION between by In a certain zone, and or remaining length both the oositional, elements absolute margins in the position if or it can be defined by relative position if assigned to be through the use-spaces the the as in MARGIN DIMENSION. site. It can be defined either assigned by CONSTRAINTS. in SECTOR DIMENSION, and against Important, is can be tested DIMENSIONS length and width of An the CONSTRAINTS of a zone, as in ZONE DIMENSION, of a sector, and the EVALUATION OF ELEMENT and check constraint a POSITION RULES, and relative oositiors Dimensional width by regulated thet Absolute positions of can be as standard. functional the the POSITION and in a given different layout. 118 a preference order between These operations have can attempt not a zone until elements in of permutation we Instance, For themselves. we know what are the position rules that assign such elements to try, to dimensions of find out them or if the elements against zones that location can be, Only then not. constructing hold the all that the overall in fact, we can permute will these different is broken down, check to and sectors, occupied by we kncw elements check and valid dimensions, arrangements, that the the margins elements in the adjointing zones, and pattern Is respected. circulation The different task decide elements we relative positions are being satisfied, that can to locations are known, we have can have valid positions and while zone. before doing any permutations or changes. Once such the location of valid alternative first what the rules are compound, we have position these If a levels into which are then in order of the Importance: TABLE. 1.- TRUTH 2.- EVALUATION OF 3.- PRECLUSION. 4.- ZONE DIMENSION. 5.- SECTOR DIMENSION. enumeration POSITION RULES. 119 6.- PERMUTATION OF ELEMENTS. 7.- MARGIN DIMENSION. 8.- EVALUATION 9.- CIRCULATION. corresponding "'1.v CONSTRAINTS. OF ELEMENT to expansion the pruning or P###w"1 4 Ive PWINAr M941/Nt AT#17N operations as shown in figure 3.20. fig ure 3.20: EXA MPLE: The how these possible following example operations layouts will interact for the standard: to be used to enumerate describe all the 120 SITE: ELEMENTS: R EL ATI ONS: ( Oea'$/4' AfC pWeM/ Z (4a4sNe CZ5< eM/r Z2) (0e zz) f I gure 3. 21: {4VO (4/vr' 4 The first operation to be POSITIONS: enumeration process functional applied to solution layouts that first partition that the all the a given standard can we can make compound rule. we can have, then to the Implicit in a corresponds alternative position rules that this, for stands space possible To do start would be then: the If z Truth Table: 3.3.2.1. the 6#1r V /3P 4(44D (sm (g are consider each *building block" In a compound rule, as a simple relation that is or is not satisfied in different alternative position rules. To each simple relation, TRUE or FALSE, according to whether or not have these positions lets say we can assign a *value*, satisfied in solution space that we want to explore. the we decide region to of the 121 All the these relations different combinations can have, represent position rules. level of actual layout, we the pursued among al I the subdivisions a solution space that can be made out of Without the at general having explored yet any what alternatives should be first decide that values of different permited position the by rules. Compound depe nding simp le position rule, be TRUE represents a valid or FALSE or the combinat ion of val ues whether on can statements an for each InvaiId posi tion rule. The alte rnative assigns TRUE that the simple rules of ch, elf< F/667 the solution space into of values can be expressed then combinations by a TRUTH TABLE, each of subdivision 0 V FALSE values to the possible c.ombinations. all in or // c c 01z 122 figure 3.22: In our examole these possibilities. the different Truth or enumerate all 0 to in 31 there binary. be can simple rule a with True o r As such, there are values, simple relations there relation with So a number that be As can from rules, large compound probl em*, *counting they 1 values, 0 or statement. and for added to the compound, two False, r epresents each column alternatives from 2 to 2. insofar "oinary counters" Tables are figure 3.22, in seen cross that as columns alternatives for a compound the or 1, and for each r ow. values "count* is single rule appear as a possibilities appear along alternative of by and the Truth Table side, take the values Tru e or False, 0 can that left matrix where each represented by a there can be 32 3. 22), The position rul e is represented tree on the a horizontal row (figure each new Is, increases the number for one simple of relatIon for a compound relation with two are four values, three simple relations for there are and so on; running into the enumeration or compound a 8 values, *counting* of large numbers very easy. For the time being, in mind but no this problem has solution has been Implemented been kept to reduce 123 to the direct combinations starting such as likely * in POSITION figure the of some problem in the definite relation for oartitions of simple OF that compound in terest us. tha t relations have a out these find the TRUTH connect ives that in sense. In expressed together - the EVALUATE each of the columns logical As have a out these find Position Rulest for the compourd statement. To value alternatives, we tie rules, EVALU ATION in valid combinations are rules These are combinations TABLE those towa rds RULES. Only TRUE How to 3.22. 