Automation in Construction 9 (2000) 117–128www.elsevier.com/ locate /autcon

Computer graphics: assistance for archaelogical hypotheses*´Stephane Potier , Jean Louis Maltret, Jacques Zoller

Ecole d’Architecture de Marseille Luminy, Marseille, France

Abstract

This paper is a contribution to the domain of computer tools for architectural and archeological restitution of ancientbuildings. We describe an application of these tools to the modeling of the 14th century AD. Thermae of Constantin in Arles,south of France. It was a diploma project in School of Architecture of Marseille-Luminy, and took place in a context definedin the European ARELATE project. The general objective of this project is to emphasize the archeological and architecturalheritage of the city of Arles; it aims, in particular, to equip the museum of ancient Arles with a computer tool enabling thestorage and consultation of archaeological archives, the communication of information and exchange by specializednetworks, and the creation of a virtual museum allowing a redescription of the monuments and a ‘‘virtual’’ visit of ancientArles.

Our approach involves a multidisciplinary approach, calling on architecture, archeology and computer science. Thearcheologist’s work is to collect information and interpret it; this is the starting point of the architect’s work who, using theseelements, suggests an architectural reconstruction. This synthesis contains the functioning analysis of the structure andbuilding. The potential provided by the computer as a tool (in this case, the POV-Ray software) with access to severalthree-dimensional visualizations, according to hypotheses formulated by the architect and archaeologists, necessitates the useof evolutive models which, thanks to the parametrization of dimensions of a building and its elements, can be adapted to allthe changes desired by the architect.

The specific contribution of POV-Ray in architectural reconstruction of thermae finds its expression in four forms of thismodeling program, which correspond to the objectives set by the architect in agreement with archeologists: (a) Theparametrization of dimensions, which contributes significantly in simplifying the reintervention process of the architecturaldata base; (b) Hierarchy and links between variables, allowing ‘‘grouped’’ modifications of modelized elements in order topreserve the consistency of the architectural building’s morphology; (c) The levels of modeling (with or without facing, forexample), which admit of the exploration of all structural and architectural trails (relationship form/function); and, (d) The‘‘model-type’’, facilitating the setting up of hypotheses by simple scaling and transformation of these models (e.g., roofingmodels) on an already modelled structure.

The methodological validation of this modeling software’s particular use in architectural formulation of hypotheses showsthat the software is the principal graphical medium of discussion between architect and archaeologist, thus confirming thehypotheses formulated at the beginning of this project.

L’infographie: une aide a la formulation d’hypotheses archeologiques

´ ´L’article que nous proposons ici presente une etude sur les apports de l’outil informatique dans la restitution architecturale et´ ´ ´ ` ` `archeologique d’un edifice antique: les Thermes de Constantin en Arles, monument date du Iveme siecle apres J.C. Ce travail

´ ´ `a constitue l’objet d’etude d’un Travail Personnel de Fin d’Etudes (TPFE) a l’Ecole d’Architecture de Marseille-Luminy

*Corresponding author.E-mail address: stephane.potier@wanadoo.fr (S. Potier)

0926-5805/00/$ – see front matter 2000 Elsevier Science B.V. All rights reserved.PI I : S0926-5805( 99 )00054-0

118 S. Potier et al. / Automation in Construction 9 (2000) 117 –128

´ ´ ´ ´(EAML) et s’est deroule dans un cadre de questionnement defini dans le projet europeen ARELATE, qui a comme objectif´ ´ ´ ´general de valoriser le patrimoine archeologique et architectural de la ville d’Arles et de doter le musee de l’Arles antique

´ ´d’un outil informatique permettant le stockage et la consultation d’archives archeologiques, la communication et l’echange´ ´ ´ ´ ´d’informations par les reseaux specialises, la creation d’un musee virtuel permettant une relecture des monuments et une

visite ‘‘virtuelle’’ de l’Arles antique.´ ´Notre approche implique un travail pluridisciplinaire rassemblant des competences archeologiques, architecturales et´ ` `informatiques. Le travail de collecte des informations et sa traduction par les archeologues sert de base a l’architecte qui, a

