Datacab: a geographical‐information‐system‐based expert system for the design of cable networks

Da acab: a geog aphical-in o ma ion-sys em- based expe sys em o he design o cable ne wo ks In ˜ igo Monede o, 1 Ca los Leo ´ n, 1 Robe Denda 2 and Joaquı´n Luque 1 (1) Escuela Te ´cnica Supe io de Ingenie ı ´a In o ma ´ ica, Depa amen o de Tecnologı ´a Elec o ´nica, A da Reina Me cedes s=n, 41012 Se illa, Spain E-mail: [email p o ec ed] (2) Manage o Technology Se ices, Endesa Ingenie ı´a, Edi icio Expo, Isla de la Ca uja s=n, 41092 Se illa, Spain E-mail: [email p o ec ed] Abs ac : Telecommunica ion ne wo ks ha e e ol ed o e ime as a esul o echnological ad ances, and ne wo k opologies and equipmen ha e become inc easingly complex. Expe sys ems a e being success ully applied o he managemen o elecommunica ion ne wo ks. Howe e , applying expe sys ems o ne wo k design is ano he especially beneficial ye s ill no e y common app oach. In his pape we p opose a ule-based expe sys em called Da acab. Da acab was de eloped a Endi el Endesa in collabo a ion wi h he Elec onic Technology Depa men o he Uni e si y o Se ille, o he au oma ic design o hyb id fib e coax (HFC) cable ne wo ks. Using da a om a geog aphical in o ma ion sys em as inpu , i au oma ically gene a es iable HFC ne wo k designs. Keywo ds: elecommunica ion ne wo ks, expe sys em, HFC ne wo k, GIS 1. In oduc ion As a esul o he g ea echnological e olu ion ha has aken place in elecommunica ions, i is possible o communica e and access in o ma- ion apidly, ia elephony and communica ion ne wo ks such as he In e ne . Telecommunica- ion ne wo ks ha e e ol ed o e ime o sa is y he demands o di e en communica ion se - ices, con inuously equi ing g ea e band- wid h and a be e quali y o se ice. Ne wo k echnology has inc eased in complexi y, gene - a ing he need o a be e adminis a ion o he esou ces o hese sys ems, which has os- e ed he e olu ion o ne wo k managemen sys ems. On he o he hand, esea ch ca ied ou in he field o a ificial in elligence has led o impo an ad ances in he ea men , s o age and exploi - ing o knowledge. These ad ances include he field o knowledge-based expe sys ems (Bucha- nan & Duda, 1982; Liebowi z, 1998; Ignizio, 1991; Awad, 1996; McG aw & Ha bison-B iggs, 2002). Expe sys ems a e being success ully applied o he ealiza ion o di e se asks (in e p e a- ion, p edic ion, diagnosis, design, planning, ins uc ion, con ol e c.) in mul iple fields such as medicine, geology, chemis y and enginee ing (Liebowi z, 1998). Expe sys ems may also be applied o he field o elecommunica ion ne wo ks, because many o he p e iously enume a ed asks equi e specialized knowledge and a e needed o he managemen and ope a- ion o hese ne wo ks, which can be au oma ed by an expe sys em. Among hese asks is he one o ne wo k design, which in pa icula equi es he knowl- edge o expe s. This pape desc ibes an expe sys em called Da acab o he au oma ic design o cable ne wo ks. The aim o Da acab is o sa e ime and wo k du ing he ask o designing a cable ne wo k. The pape fi s gi es an in oduc- ion o he s a e o he a o expe sys ems applied o elecommunica ion ne wo ks; sec- ond, he hyb id fib e coax (HFC) ne wo k s uc u e o which Da acab has been applied is p esen ed, and las ly, he s uc u e and design o Da acab as well as a base case and expe imen al esul s a e desc ibed. 2. Applica ions o expe sys ems o elecommunica ion ne wo ks Figu e 1 schema ically shows he di e en applica ion domains o expe sys ems ela ed o elecommunica ion ne wo ks (Liebowi z, 1988; ITU-T, 1991). The e ficiency o he asks o ope a ion and main enance o cu en ele- communica ion ne wo ks la gely depends on he deg