3.3.2.2. Evaluation of position be first valid co mb natior, from the is combinations valid values to produce valid Positi on our 'counting" column of assignment must could alternatives. One possibility this generation of simole 3.1.3.2, the into compound relations statements meaning: for each, have to be AND*, *true' ,1a the two to have 4 / element s in the the whol e relation 124 evaluated to "true*, for - elements have each exclusive *0R*, to be either one the two of "true' to make the compound statement *true". both for each Inclusive elements being 'OR', one produces *true' of two the or a *true" compound element makes relation. *true* relation these simple tables. are nested 'false* a and viceversa. *ANO,*0R*,'NOT*, to a for each 'NOT*, - in compound are evaluated When several statements, then according *AND*,*OR' or we first find 'NOT's out the 125 values "lower" relations the hierarchy, in the the resulting values as values of then the the for next the elements relation up, and continue doing so until and relation, final have the whole pass for we reach statement 4 JoeC <'v4 as shown evaluated, in figure 3.23. figure 3.23: In of example to in figure figure 3.21, rule standard the position rules that evaluate 26,27,28,30,31 and columns: 3.24. Only these relations represent valid position our truth-table generated are only *true" shown the at the side of these combinations make any sense 32, combinations of alternatives left for to of the the table, continue as simple compound and only exploring 126 for possible in the columns root, If we branches going out layouts. different matrix as we can preclude /0 then from further 3 4 5, 7 O/ the regions that / / extend down the tree our consideration all r303/32 524 25627 / 0 / ' 00 011 _VW74 of 89 /'///2/3/4/5//8/.92ZI222 --A/ of think those paths. figure 3.24: Through compcund relations we into problem elements. the different possible Through the assignment of for each problem. Through the evaluation of considered valid can that can be made all decide which of we truth values, the possible decompositions a locations for the explore we can decompose can these the alternative positions values should be and continued being explored. 3.3.2.3. Preclusion of repeating elements: 127 across manner look different position alternatives, we would the compound as 26, branch now Into a left-to-right example in In our Moving statement true. evaluated to be This statement and the rest that have passed now checked are our previous test, all of PRECLUSION for reoeating elements. As can be better seen In brach 30, little ahead, there are s ome cases when compound element, true, to evaluated be to here *bed*, In positions for one element can element An Even be not though evaluated to makes no sense when The test different and Z zone environment: assign but zone Z2. in positions valid of accepted. in two places at the same this the This repetition compound Inter preted as a real for rules can times the same two cannot be, obviously, TRUE, a lumping branches time. statement position rule. with repeating 4 S, /4, /00, 7Z4 elements, would preclude then columns, or branchesj30 and 32, from further considerations and reduce the search for functional In layouts to branches,1 26,27,28 and figure 3.25. figure 3.25: 3.3.2.4. Zone Dimensions: 31, as shown 129 / (PW "-- '6 34 ' After !V the site position The possible spaces now of in first elements is assigned to only one of tte the environment, this that at width- the least one of the Is equal or check. assignment Is correct. zones or If, of sectors element. dimensions larger to the width of is going to be positioned, and equal width of both zone or and adjoining margin. element and elements for Instance, allowed to end in margins, then we have only to and we have a rule that defines so, and second, is should be considered. site position can be positioned. is an agreement on how the dimensions there it there of repeating elements, there is only Elements can be positioned in if, k II combination where each element dimensionwise if one selecting l/ 1Z 1 /;' +fl rules, valid and without in 3/.YZ // :44 (4tgi-) one part 17 .