´ ´ `partir de ces elements, propose une restitution architecturale, travail de synthese comprenant l’analyse des structures et du´ ´ ´fonctionnement de l’edifice. La possibilite donnee par l’outil informatique (ici, le logiciel POV-Ray) de disposer de plusieurs

` ´ ´visualisations tridimensionnelles dans le temps, en fonction d’hypotheses formulees par l’architecte et les archeologues,` ´ ` ´permet l’emploi de modeles evolutifs qui peuvent facilement s’adapter a toutes les transformations souhaitees par

l’architecte.´ ` ´L’apport specifique de POV-Ray dans la restitution des thermes, comme outil de formulation d’hypotheses archeologiques,

´se traduit par quatre modes d’utilisation de ce modeleur, qui correspondent aux objectifs fixes par l’architecte en accord avec´ ´ ´ ´les archeologues: (a) Le parametrage des dimensions, qui apporte une grande simplicite au processus de reintervention sur la

´ ´ ´base de donnees architecturale; (b) La hierarchie et le liens entre variables, permettant des modifications ‘‘groupees’’ˆ´ ´ ´ ´ ´d’elements modelises pour maintenir la coherence de la morphologie architecturale du batiment; (c) Les niveaux de

´modelisation (avec ou sans parement par exemple), qui autorisent l’exploration de toutes les pistes structurelles et` ` `architecturales (rapport forme/ fonction); et (d) Les modeles-types, facilitant la mise en place d’hypotheses, par simple mise a

` ´ ` ´ ´l’echelle et transformation de modeles-types (de couvrement par exemple) sur une structure deja modelisee.´ ´ ` `La validite methodologique de cette utilisation particuliere du logiciel POV-Ray dans la formulation d’hypotheses

´architecturales, se confirme en tant que support graphique principal de discussion entre architecte et archeologues, ce qui a` ´permis d’attester les hypotheses formulees dans ce projet. 2000 Elsevier Science B.V. All rights reserved.

1. Introduction age, studied with the aid of computer graphics. Ourapproach implies a multidisciplinary work calling on

This paper is a contribution to the domain of architecture, archeology and computer science. Thecomputer tools for architectural and archaeological archeologist’s work is to collect information andreconstruction of an ancient building. We describe an interpret it; this is the starting point of the architect’sapplication of these tools to the modeling of the work who, using these elements, suggests an ar-Thermae of Constantin in Arles, south of France, chitectural reconstruction. This synthesis contains thedating from the 14th century AD. It was a diploma functioning analysis of the structure and building.project in School of Architecture of Marseille- Before studying the implications of the computerLuminy, and took place in a context defined in the as a tool in the architectural reconstruction process, itEuropean ARELATE project. The general objective is necessary to firstly analyse the medium throughof this project is to valorize the archeological and which this process will be applied. After an intro-architectural heritage of the city of Arles; it aims in duction to the computer science context used for theparticular at equipping the museum of ancient Arles restitution, the main part of this paper is dedicated towith a computer tool enabling the storage and the analysis of the archaeological context. Theconsultation of archaeological archives, the com- principles of archaeological analysis are explained,munication of information and exchange by special- the synthesis of which constitutes the basis ofized networks, and the creation of a virtual museum architectural reconstruction. The principles govern-allowing a redescription of the monuments and a ing this reconstruction are then discussed, using‘‘virtual’’ visit of antique Arles. This kind of ar- practical examples of the potential provided by thechitectural reconstruction poses a central problem computer as a teaching and research medium. Thetoday: the Delphes Tholos (restituted by the School elements are then in place to illustrate the mechanicsof Architecture of Nancy) and the ancient Lausanne governing the relationships between the architect, the(by the Cantonal Art School of Architecture), for archeologists, and the computer. Finally, the study ofexample, show the importance of architectural herit- the specific contribution of computer graphics (in

S. Potier et al. / Automation in Construction 9 (2000) 117 –128 119

this case, the POV-Ray software) as a tool forarchaeological hypothesis formulation will permit usto establish how methodologies, implemented inmodeling strategies, have an influence on the ar-chitectural reconstruction (and inversely).