ee o coope a ion be ween he manage- men sys em and he human ope a o s. The managemen sys em plays a e y impo an pa , in o ming on possible anomalies and execu ing he commands o con ol. The di e en asks in ol ed in his manage- men and o which comme cial expe sys ems cu en ly exis a e as ollows. Failu e managemen . This ask e e s o he g oup o unc ions ela ed o ne wo k man- agemen necessa y o he de ec ion, diagno- sis, eco e y and co ec ion o sho comings o he elemen s ha o m a elecommunica- ion ne wo k. Examples include Max and Op i-Max (an expe adminis a o o ele- phony line main enance de eloped by NYNEX) (Rabinowi z e al., 1991), T ouble Loca o (an in elligen sys em de eloped by Pacific Bell o loca ing ailu es a he physi- cal le el o a local elephony ne wo k) (Chen e al., 1996), ANSWER (an expe sys em de eloped by AT&T and in cha ge o supe - ising hei 4ESS swi ches) and Scou (de el- oped by AT&T o iden i ying pe sis en ailu es in he ne wo k) (Raman, 1999). Configu a ion managemen . This unc ion helps he adminis a o o exe cise con ol on he configu a ion o he componen s o he ne wo k. Changes in he configu a ion a e ca ied ou o elimina e conges ion, o isola e sho comings o o make changes equi ed by use s. Examples o expe sys- ems suppo ing configu a ion managemen include ECXpe (de ec ion and epa a ion o ailu es, de eloped by Lucen Technolo- gies) (Raman, 1999), ACE (an expe sys em de eloped by AT&T o help in he de ec ion and diagnosis o ailu es in cables) (Liebo- wi z, 1988) and NEMESYS (also de eloped by AT&T o a oiding ne wo k conges ion) (Raman, 1999). Accoun ing managemen . This unc ion al- lows posi ions and cos s o he use o he esou ces o he ne wo k o be de e mined and loca ed, e.g. APRI (de eloped by AT&T o o ecas ing he p obabili y o alling in o new deb ) (Ezawa & No on, 1996). P esen a ion managemen . This ask helps he ne wo k adminis a o s o moni o and e alua e he benefi s o he sys em. Example Failu e managemen Con igu a ion managemen Ne wo k managemen Accoun ing managemen P esen a ion managemen Secu i y managemen Expe sys ems applied o Use in e aces Final use applica ion Communica ion sa elli es Ne wo k design P o ocol e i ica ion Rou ing Figu e 1: Expe sys ems in elecommunica ion ne wo ks. expe sys ems o p esen a ion managemen a e Ne =Ad iso and Ne Command ( o moni o ing he s a e o he ne wo k in eal ime) (Raman, 1999). Secu i y managemen . This unc ion helps he ne wo k adminis a o s manage he se - ices ha p o ide p o ec ion in he access o he communica ion esou ces. Examples o expe sys ems ela ed o secu i y manage- men include EXSYS (a se o expe sys ems de eloped by Pacific Bell which execu es on an AT&T 3B2=600 compu e o supe ising he loop main enance ope a ing sys em om a se e ) (Raman, 1999). In addi ion, he e exis o he expe sys ems such as Ne HELP (which p o ides assis ance o use oubleshoo ing) and ExSim (assis ance in he p ocess o ou ing a ne wo k) applied o o he fields (Raman, 1999), enume a ed in Figu e 1. The e is an especially in e es ing ye less in es iga ed field o he applica ion o expe sys ems o he wo ld o elecommunica ion ne wo ks: sys ems applied o assis du ing he asks o ne wo k design. Wi hin he design p ocess, he use o knowledge-based sys ems is o pa icula in e es o he opological design o he ne wo k. Among hese asks a e he ollowing:  he choice o he ype o ansmission: ans- mission by physical suppo (cable, mic o- s ip, wa e-guided) and=o ansmission by adio (elec omagne ic wa es).  