-- ru ) are to check now -length or the zone where smaller than 129 figure For suficclently 3.26: our small example, {j MM-449AM F test as shown In figure the Zi and four is passed by al I the 4/M4W#N4'6 branches zone to contain any of any position. So this -olow4w7# both 3.27. Z2 elements are in remaining 13o figure 3.27: Sector DimensIonst 3.3.2.5. Valid positions the other along have to be ch ecked in zones dimension: Elemen ts, length. be lo cated in zones Zi and Z2, for branche s 31, w Ithout repetition now we know, can 26, 27,28 and and fiting within t he wid th of both zones and margins. test The the for to assigned elements check s Sector Dimensions, al so a zone can fit I engt h, or along the length of the sector where if all along its have they been assigned if this would have been the case. Sector Dimensions checks depending on how they are or widths, the of such part of figure the our zone or sector, sum of lengths between It is widths or all length of all than the corresponding dimension of the difference is occupied by an empty length or width. As a convention, this space is entity. the of 3.28: smaller space with that exceed lengths site. the When is positioned, In a zone/sector does not elements elements that the sum of treated not broken down into several elements but appears as one unit as one empty spaces that can be 131 In changed in position but this Under further expansion, 7 RKECM.0N ~~a# ,V'rMS P/w- I loi keeps always its dimension. test, but branch,31 all is excluded from rest continue as valid the 23 T,6M I'A'- 132 options where furniture layouts might figure 3.29: .1 * exist. 133 3.3.2.7. Permutation of Elements: we have SORTED each element before, said these options, as was For each of in the rule to only one valid the site. and possible space in of assignment this For branch 26, to elements site would bet Z2 with closet and desk. Z4 with bed. interesting two present sort This characteristics: into produces a CLASS of furniture 1.- It 2.- It permits sorting the constraints that can be used for constraints LOCAL or GLOBAL element layouts. pruning criteria. We know that as I.- already a valid be represented This as: layout satisfies one fit in they have been assigned, and the only sector If that the width of they also fit the zones where length the of that each zone has. we forget for a furniture variant. each position alternative moment that elements the zones can switch positions, we can say we found a have we which schematically can furniture layout, elements Its rule, and sectors are margins, is without considering that concerned, far as zones element If can vary on its have already the other hand, location in within we accept the zone, we 134 can say then We layouts. space where several rule of and indeed, found, have found have that we a a region in arrangements share the furniture of CLASS the solution position same assigred to the same sector are validly or zones a site. The different arrangements formed the by permutations Zone Z2 can be, each zone. for of the element are CLASS this In locations in Instance, either: and Zone Z4 can be either: r--- 1~10 these The combination of equivalent several generates same elements in EQUIVALENCE different arrangements that have the constitute the same zones, and that of CLASS locations, furniture layouts and in terms of the relation *position*. Exploring *top-to-bottom*, we our solution have partitioned from space the set of all 135 are yet positioning any exponential explotion putting the bed in the put we reduce at all, corner of lower tr led then Z2 to so on, for and Z4, the we star ted could have had, had we the upper corner of in desk the piece furniture our arrangements *merging" that can be tested without 2.2.4.2) *re. branches into By by similarities. grouped layouts where CLASSES into EQUIVALENCE layouts possible each possible combination. have we class, terms of our class by generate all class would be generated 1.- The elements In branch by the in our that can layouts. the 26, equivalence foil owing representation: *state-descriptor' plus the positioned model state and a series of rule s example or t ree, we representation, the equivalent furniture For our e quivalence the permu tatlons of We build a combinatorial State-Space an Initial each in to construct now all the site. elements in layouts enumerate the To Is site formal the elements. 2.- The Initial 3.- The goal is t he empty site. state state is the site all with the positioned. 4.- The state-operators are the list of simple 136 prosition column with rules that true* values £3- 4 ) Where elements llq c the operators simply state that a can be -jY F- formed by the In zones Z2 and Z4. should be 4 A11, S ZE2 be the Truth Table we are exoloring: V4 should in two possible arrangements of That formed by two elements El a desk or a closet, and layout an arrangement in Z+ and E2 either of which that an arrangement for Z2 137 kqA A9 Ai6O~ 3~33: SA 138 should be either formed by two elements E3 and E4 El following 4 figure can be a Ded or an empty soace. Applying these first which then E2 layouts (3.34). then operators to the E3 as shown and then in E4, initial we the bottom of layout, generate the our tree In 139 way, elements however, zone. same within the F/Mc assigned to can The be bed piece which as: smaller than corresponds 09 II th at and ther is a decide on any of layout, and f $3 change the operators E3 and E4 The elements fit exactly know we OR [JemI~ F76OME (Z4), to its four 90 In the furniture appear to positions its site Therefore we can remaining empty space. these be 33 Z4 is could example degrees rotations. From SECTOR DIMENSIONS, this same differently positioned for the in positIoned here always are Elements in zone Z2, Into desk and closet, on the other hand, therefore they can only be rotated 140 180 degrees, which keeps the zone tree like their dimensions along flIB0J 0 ir/Ajj t he same 9//,El O H -~ ~E liDo ~ changing the operators El and E2 Into produce a Which could the one Qqeierating partially represented 64 possible combinatorial in the layouts with sirrilar following oicture, positions. 