2. Architectural modeling

Fig. 1. Examples of structure obtained by extrusion and rotationFor the design of new buildings or for the with POV-Ray.

reconstruction of ancient ones, architectural model-ing needs easily usable modelers. The evolution ofthe model depends upon building regulations or on very numerous except as design elements, becausehypotheses for ancient buildings. The modeler must they are often constructed in place or in the factory.be highly interactive to permit easy tuning. This A modeler based upon these primitives (polyedricconcerns forms and associations, but does not imply forms, cylinders, cones, spheres, torus) permitscommon sense interactivity with a peripheral device, modeling of the major part of contemporary andallowing direct modification of the screen representa- ancient architectural production. For real situations,tion. The modification of architectural dimensions the basic elements are modified by holes and cuts,with their consequences on all parts of the model is operations which can also be executed by previouscertainly more important than the ability of modify- primitives. From computer and mathematical pointsing an element on screen. Parametric modelers, of view, boolean operations are the correct way torecently appeared in mechanical CAD, would permit realize these actions. In most cases four operationsa similar approach. Today, this approach is at sketch are used: set complement, union, intersection, andor rough-drawing level and not in production phase difference. Every element obtained by these opera-of execution plans, but even at this level it allows tions can later be combined with others, possiblyboth rapid validation of reasonable hypotheses and modified by transformations or new parameters (Fig.coherent choice of dimensions to be used in a more 2). Description of an element is thus represented by apowerful modeler. In the first stage of this work tree of successive operations, modified by geometricPOV-Ray software has been utilized; this is both a transformations. This description has the advantagemodeler to describe a scene by means of a text, and a of being concise and can be evaluated at every stagerealistic rendering software using the ray tracingmethod.

3. Characteristics of POV-Ray

1The user of POV-Ray can use a set of geometricprimitives and volumetric operations to representobjects. Among these primitives are quadrics (2nddegree equations), quartics (4th degree equations),surfaces of revolution, extrusions, and more specialforms such as blobs and height fields which are notvery well adapted to architectural modeling (Fig. 1).Basic geometric forms used in architecture are not

1POV-RayE is based on DKBTrace 2.12 by David K. Buck andAaron A. Collins. See http: / /www.povray.org /povcd. Fig. 2. Examples of boolean operations with POV-Ray.

120 S. Potier et al. / Automation in Construction 9 (2000) 117 –128

if some dimensions are changed. Text description is the beginning of the century before restoration, ita simple way to describe these trees and allows an was noted that some large modifications were made,easy modeling, after convenient decomposition of which eradicated many clues. Therefore, the ancientthe objects in elementary forms and operations. Any traces and clues must be verified by constant com-part of a CSG tree corresponds to a subset of the parison with infographical layout and plotting.object: in POV it is possible to extract any sub-tree,name it, and re-use it after modification of parame- 4.2. Principles of archaeological analysisters or application of geometric transformations. Thiscapability allows multiple usages of generic objects The aim of this analysis is to determine whichused in many parts of a scene with some variations. elements guided the original conception and con-

struction Thermae of Constantin: the architecturalinfluences, constraints, and technical at architectural

4. Archaeological context experience of the project manager at the period.These elements are:

4.1. Presentation of the building • the history of Arelate Roman city, of which the4th century reveals a program of civil monumen-

The building, such that one can observe at present, tal constructions;is located in the north of the city of Arles, near the • the terminology of thermal baths, which classifies