he layou o he ne wo k: in adio ne wo ks i would be necessa y o es ablish he layou depending on he ype o communica- ion (bidi ec ional o omni-di ec ional) and signal p opaga ion and in e e ence cha ac e is ics. In cable ne wo ks, his ask comp ises he p ocess o cable layou . placemen o he ne wo k elemen s: es ablish- men and localiza ion o he di e en nodes and necessa y elemen s in o de o ob ain a sui able signal le el (spli e s, amplifie s e c.). The las wo asks in pa icula equi e he wo k o expe s who apply a ious design ules and hei empi ical knowledge. Because o he g ea numbe o limi a ions o he physical media, hese asks a e pa icula ly complex in cable ne wo ks. On he o he hand, due o he e y specific cha ac e is ics o he design ules o each ne - wo k, as well as he high complexi y o he applica ion o he knowledge, he de elopmen and comme cial use o expe sys ems o cable ne wo k design is s ill no e y widely ex ended. Thus, he cu en ly exis ing app oaches mainly ocus on esea ch (Lo & Ghau i, 1991; Wu & Lee, 1993; El-Fishawy & Khamis, 2000). In some cases, he design is only pa ial (Lo & Ghau i, 1991); o he app oaches (Wu & Lee, 1993; El-Fishawy & Khamis, 2000) a e limi ed o he design o local ne wo ks. In his pape we p opose an expe sys em o he comple e de- sign o a cable ne wo k. 3. The HFC cable ne wo k A ypical cable ne wo k, which co e s a med- ium-sized ci y, consis s o an HFC ne wo k ha is made up o an op ical fib e ne wo k and i s con inua ion in o a coax ne wo k. Po en ially, a pa allel elephony ne wo k is deployed (Tun- mann, 1995; O adia, 2001). The cable ne wo k s uc u e suppo s cable ele ision, In e ne access and elephony se - ices; i is ep esen ed in Figu e 2. The s uc u e Main nodes 30.000 o 90.000 Fib e P ima y Ne wo k Seconda y nodes 2000 homes Seconda y Ne wo k ONT ONT ONT Fib e Op ical Ne wo k Te mina ion 500 homes Dis ibu ion Ne wo k Remo e Telephony cen e Figu e 2: The HFC ne wo k o be designed. o he ne wo k depends on each coun y and ope a o , bu ypically a fib e op ic ne wo k is made up o a head-end connec ed o a p ima y ne wo k, which con ains a numbe o p ima y nodes, e.g. se icing be ween 30,000 and 90,000 homes each. These main nodes a e connec ed o a ne wo k o se e al seconda y nodes (second- a y ne wo ks), each o which ypically se es a ound 10,000 homes. In gene al, he seconda y nodes also con ain a emo e elephony cen e (RTC). The e ia y ne wo ks (dis ibu ion ne - wo ks) connec o he seconda y node and end in op ical ne wo k e mina ions (ONTs), which a e connec ed o a coax ne wo k. This coax ne wo k finally dis ibu es he signal o he subsc ibe s. The elephony ne wo k ypically s a s a each RTC and has h ee le els. The fi s le el con- nec s each RTC o an ONT, he second le el eaches a g oup o homes and, finally, he hi d le el connec s wi h he subsc ibe s’ homes. In Figu e 3, a ep esen a ion o a ypical coax a ea o a cable ne wo k ex ac ed om a geo- g aphical in o ma ion sys em (GIS) (Heywood e al., 1999; Ko e, 2001) is shown. In his figu e, he buildings a e ep esen ed as shaded poly- gons, he cables as lines, he connec ions as small ci cles and boxes, and signal amplifie s as iangles. In he design o an HFC ne wo k he layou o he fi s , seconda y and dis ibu ion fib e ne wo ks a e no mally easy o ca y ou because he loca ion o hei elemen s depends on he a ailabili y o adequa e p emises. Thus, he pa o he ne wo k ha pa icula ly equi es deep expe knowledge du ing design is he coax ne - wo k. When a human designe ca ies ou a coax ne wo k design he in many cases is only guided by his expe knowledge and in ui ion, and e y basic suppo ools. The designe wo ks his way: wi h he help o specific so wa e which eads and ep esen s he geog aphical in o ma- ion om a GIS he begins he layou a he ONT p og essi ely placing he cables and ele- men s and ca ying ou he calcula ion