141 2.- Classes of sort the constraints. arrangement element If furniture layouts constraints into help Global we think of a furnitdre layout as formed of different then we can break arrangements down our us or also Local a at room the previous zone-sector level, State-Space representation into the following model: 142 room levelt - state descriptor = same, - Initial - goal - state = same, state = same, = state-operators assignment of zone arrangements generated by: zone level: - state descrlptor = Z2, - Initial state = empty Z2, - goal state = complete Z2, - state-operators = assignment Z4Z A764 4Ai;a elements to zones as? zone level? - state descriptor = Z4, - Initial - goal state = complete Z4, state = empty Z4, of 143 - state-operators Zr - = assgnm1e"t I of - Iw~ I74J34*u elements to zone as: woul d which ,OAC produce the following room 404O~ le state-operators: In apply one of the generation we would the operators at the room proceed level, first i.e.: to 144 ROOM - but in out first Z2, order Z to do this we would an arrangement at this operator, therefore we zone Z4 which can be used as to have construct o at which produce: the zone by we can apply to 3-40f) to get: {90,d4) level: (344 (1oe) form: and continue with ZZ In with: It a applying the operators that find to have a (5-4of similar way, first 145 9 .1e 165A4'b 4 and then:(3,4'I) E0 to form: This descriptions, operators same for of representation where the result of equivalence class, and one search oroduces the level up, permits us the next search one State-Space NESTED produces the to the sort 146 in the following way: constraints A CONSTRAINT GLOBAL in elements relates zones or sectors, different same the In A LOCAL CONSTRAINT relates elements zone or sector. As wiil first the at room State-Space relations that apply between those relations third help include those while would include for and constraint th-e element Z4. These sorted DIMENSIONS Z2 in that apply to element OF criteria for elements In Z2 and Z4, State-Space those relations that positions in can would level the second State-Space in constraints the EVALUATION section the constraint (3 .3.2.9), CONSTRAINTS ELEMENT the be seen in and CIRCULATION us prune constraints, Element branches are In with the remaining tests ttat exploration our MMARGIN of the Solution Space. 3.3.2.8. Margin Dimensions: When two or more zones share them, might are the elements larger than the margin between can be positione in each zone, that overlap portions of a the margin width of As the functior of if their dimenslors the zone. a margin is precisely to allow the position of elements with different widths, when a furniture piece extends beyond the width of its zone It 147 the margin. occupies a portion of common conflicts margin, is the elements fit, elements have to we check overlapping this If For overlapping overlap. elements the then both margin the than larger is the sum of if a than the margin then or equal the sum if occurr: end within zones opposite might smaller overlappings dimensions in two elements When permited Is or not. the generation of the As room, zone levels, against previous conflict that stops branch In would the search possible would pass it continue tree, only our search all passlog the tests the of arrangements two way down the 64 we case in conflict any a further. in branches 26,27,28 all withou Is there from going any for example, the test while pass if see to dimensions margin the test arrangements 7 assign an or sector, everytime we its position, we element to of layouts proceeds at any to the bottom of our EVALUATION OF and CONSTRAINTS CIRCULATION. 3.3.2.9. Evaluation of Element Constraints: Element constraints, expressed by compound like statements which can be TRUE FALSE, depending on particular combinations of values in rules, however, their simple components. the assignment of Different are roles, position or TRUE-FALSE from position these values Is not done MITLibraries Document Services Room 14-0551 77 Massachusetts Avenue Cambridge, MA 02139 Ph: 617.253.2800 Email: docs@mit.edu http://libraries.mit.edu/docs DISCLAIMER Pagelas been ommitted due to a pagination error by the author. 