ˆriver of Rhone. What remains of the edifice are walls this building as ‘‘thermae’’ instead of ‘‘balneae’’;up to 17 m high; the north part is the well-preserved • the terminology of thermal rooms (caldarium –one (Fig. 3). One can distinguish several well hot room, tepidarium – tepid room, frigidarium –delimited rooms in the zone open to the public. To cold room, . . . ), of which the etymology of room-the south, one room of the thermae is occupied by a types can be described by the strict progress of andwelling. Further to the south are other ancient ancient bather, and therefore the function of eachfragments situated in cellars and in elevations of room of the building;these dwelling. The graphic documents forming the • the general evolution of Italic thermal baths, as anbasis of this study are the plans provided by Arles examination of the thermal edifices in westernMuseum, drawn by the archaeologists J.Bremond provinces from the 1st century to the beginning ofand M.Heijmans (Fig. 4), and two perpendicular the 3rd century, confirming the adaptabilty of thecross-sections of the caldarium by an historic monu- thermal edifice to many parameters (topology,ments architect. The Thermae of Constantin were town planning, etc.);largely excavated during the last century, leading to • and finally, the description of the Thermae ofinterventions in the site. In viewing documents from Constantin, in comparison with other thermal

Fig. 3. View from northwest of the Thermae of Constantin.

S. Potier et al. / Automation in Construction 9 (2000) 117 –128 121

Fig. 4. Plan of the current state of the Thermae of Constantin, showing the antique traces in black.

baths of the ‘‘Narbonnaise’’ area (the ancient ing other directions of research, thanks to the three-south of France), which allows the specification dimensional approach. Under these conditions, theof the morphology of each room, the dating of the part of archeologists is not to give an absoluteedifice, and the constructive antecedents of the guarantee to all the propositions, but to cast asite. ‘‘critical’’ eye, essential for a productive exchange.Many of these elements in the general context of The architectural reconstruction relies on an ar-

the Thermae of Constantin remain hypothetical, but chaeological synthesis, giving us a basis for for-they obey a certain coherence; it is this consistency mulating, and validating (or not) a hypothesis, whilstwhich has guided in the architectural restitution, as a respecting the architectural coherence. In the precise‘‘schedule of conditions’’ given from the past. case of the Thermae of Constantin, the good state of

preservation of a large part of the caldarium allowsus the use of a structural and morphological attested

4.3. Principles of architectural reconstruction basis, which constitutes an ‘‘objective’’ starting pointfor the restitution relating to general themes (con-

During architectural reconstruction, care must be structive systems, water circulation, roofing, etc.) andtaken that only hypothesis are formulated and not a room-by-room study. The situation is comparableunjustified assertions. The aim is not to restore to a jigsaw in which some parts are already in place!precisely an edifice as it was originally; this work is Thus, one can establish general principles (proper toabove all useful for new discussions on different our building) from particular ones, with a meth-restitutions, reinterpretation of excavations, and find- odology similar to the process of a ‘‘usual’’ architec-

122 S. Potier et al. / Automation in Construction 9 (2000) 117 –128

Fig. 5. General principles of architectural reconstruction.

tural project (Fig. 5). This is the point wherecomputer graphics begin their use as a tool forarchitectural hypothesis formulation, with the dualmission of research and communication, acting as areal graphic medium between architect and ar-cheologists. Using this technique, the situation ofheating in the Thermae is significant at the teachingand visual level of image synthesis. The heating byhypocauste in the Thermae of Constantin is one ofthe best-preserved of Roman Gaul. This is under-floor heating where smoke is sucked by tubulures inthe walls. Many elements of this device are still inevidence, the fireplace, heating room, etc. (Fig. 6). Inaddition to assisting the hypothesis formulation,

Fig. 7. Partial reconstruction of the whole hypocauste system ofimage synthesis also becomes a teaching tool in the Thermae of Constantin, exept for the heating rooms. The redreconstructing the heating course in the building arrows symbolize the course of the smoke, from fireplaces to the

flues.(Fig. 7).Another practical use of computer graphics is the

provided by a three-dimensional vizualisation of aarchaeological hypothesis formulation. In the room-hypothesis, very difficult to formalize in two dimen-by-room restitution of Thermae of Constantin, thesions. Originally, some incoherent traces on southfrigidarium east apse is typical of the potential

Fig. 6. Traces of the piles-circulatin ground system in the caldarium of the Thermae of Constantin.