o he signal le el. I a some poin i is no possible o each he desi ed signal le el on a pa icula Figu e 3: Design o a zone o he cable ne wo k. connec ion wi h he layou in p ocess, he de- signe has o edesign all o a pa o he layou . Following his commonly used app oach, he a e age ime o p oducing a finished design o an ONT zone o a ound 500 connec ions is a ound 8 hou s. 4. Da acab: an expe sys em o designing a cable ne wo k 4.1. In oduc ion Da acab is an expe sys em we designed and implemen ed based on ules o he au oma ic design o an HFC elecommunica ion ne wo k and which wo ks in an in eg a ed manne wi h a GIS. The majo ad an age o he p oposed expe sys em app oach is he g ea educ ion o he human wo k o ce needed o he design o he HFC ne wo k. Wi h Da acab, ins ead o doing he designs himsel , he human expe only has o e i y he di e en designs gene a ed by Da acab. Ano he impo an ad an age is i s adap abili y, being possible o easily modi y he knowledge base i he design c i e ia change, and i s high in eg a ion, wo king as add-on so wa e o he exis ing GIS. Finally, simila o o he expe sys ems, he de eloped sys em can be used o ain new s a . The GIS in o which Da acab is in eg a ed is GE SmallWo ld (Rigaux e al., 2002), a powe - ul comme cial GIS wi h special deploymen in he a eas o dis ibu ion ne wo ks like wa e , elec ici y, gas and elecommunica ion ne - wo ks. 4.2. In e ac ion wi h GIS Da acab pe o ms he design o he HFC ne wo k h ough he phases ep esen ed in Figu e 4. Phase 1: Da acab p ocesses he geog aphical in o ma ion co esponding o he zone o be designed, adding i o i s knowledge base. The inpu da a o he di e en elemen s o he zone wi h hei loca ion and all edesign in o ma ion (i.e. possible access o accom- moda ions, p e ious channel loca ions, non- pe mi ed ac¸ ades e c.) is ob ained om he GIS. Phase 2: By means o a g aph ep esen a ion o he GIS elemen s o he zone and he mechanisms o applica ion o he knowl- edge, Da acab ca ies ou he necessa y cal- cula ions o achie e an op imal solu ion o he cable ne wo k design. P e iously, a da- abase access is ealized in o de o ob ain he di e en pa ame e s, such as he deg ee o a enua ion o he cables o he gain o he a ailable amplifie s o he di e en com- me cial ne wo k elemen s ypically ins alled by he HFC ope a o . Phase 3: Da acab p ocesses he ob ained in o ma ion, gene a es he ne wo k design based on he ule da abase and p o ides he GIS wi h he geog aphical in o ma ion o he elemen s belonging o he cable ne wo k. 4.3. Design Da acab’s a chi ec u e consis s, like mos expe sys ems, o a knowledge base and an in e ence engine. The knowledge base consis s o a class s uc u e and a ule-based sys em which a e desc ibed u he below. The in e ence engine, which akes ca e o he fi ing o he ules, is Da abase o elemen s GIS da a Ci y in o ma ion Redesign da a Da acab inpu Da acab da abase que ies GIS da a Ne wo k elemen s Da acab ou pu Figu e 4: Da acab’s in e ac ion wi h GIS. ca ied ou by an expe sys em en i onmen called ART*En e p ise which is desc ibed in he coding sec ion. In addi ion, a salience mechan- ism was de eloped and added o ule in e en- cing. This mechanism makes i possible o fi e he ules in he o de ha co esponds o he p io i y defined o each ule. We selec ed a knowledge ep esen a ion, ak- ing in o accoun ha he p oblem was o med by di e en ne wo k elemen s, and a se o ules and condi ions necessa y o place hese elemen s in he layou . These ules, which a e defined in in e nal design guides and p ocedu es o he company, we e s udied and defined du ing he knowledge acquisi ion phase. The geog aphic elemen s a e ep esen ed by objec s acco ding o he class diag am o he applica ion. The necessa y knowledge da a o ca y ou he design is con ained in he ule- based sys em. The me hods which execu e he fi ing o he ules a e enclosed in a class called dc:algo i hm (ke nel package). 