149. through a *blnary counter', alternatives to explore, buth actual, present, re Iat Ions. ,Wq- or a through decision the on which evaluation of 150 defined elerent constraint The evaluation like done bottom-up, that statements as in If the when all to wait for from Is compound simpler so to relations from the compound the relations, however, can be evaluated only values are set a complete a that but series of each for the results branch 26, (OR is State-Space as was said representation In any other that the relation zone to element constraintst this would mean (adjacent desk bed) (adjkanti have nested the constraints evaluate layout a if . or sector arrang ( AND then we have or FALSE, to TRUE layout, can we without having to wait In constraints element (3.25). figure representations, we can break before, the is set to FALSE. the evaluation of position rules, through generated of among satisfied not is the relation present elements then is it TRUE, if a the relations have a value it does, If not. then the site some relations to other existing elements either satisfies or is positioned in an element If be &ak ch'set )) broken down into the following 151 room: (441 (4,(t'g*;#n d't/k d"S9) z one Z 2 117011V zone Z4 We 'gg4 A4') can do f/ekwe transformations the this by applying 7 ~-e/,#77ffi shown in the next tables: tree which reduce our into several that have each trees, at to be satisfied do not possible satisfy layouts to of the the constraints element to the relation one corresponding pruning away all that zerieZ level of the site. in those combinations constraints 12, from the original 64 that we if reducing and there were no could have 26 branch the further tests had in figure 152 153 0 0U 3.39. In produces this valid case the State-Space of zone Z2 always arrangements because its two elements are 154 and always adjacent, When adjacency bed-desk. the case of not and that Z4 in constraint, over it search, when we don't have still a relation of like positioned, branc 26, then we do assumed that the is the element constraints. values long as we can have 'true' As at comes the a piece is not have priority position rules however, preclusion the chair, in our example evaluate with our X2 between level, global therefore we stop. True we proceed constraints, problems because the desk *adjacent' EUl In: to the closet as after Therefore, check blocks the access to its use space. we have to checking element constraints for CIRCULATION. 3.3.2.10 Circulation: we positioned has to have Each element its use element space. that is IF a spatial in a zone, then we treat it as any access is other element, and specify the relations have with the If to defined as this located acces furniture Pieces it will the circulation that it should as access serve. Is defined simply 155 use space withour any particular specification, every to then we have to check that there Is a chain of use-spaces, or be can access this which through spaces leftover solved. last This we position a Everytime path. If we think of the a finding graph, furniture piece we check for this a path precisely be a that spaces use-spaces, margins or leftover are linked bh a certain that minimum. us to reach around, and that should allow we can walk each piece of of adjacencles furniture the room. in Finding a spanning tree for a graph problem solved to has path of series spaces links the of that goues through some of tree A spanning that graph. its nodes. Our circulation of but visits all for a problem then path is circulation a as constructed being layout finding a "spanning-tree' is case. second test corresponds to the with several Is well a alforithms that can be used. (2). We apply this as a prunning criteria In the following way: -everytime an element spanning path or valid circulation. tree for is added, it, if we construct we succeed we a have a 156 our brancn 26, this come furniture variants which are elements In layouts for tree. combinatorial With we stop any position of fall we -If which 4 to down for possible our test of one case the end of our long When we apply this 12 represent position our basic rules, and search. complete procedure to all the other branches 7, 27 layouts and 28, that we end up with the 19 our standard permits. 27.- e * eW ~e %S ol/m Ine m /Afendr 26 - basic fdrniture 57 158 4. COMPUTER SYSTEM: A computer system was the process defined in operation computer of SPACE and or be FUNCTION programs being implemented generation of Basic to carry last chapter. It can "furniture sl-u fler", an independent the the implemented Variants at be used as incorporated as ANALYSIS at on M.I.T. in for the Housing Level the the by M. Gerzso. From the user point of view, the system as having two definition of of existing relatlors one corresponding querying this Irformations, in the site, such as sizes of et ., and to the second for cuerying first the for furnitdre, adjacencies Implicit configurations furniture variants. Internally - the partst standards, and another corresponding to process or main appears basic it is organized in front end "SARCASM", a variants program, or four modules: user written by M. Interface of Gerzso and M. Gross. - a Relational - a - and RDB set Data Base, "RDB", manipulation LANGUAGE", the SEARCH programs, routines, OQUERY 159 4,1: O6C43A. 4.1. Relational standards The during needed the process. the Information the 1r for a RDB(1) constitutes which for the system. There Is only generation rules used in the (in the Permutztion of Elements). A small, for formulation, Information representation another Base: enumeration process and the resulting configurations are stored the maln bank of Data in core, ROB was furniture shuffler written specifically and the S.A.R. Basic Variants program. In this as entities and after the kind of data base, their relations. As description of S.A.R. Information Is Is aulte obvious and the stored now, enumeration 160 process, both our standards are basically ,%/*7iE4a the simple / that: entitles 52A,(c(~ i* function Where we have have a entIties, list "state-descriptors' and relations. For example, would be represented site our and lists of In our ROB as:5/) entities, one for tt~e another for our elements entities, and we of relatiors, in this case binary relations, one for the position rules, the other for the adjacencies. 