S. Potier et al. / Automation in Construction 9 (2000) 117 –128 123

Fig. 8. The visible traces of the laconium south wall of the Thermae of Constantin.

wall of laconicum (dry steamroom, Fig. 8) made any making the roofing of this apse by a half-cupolaroofing impossible for this room. It is thanks to possible (Fig. 10-3 and 10-4). The image, more thanmodeling that the solution appeared (Figs. 9 and being a discussion medium, becomes a real part of10-1): the laconicum south wall had to be doubled the reconstruction process (Figs. 11 and 12), and anby a ‘‘filling’’ wall (Fig. 10-2), of which an elevation essential communication medium between architectpart (not taken into account before) can be seen, and archaeologists.

Fig. 9. Plan of the current state of the frigidarium east apse of the Thermae of Constantin.

124 S. Potier et al. / Automation in Construction 9 (2000) 117 –128

5. Computer graphics as a tool

Thanks to the software currently available, thearchitect, working not only with plans, cross-sectionsor elevations, has access to a three-dimensional viewof a building. This approach allows him to develophis capacity for analysis, understanding and con-ception. The validation of architectural hypotheses,restricted to a graphic plan /cross-section /elevationrepresentation, is much less controllable than three-dimensional visualizations using parameters based ongeometry. This potential provided by the computer,with access to several three-dimensional visualiza-tions at one time according to hypotheses formulated

Fig. 10. Reconstruction hypothesis of the frigidarium east apseby architect and archaeologists, necessitates the useroofing.of evolutive models which, due to the parametriza-tion of the dimensions of a building and its elements,can be adapted to all the changes wished for thearchitect. This essential aspect of parametrizationmade us choose the POV-Ray software for thearchitectural hypotheses formulation, as it is one ofthe few which offers this possibility. The specificcontribution of POV-Ray in the architectural restitu-tion of the Thermae of Constantin, provides four ableforms of this modeler, corresponding to the objec-tives set by the architect in agreement with thearchaeologists: parametrization of dimensions,hierarchy and links between variables, levels ofmodeling, models and hypotheses.

5.1. Parametrization of dimensionsFig. 11. Perspective from east.

The modeling of the building with POV-Ray isdivided into three distinct elements:• A main file including the attributes of light and

texture, a statement of the objects to be repre-sented, and the observer’s position and his view-point.

• A file of the variable declarations correspondingto the numeric dimensions of the building’selements (attributes of height, length, width,angle) (see Listing 1).

• The modeling files themselves describing objects’coordinates and their transformations, which are,in fact, the variables described above (see Listing2).This parametrization of the database is performed

Fig. 12. Inner cross-section of the complete model. by the command [declare; as such, variable setting

S. Potier et al. / Automation in Construction 9 (2000) 117 –128 125

5.2. Hierarchy and links between variables

This notion of parametrization is coupled with thelinks between variables, these being used to defineseveral objects at once. The hierarchical organizationof these relations is possible with POV-Ray specificcommands, such as inclusion ([include) allowingthe hierarchical organization of data inside a file.This inclusion principle allows a concrete divisionbetween building structures, e.g. caldarium,frigidarium, tepidarium, etc, and architectonic ele-ments e.g. upper cornices, lower cornices, joineries,roofing, etc. (Fig. 13). The links between the vari-ables constitute an essential element for the model-ing: thus, the radius of North apse of caldariumListing 1. Part of the variable declaration file of the caldarium