4.3.1. Class diag am The class diag am o Da acab was designed wi h he help o he Ra ional Rose ool using UML (E iksson & Penke , 1997; Booch e al., 1999) and is o med o se en packages, which a e linked as shown in Figu e 5. The class diag am includes geog aphi- cal elemen s (s ee s, buildings e c.), edesign in o ma ion (i.e. access poin s o buildings), coax elemen s (wi es, amplifie s e c.), ci il ele- men s (caske s, channels e c.) as well as he elemen s co esponding o he elephony and fib e ne wo ks, in addi ion o supe classes, which we e c ea ed o gene alize some common p ope ies o ou objec s such as dis ances in linea elemen s and signal losses in elemen s o he coax ne wo k. The e is an impo an package called Base- Type. This package ep esen s he g oup o geome ic elemen s which a e inhe i ed by he di e en elemen s om he map, ci il, adio equency and ke nel packages. In Figu e 6 he classes ha o m his package and ha a e needed o deal wi h he geog aphic p ope ies o he elemen s a e illus a ed. The class dc: segmen ep esen s a segmen o geome ic elemen s. An objec o his class is composed o wo objec s o he class dc:poin , which ep e- sen s a geome ic poin and which is cons i u ed by a pai o coo dina es. The class dc:a ea which ep esen s he elemen geome ic a ea is Ci il Coax Fib e BaseType Ke nel Map global Telephony Figu e 5: Package diag am. dc:gis id : Enum{...} s a e:Enum{...} 2..*0..* dc:segmen poin s : Se (Poin ) 0..* dc:a ea 3..* dc:poin x : Floa y : Floa 2..*0..* Is o med by 0..* 3..* Is o med by <<o de ed>> poin Lis : Sequence(Poin ) Figu e 6: BaseType package. composed o a so ed lis o objec s o he dc:poin class. Finally, he e is a class called dc:gis om which all he men ioned classes de i e. The coax package has a supe class called Coax, which con ains h ee classes co espond- ing o he ype o coax elemen : passi e compo- nen (i.e. wi es), ac i e componen (amplifie s) and powe sou ce ( o eed ac i e componen s). Thus, he ad an ages o inhe i ance a e impo - an , o example, in classes inhe i ed om passi e elemen s which ob ain cha ac e is ics on he one hand om coax elemen s and on he o he hand om hei geome ic da a. Se ing as an illus a i e example, he class diag am o he passi e elemen s is shown in Figu e 7. Da acab includes a da abase o elemen s. This da abase con ains all he cha ac e is ic da a om he ne wo k elemen s o he co esponding ci y o zone. Thus, o ins ance, i he company changes a model o amplifie i is no necessa y o ep og am Da acab bu simply o upda e he da abase. In his way, Da acab will ins an ia e one copy o each di e en child class o he coax diag am and he a ibu es wi h echnical cha ac e is ics a e filled in acco ding o hei cons uc o ia da abase accesses. 4.3.2. Algo i hm When he geog aphic in o - ma ion is ead om he GIS, each elemen o he zone, like a s ee o a building, is con e ed in o an objec o he class diag am and a connec ed g aph is gene a ed. Ini ially we in es iga ed wo al e na i es o he design o he algo i hm, based on he no ion o algo i hms o design o an expansion ee ela ed o a non-di ec ed g aph (Sedgewick, 1992; Cas illo e al., 1997; B a ko, 2001). The al e na i es we e as ollows. The fi s al e na i e was based on he idea o K uskal’s algo i hm, which selec s he edges so ed by cos a oiding cycles in he layou . K uskal’s algo i hm selec s he lowes cos edge in each o he i e a ions, as long as i does no gene a e any cycles. The second al e na i e was