161 AO-744V7 E 56 r 5 2 A RDB consists Information, entities and is lists before, of a general a provides it lists of relations. to up us. search process, as: What can Input We or we can input of way //E4-) in the logical representatlon structure of In those a standard as we did we put during our that have concerned with elements as El and S1. the site This general of - Where we can keep track of been positioned lists cefining state-descriptor, a for reoresentation is the data, rather contents, we can change our descriptions than as its actual we please, 162 include partial complex as formulaticns relations or construct new as retrieve desired, lists out existirg of ones. made relational through this description of Precise together with a set of structure algebra or relational can be calcu Ius, operations which can be applied to retrieve or In our case the RDB was implemented with the following data structure: where we keep the entities and as two entries for entity contain which lists external lists, and 40 their relations respectively entries for 20 relaticn lists In entity site, such entr les we keep chains el ements) basic data or the relation chains, such as (i.e. position rules, t-e about name (i.e. adjacencies ), a ena maan (re~w ~~474v) fe/*eader~ I /A4 MM#{ by (kf,4wr ____I r (/f0Ar I doeA4/?# __ I 4 .9 Yt/,,r~,, hill"2~~ad' Aea," ~~ ~ po ,~r I ) rhcA d-dA ~~&~U L~ /I~W'% or 4 p 7774 -61-. 1 I -~ r-<-~ -~ 163 Rewwt;?.. Rea4/Oay as r 17ZIIZ A 4r P )Wa.. A4-rW ansa 164 pointer end of to the beginning of of format the were not and For each Infcrmation each for entry has a name, the previous or for of this to the next entry in each element list has that we want to store. entl y to the same additio al list, link it flexible Information as to list. The besides its name and type. been to be as Each list property a in entry have an down, and two pointers that In a property kept list, we a type, a pointer further ,explained and type as such thought were relations. entities chain slots that all, at used to - he and a pointer list plus additional each list, necessary but each possible Is idea In terms of the elements we use for our reDresentations. From the description Information can be divided can see that the relational in 3.1 we can see in the part is that follcwing way: taken care of our and we by the 165 and entity is information row, here we can have, spatial *property' link several which the structures), this For each atribute keeps actual track we have a of the name of the dimensions, restrictions, data (in and th-e needed pointers to In the property In list. property 'atributes' that an entry might examplet for etc.; graphics, in or nonspatial. entry Information, incluced particular the while lists, relation the different data elements further list. By subdividing information store different kinds of elements keep their different data in In this way in the entity can we lists, different entries of ard property lists. In the other part of our relation list, being related by RDB, we have for each an entry representing a pair of elements it. In this entry we donot need to store 166 the elements again, but we store elements, a reference information lists where of location they are. two the data entry Several in in delete In the routines *id" for such them In tte doing so we avoid *1d' is the core, and we have location In other Items are poIrters that previous and next pairs retrieve, By to our case this therefore a pointer for each element The an that can help us get redundancy of Informatlor. In address Insteaa link the pair. our pair to the chair. were written or query elements to and relations, Insert, as shown figure 4.6 and in more detail, With them we can relations(PUTRDB); delete (DELENT) pairs relation or relation pairs in retrieve the values of combined queries as will all they In sert are; elements entries in relation th e Information lists relation some r elations be h own in (PUTSPC), (VALNAM); In QUERY lists (DELREL) lists or or (DFLRDB); or LANGUAGE. make - - - &- - - - KA 11reel77 W 5 - U U r4~ P 1 I dl. W. AO AWWW 1' AMP I I III U I 1 0~ A 1~JL I U I -I AO AOAW do LiEJ Wat --- -- - Gdbkv/ 168 programmers manual not these of detalIs The included in as easier understand, to data Implementation, languages manipulation 4.2. Is are Spatial flexibility, the ease