north apse of the Thermae of Constantin. rayonCoupole is the same as the radius of the arch’sin front of this apse, decreased from the ‘‘redan’’(Fig. 14): →rayonArc5rayonCoupole2

of the numeric data is possible. The declarations, like redanCoupole. The variable rayonCoupole is there-most of language instructions, can appear anywhere fore used for the definition of these two elements asin a file, and allow direct relationship between for the apse’s rear, but this variable also is used laterseveral variables. These distinctions, allowing to describe other parts of building as integral part ofparametrization of dimensions, has a distinct advan- the coordinates in relation to origin. For instance, thetage in the ability to check dimensions, e.g., follow- x coordinate of the East Piscina is given by theing a first numeric model built from the documents variable rayonCoupole added to the one of the north-produced by archaeologists, a dimensions plotting east wall’s length. This principle allows the propaga-made on the spot has shown some distortions tion of modifications within the model. For example,compared with the existing plans. Due to the parti- the modification of the half-cupola’s radius (bytion of the modeling files and of the parametriza- simple numeric insertion) leads to the arch cupolation’s variable one, the data modification can be then to the apse’s rear, while respecting the structuraldone by simple numeric insertion in a single opera-tion, without modification of other files.

Listing 2. Part of the model file of the caldarium north apse of the Fig. 13. Arch elements of the original plan, modeled in anThermae of Constantin. ‘‘include’’ file.

126 S. Potier et al. / Automation in Construction 9 (2000) 117 –128

Fig. 14. Data variables of the caldarium north apse: north-south cross-section.

and morphological integrity of the whole thermae.However, it is necessary to organize into a hierarchyonly compatible elements with each other: two veryidentical elements from morphological and dimen-sional standpoint, but independent (like the half-cupolas of caldarium and tepidarium), must corre-spond to independent variables in that precise case.The data modification is not restricted by the soft-ware: all the operations are reversible, following theexample of the boolean operations. This principle ofre-intervention is conditioned by the arborescentstructure of POV-Ray modeling, described by textualform, modifiable by nature. So, with two sets of

Fig. 15. The junction between the roofing and flues apertures in plans (numerical and with variables), the user canthe north-west angle of the caldarium: current state.

proceed through modifications if there is a plotting

Fig. 16. The junction between the roofing and flue apertures in the north-west angle of the caldarium: restoration using computer graphics.

S. Potier et al. / Automation in Construction 9 (2000) 117 –128 127

error, or for refining an archaeological hypothesis trace of the west vault does not correspond to thealready modeled. image created with POV-RAY, especially for the

insertion depth of the drain pipes. This archeological5.3. Levels of modeling reconstruction can therefore be considered as erro-

neous because of the false vault’s curvature, provedA primary model of the ‘‘crude’’ structure (with- by the differences of insertion depth of the drain

out facing) of Thermae allows a primary check of pipes (Figs. 15 and 16). This example occurs at thethe structural hypothesis. The use of parameters and first level of modeling. At the second level ofthe relations between variables of the ‘‘crude’’ modeling, the junction of the tubulis with the drainstructure, used to model the inner facing, allows a pipes located in the vault base of the caldarium northsecondary level of models to refine previous hypoth- wall poses a problem which is solved with theeses. The parametrization of dimensions then allows attachment of the facing thickness or of the interiorswitching from ‘‘crude’’ to facing structure, which coating: the necessity a regular appearance of thegives more flexibility in hypotheses validation. For caldarium arches curvatures induces the position ofinstance, in the first level of models, one can also the junction between tubulis and their evacuations.verify the presence or absence of tubulis (flues The possibility of several three-dimensional views ofhanging on the heated room walls) by the traces of this junction allows a fast validation of the correctbronze tenons which served as support. These were hypotheses.partial and scattered in the site. The walls placed The possibility of a distinction between ‘‘crude’’further back make the presence of tubulis on the and ‘‘covered’’ structures in the phase of modeling iswalls highly probable, as thus show the cohesion of straight forward with parametrization of dimensions,facing in relation to the basic structure. The restora- and allows the exploration of all the structural andtion of the caldarium’s vault has shown that the base architectural trails (relations between form and func-

Fig. 17. The architectural reconstruction process of the Thermae of Constantin.