based on he unning o P im’s algo i hm, which s a s om a poin and selec s se ial edges o mino cos , building an expansion ee in his way. A e s udying and applying hese wo al e na- i es o ou pa icula p oblem, and designing p o o ypes based on bo h o hem, we decided o implemen he fi s al e na i e which begins by de e mining he layou and hen es ablishes he si ua ion o he elemen s o adio equency. I i is no possible o loca e he elemen s in he selec ed ou e, a sea ch o a di e en ou e is dc:poin ( om BaseType) dc:passi e Cu en Blocke dc:coaxcable dc:segmen ( om BaseType) Te minalCha geEqualise Tapdc:spli e Figu e 7: Coax passi e elemen s. ca ied ou and an a emp is made o eloca e he elemen s. No e ha when a human expe ca ies ou he design o a ne wo k, he adio equency elemen s a e placed a he same ime as he layou is de e mined, which esembles mo e he second algo i hm. Howe e , his app oach was disca ded because o i s highe sys em equi e- men s and because he fi s algo i hm p o ed o esul in simila ly good design esul s. The di e ence be ween K uskal’s algo i hm and ou algo i hm lies in he dynamic cha ac e - is ics o he la e . While he edge cos in K uskal’s algo i hm does no depend on he p e iously selec ed edges, in ou algo i hm he cos is allowed o a y. We dis inguished wo kinds o edge cos in he algo i hm. The s a ic cos o a ce ain edge is defined as he cos ha does no depend on he p e- iously selec ed edges o he layou by he algo i hm, i.e. he cos ha is independen o he s a e o he layou such as he longi ude o a ce ain s ee and he necessa y cable o co e i . The dynamic cos is defined as he cos which a same poin depends on he p e iously selec ed edges o he layou ; i.e. he cos which is ela ed o he s a e o he layou as, o example, he cos o a c ossing be ween a placed cable and ano he cable ( he com- pany ies o a oid such cases). These ules use me hods belonging o he di e en adio equency objec s which a e de- i ed om he class adio- equency. The e a e wo especially use ul me hods o signal-le el calcula ion. Reg essi e signal-le el calcula ion calcula es eg essi ely he le el o signal a a ce ain objec . The e o e, i calcula es he signal le el belonging o an elemen and all he elemen s connec ed om i un il he ONT. P og essi e signal-le el calcula ion calcu- la es p og essi ely he le el o signal a an objec . Thus, i calcula es he signal le el belonging o an elemen and all he elemen s connec ed om i un il he di e en poin s o connec ion in which i is in ol ed. 4.3.3. Rule-based sys em The knowledge da a a e ep esen ed in he o m o i – hen ules o , o pu i ano he way, o ules made up o p emises and conclusions o which a salience mechanism was added in o de o p o ide hem wi h a p io i y o de . This sys em o easoning has been chosen ins ead o o he kinds o easoning, such as easoning based on cases, because he design o an HFC ne wo k is based on ules defined by he expe ience o he expe s. The e a e fi e main g oups o ules in ou sys em. Rules o ini ializa ion (R1) a e a g oup o ules which ca y ou he p ep ocessing o he in o ma ion o he s ee s om he GIS, con e ing hese da a in o Da acab in o ma- ion. In addi ion, hey ca y ou a se ies o p e iously necessa y calcula ions o he ex- ecu ion o he algo i hms. Rules o edge so ing by s a ic cos (R2) a e a g oup o ules which ca y ou a so ing by s a ic cos o he di e en s ee s o he zone. This g oup o ules comp ises he ules o he di ision o zones, e.g. om 500 homes in o ou zones o abou 125 homes. Rules o so ing by dynamic cos and selec- ion o edges (R3) so by dynamic cos he di e en s ee s o he zone. In addi ion, his g oup con ains he necessa y ules o selec On he o he hand, he p e ious cos