irdependence, of 4.2 are of precision, ease clarity and the data (mieically present in 4.3). Representation? representation of aroung L.Teague's Ph.D. representation of thent the ones in figure The spatial organized (2) Thesis on "The Relationships Spatial layouts furniture In a Comouter building design". System for Teague network like tables simple in these Thesis. The advantages of an RDB(i) use, are routines of rectangles as a larger rectangle. Based in spaces describe within a Tutte's network representation of it extends this description the following representatlor. It organization express and network theory for in to in a squared building rectangles(3), three dimensions. network terms the By using spatial makes therefore available the results of the analysis and synthesis of spatial 169 A f3 A relationships. In this network, spaces are described by or *directed links" wtich 'fIow* correspond 'arcs' to thIe I 7170 which receive describeo as 'nodes' left the of *flow* the origin for between adjacercles The space. nori'ontal(xy) the (z) or vertical 'arc space, flow" for tre sides of two soaces are on ore side the lets say, ano the of dimsnslans which are the 'n soaces arc' right the I side, like mode This to Instance, for of Eastman's opoosed repres erta t Ion (4) or Yess io's (3), was selected two reasonst 1.- Its limitatiors to rectangle SECTORS In rot f3ct, nave complex sfaPes, they are always rectanIles because in the end, ZONES we though composed of d3 shaoes Interfere with the orincicles in the methodoloqy. ever for restrict the analysis to ortogonal soaces, ard ard two directions. 2.- Its network description blends Itself ilte 171 well our with and RDB the network becomes one more relation in our data'fo r config urations. One represent ing was change as spaces switched soaces as node s mad e, arcs sides and and si des as graph wit h geometric ch aracterI stics dimersions amount for each a nd node, representation, is some of Elements), 1iritat ions to keep spaces are rectangles always. a convention representation. add on In corresponds to the generation get an L shaped room. * free' corner, space, and subdivide the S 0, room Teague*s that as all (th we establish this *squarec" we can see how when we is layout o f I ayout we do o f c hair, I nto *room2* ele gart Permuta ticn 1 by the In branch 28) then to the end of *room' and way we is extend the our *room1*. The same happens with chair-use-space roomi" ard the resulting he does, 0 btain Wha t i.e. SouttEast shape layouts (at examp le our chair-physical-unit our to t rac k to how as this graph seauenti ally when we corstr uct we have Having th en a links. such we having at each link the is done during generation of as it nodes, as of spaces. of adjacency between two There Instead however, and we get and *room* When we put the bed-p.u. then we just reduce 172 O4AIR.0 Om~IR.LS 1!41 -, 173 'roomi' and "room2*, CHAIRJ "IR.UsKO which d sappear with the position of PL Uq!PK PU KD.PUi c"IAR.F OHAJR.L6 DESK.LU4 D~s 'ED."U the bed-use-space. The network In the right side should serve to Illustrate that there tree at each to another. level of our figures is always a spanning which allow us to move from one element 174 until we get the COAIR. final CMAIRiG layout which CM . A a is basic furniture DEgK. PU PU CMAiR. J SK AO DESI CMAI.A V lu ED.Pu CLOSU EDAIS CM CLOS var Iant. As a layer in our ROB there is a set or that keep track of this NETWORK control. corners of overlapping adlustments spaces of fall, other what spaces, in our representation is and outines They chec the ffahe where containment the or necessary as shown before 4.3. Query Language: Through this module, I shoLld say, pretentiously called, QUERY LANGUAGE, simple aueries can be constructed 47f A0,9 7%Ar 4 Q/Af -7-- ~eAf~e-o1/6~e4s~ 175 60w~evgy out of combinations of MEMBERSHIP, INTERSECTION new information in basic set and UNTON, to the R01. Together operations such ast retrieve or with VALNAM in form our 176 previous routines, relational they constitute a reduced version of we which calculus, similarly to our previous rules, the SARCASM I (Is and we exDress, following queries tbe in enti relationname entityl of 2 y2) the for look under Pair relation the relation the logical same members connective This query would be answered before. used have ) 45=,Q47 AA'AA'C7-46M~C ('/ "relationname", and OR is that to use syntax: ('CII where can a fourth element TRUE after we positioned the in our layout 1 or branch 28. (W4P4DAA (6/uAe A4,op|awr.