128 S. Potier et al. / Automation in Construction 9 (2000) 117 –128

tion) on the one hand, and on the other hand to use the process of architectural reconstruction of theof data of a ‘‘finished’’ building exploitable during Thermae of Constantin can be formalized in Fig.the transition to a more powerful modeler. 17.

5.4. Models and hypotheses

6. ConclusionThe preparation of ‘‘typical models’’ using param-eters such as roofing, facilitates the setting up of

With the use of simple geometric models, a toolhypotheses by simple scaling and transformation oflike POV-Ray allows the formulation of architecturalan existing model. The use of models is also veryhypotheses comprehensible by the participants in thisuseful to conceive a satisfactory lighting principle forproject. Acting as a dialogue tool between architectthe different rooms of Thermae. The opening traces,and archaeologists, the software can not take part inonly visible in the caldarium and the laconicum,the restoration of a building, which is the field ofallow the creation of an open model proper to theother methods such as stereophotogrammetric plot-Thermae of Constantin. For the realization of theting or laser. However, it yields a useful model for‘‘typical models’’, the ‘‘plotting’’ of existing ele-general public simulations such as virtual reality, onments, compared to other building models, allowsthe condition that a model compatible with the realthe formulation of new ‘‘dimensional’’ archaeologi-time navigation, for example the VRML language,cal hypotheses.can be produced. In this context, the model used inthis exercise, requires conversion into an environ-5.5. Methodological validationment which can produce this kind of data, and this• The main constraint in computer modeling is thework is facilitated by the use of a finished modelregularity of the geometrical shapes inherent inbased on hypothesis formulation made with POV-the model. This regularity is often in contradictionRay.with the site plotting, which supposes many

geometrical exceptions. In the case of POV-Ray,the use of a model with regular primitives, whichis the only approach usable for a building such as

Bibliographythe Thermae of Constantin, has no consequencesfor hypotheses verification with archaeologists in

´J.-P. Adam, La construction Romaine – Materiaux et techniques,relation to the reconstruction scale.Grands Manuels Picard.• The complete model files do not constitute an

P. Gros, L’Architecture Romaine – Les Monuments Publics, Lesexploitable database for a site plotting, but are an ´ ´manuels d’Art et d’Archeologie antique, Picard editeur, 1996.essential tool for archaeological hypotheses ˆA. Choisv, L’Art de batir chez les Romains, Arnaldo Forni editore

´– Reedition, 1984.formulation at the time of architectural recon-P. Ward, B. John, Architecture Romaine, Editions Gallimard,struction.

Milano, 1994.• The absence of strict structural rules for plan and´C. Sintes, M. Moutashar, Musee de l’Arles Antique, Actes Sud,

elevation is a characteristic of the Thermae of 1996.Constantin. The systematization and the respect ´l’EFR, Les thermes Romains, Actes de la table ronde organisee

par l’EFR, Rome, 11–12 Novembre 1988, Collection defor modular principle of a building like the Arlesl’Ecole Francaise de Rome, 142, 1988.Roman amphitheatre (or of the imperial baths) can

´A. Bouet, Les thermes prives et public en Gaule Narbonnaise,not be adapted to the Thermae, because of the´Doctorat de l’Universite de Provence, Aix, Marseille.

unusual characteristic multiplicity of our building, ´UFR Civilisations et Humanites, Centre Camille Jullian, 1997.in contrary to a building such the amphitheatre, P. Quintrand et al., La CAO en architecture, Hermes, 1985.

´which made from modular repetition. To sum up, O. Aune, Vers un Musee Virtuel, 1996.