s a e de e mined by a p io i y numbe which was defined o he fi ing o each o he ules. Thus, i he fi ing o wo ules a he same poin was possible, he one wi h he highe p io i y would be fi ed and he e o e i would ge assigned a lowe cos . Once he layou is achie ed, he ne wo k elemen s a e placed by means o a g oup o ules which calcula e he signal le els co esponding o each o he nodes o he layou . These ules hus de e mine he posi ioning o he amplifica- ion elemen s based on he calcula ed a enua- ion, in o de o p o ide he subsc ibe wi h he igh signal le el. om he p e ious o de he edges necessa y o he layou . Rules o loca ion o elemen s (R4) is a g oup o ules belonging o he placemen o he di e en necessa y adio equency elemen s o he co ec wi ing o he zone. By apply- ing hese ules, which calcula e he signal le el a each node, he di e en adio e- quency elemen s (i.e. amplifie s, close s e c.) a e loca ed. I i is no possible o loca e he elemen s, i is necessa y o go back and fi e a new g oup o ules. Rules o gene a ion o he ou pu da a o- wa d he GIS (R5) gene a e he files wi h he necessa y GIS in o ma ion o he specifica- ion o he di e en elemen s belonging o he esul ing layou , as well as hei co e- sponding pa ame e s. The o de o fi ing o he abo e-men ioned g oups o ules is shown in Figu e 8. Fi s , Da acab is ini ialized and he GIS da a a e ead by means o fi ing g oup R1. La e , he o de ing o he se e al edges is ca ied ou by means o g oups R2 and R3. The o de ing om dynamic so ing gene a es a layou . The elemen s a e placed and he signal calcula ions a e ca ied ou by means o he g oup R4. I i is possible o each he desi ed signal le el wi h he gene a ed layou om he p e ious phase, he design is conside ed good and he GIS da a a e gene a ed (R5). O he wise, ano he s a ic o de is o ced and he e o e ano he layou is designed. Cu en ly, he o al numbe o ules is 147. Table 1 shows some examples o ules o each g oup. The able con ains ou columns wi h in o ma ion o each o he ules. The fi s column indica es he phase in which he ule is fi ed, he second con ains he name o he ule, he hi d shows i s p io i y, wi h a lowe numbe meaning a lowe p io i y, and finally in he las column a b ie desc ip ion o he ule is gi en. 4.3.4. Coding Fo coding hese ules, we ha e used a ool o expe sys ems called ART*En- e p ise. F om he p og amming pa adigm pe spec i e, ART*En e p ise p o ides objec - o ien ed p og amming, ule-based p og am- ming, case-based easoning and p ocedu al p og amming. I s p og amming language is e y simila o CLIPS (Gia a ano & Riley, 1994), main aining he same s uc u e. In Figu e 9 we can see an example o he coding o a ule in ou sys em, which p e e s he edges ha ep e- sen ac¸ ade cables o e edges ep esen ing cables equi ing ubes. The ule has a p io i y o 700 which is defined in he code a e ‘salience’, in he p emises zone (be o e ¼>). The ule chooses be ween wo possible objec s (s ee 1 and s ee 2), which ep esen wo di e en kinds o edges ( ac¸ ade o ubes), and p e e ably selec s he one which is o e ac¸ ade. An ART*En e p ise applica ion gene ally consis s o h ee laye s: a g aphical use in e - ace laye , an applica ion model laye and a da abase mapping laye . We did no implemen he g aphical use in e ace o Da acab since his unc ion was ca ied ou by he GIS appli- ca ion. ART*En e p ise has file managemen capa- bili ies ha make i possible o easily ca y ou he exchange o da a be ween he GIS and Rules o ini ializa ion Rules o so ing o edges by s a ic cos Rules o so ing by dynamic cos and selec ion o edges Rules o loca ion o elemen s Rules o gene a ion o he ou pu da a owa d he GIS Figu e 8: O de o fi ing he g oup o ules.