-7o,W example, A different e lements elements by of the the to the watl2 two, adjacent to the Will to left of left and get find the list of the value the value of the bed, will produce a of will awO set of al all tte Inter sec tion elements each of them the bed and by the wall2. return the elements that satisfy any of the 177 -ae' ~bp' i 1 ~ z)) two relations, or *adjacency' 'by", performing a set uni on. o~e4ne~'M The routines ttat do this work are: 4.4. Search: Search is a recursive, backtracking proceduce 178 which carries enumeration process defined before. on the each to Its parts correspond of explained in there, coordinated This possibilities we procedure. the solution space our operations by a general exolores procedure in 9 the of tree by applying the following principles: valid starting at the root, alternative among the binary-counter or - If one level or have of furniture after advancing one level, and doesn*t find a solutior successors at the next -when there and solution, starts to - demanded, If of it look for we do, piece. it checks backtracks It it starts again, level looking if we to the If It for the down. are no valid successors to extend backtracks other nodes In when we have it succeeds It TRUE-FALSE tries to find a next successor. level possible alternative, assignment a previous In of elements. acceptable an a solution or not, one down in the tree, and applies the the positioning - for the successor nodes, whether respective operator, values, looks first the permutatior It finds then one advances it In a solution, to a a previous level different branches. and its search and ends only one is the process; 179 when are all successors, asked it continues lookin for valid advancing, backtracking and recording all other solutions until there are no more branches left explore. Its general parts are and the operations of thent enumeration correspond tte to 18o Chapter (1) Four, Martin Notes: J., "Computer Data-ease Prentice-Hall, (2) Organization", Englewood-Cliffs N.J. 1975. Teague L., "The a Computer Representation System of Spatial Relationships for Building Design". Ph.D. Thesis, Civil Engineering, M.I.T. 1968. in 181 (3) Tutte W.T., Rectangles", "Squared Proceedings IBM Scientific on Problems, Combinatorial Computing White Plains, N.Y., SymposIum IBM Data Processing Division, pp.3-9 Eastman C., (4) "Representations Communications April (5) for Space Planning", of the A.C.M., Vol. 13, No. 4, 1970. Yessios C., and Structures "Syntactic Proced ures f or Computable Site Planning", Ph.D. Thesis, of School Urban an d PubIic Affairs, Carnegie-Mellon University, Pittsburgh. 5.- CONCLUSIONS: With this there is possibe one arrangement, generative method we can or we can find configuration or fin d out all the out if furniture oossible configurations for a given spa ce. We can show then wha t layout consequences are 182 functional a in implicit in ELEMENTS or the configurations If that can a it with the S.A.R. OF CHART for a rectangular site where dimensiors func tion, ther %e shou I d have to contain a construct new disapoear. the minima I In the which observe which configurations or we are In teres space LAYOUTS, and CON STRAINTS, appear change the POSITION RULES, or or SITE, the and definition our of then with the four parts parameters standard. We can start playing we sho w CRITICAL in a matrix form, a set of rooms w ith a certain increment in the x y and we disp lay dimensions, can ho dimensions that (sequentially) the first com b ination of standard, d ou r layouts that in or f1r s t the or all are possible. I The C HART no w is used represent to spaces, and as such it I s the first of to principles in our produce not one, precise exist With supports. the functional but all the of mathematical . norms t he design t he S.A.R. can and be now be able to that arrangements CHART. sense in this "f . we norms, the furniture in ot definitions, This exhaustive caoability Is the steo apolicatior formultlon in each room of the in mind when he assigns dlirenslors an archi tect ha that to not Thesis, but *oroved" merely 'felt' in 183 that It might be true. explorations. arrangements. If arrangements, we have f or tool a for the unelegant way, but to experience exciting v ery much been made from mixing and needed, were they analyzing spatial grabing concepts manner, oersonal for norms, or standards. The generation thas hoc* a design standarcs dialogue meanlngf ul a new spatial formulating differert we can talk a bout economic perfor mance of and preferences the to factors we show the consea uences of our If can c ost we assigr more desirable* less or of terms In layouts obtain configuration s "desirable* space. we assign preference values to position If we can or relations, Inters for basis a however have We time being to 'ad olaces as in a very perhao them able be different in an been h as it enume ra te an design alternatives. If fields, left it it started poisI-1tIes barely that Interests did not end with answers in all. many questions terms from its permit S touched.* How to In difterert for each, b ut design In enu meration of a define exhaustive Instead Thesis, but How to advance In the direction of th Is a nrnhlem If I . not solved in this 184 Theory of Spatial manipulating Configurations themmight not worth to continue explor1rg. If . . be and our reasoning clear now but In certainly 185 APPENDIX: #1 / 0 / 0 / 0 0 0 3 0 (aloel"-) 0 0 4r/o*A 4W 4 J'AWAOX471 A 7-/efYOr w eA r -Al Z'kg p\Asg4ars\ L 1 gg44~'4ff4v ~~eti~v ~ -j 000zz *ot74M / : vi] FL L 4 ElI E 187 salaWM2 '}-4 1 1 %1, }%- {] cEl { }-E] E~Ei~EW I C a' H !2 I, 191 A4 I 05 wt LiJ r-I 195 -4f AS LI 4S4t4C,07rOAV AF: .1 - C' r*-. CYN '-i H