Educative Distributed Virtual Environments for Childrentisseau/pdf/paper/jt-art-int-7.pdf ·...

18 Journal of Distance Education Technologies, 2(4), 18-40, Oct-Dec 2004

Copyright © 2004, Idea Group Inc. Copying or distributing in print or electronic forms without writtenpermission of Idea Group Inc. is prohibited.

ABSTRACT

This paper presents a distributed virtual reality environment for children called EVE—Environnements Virtuels pour Enfants. The virtual environment architecture is reactive agentsbased. The FCM-like dynamic action planning mechanism assures agent’s adaptability to itsenvironment changes. This virtual environment supports cooperation among members of adispersed team engaged in a concurrent context. By the means of their avatars, special cases ofagents, users are allowed to interact and to give decisions using cooperative mechanisms. Auser-friendly interface enables teachers to create their own stories that fit with children’spedagogical requirements and generate new virtual environments according to the teacher’sspecifications. The implementation is based on DeepMatrix as environment server, VRML andJava as languages and Cortona VRML plug-in from ParallelGraphics. It is actually running onthe Internet: http://eve.enib.fr

Keywords: autonomous reactive agents, distributed virtual reality, fuzzy cognitive map, virtualenvironment

Educative Distributed VirtualEnvironments for Children

Dorin-Mircea Popovici, ENIB/CERV, Franceand OVIDIUS University of Constanta, Romania

Jean-Pierre Gerval, ISEN, FrancePierre Chevaillier, ENIB/CERV, FranceJacques Tisseau, ENIB/CERV, France

Luca-Dan Serbanati, OVIDIUS University of Constanta, RomaniaPatrick Gueguen, Ecole Kroas Saliou, France

INTRODUCTION

Learning to read is a difficult but im-portant task for a child. It takes time andsupposes a constant effort from its part.Like many other school activities, readinginvolves child’s capacities as attention andmemory, knowledge (of letters) and know-how as searching and discovering theword’s sense in a given context. Small ca-pacity in attention and incapacity to rein-vest in a new task of already known no-tions are just some examples of obstacles

in a child’s learning process. More, a taskfailure can generate a fear from the child’spart concerning particular notions, or no-tions related with the unaccomplished task.

At the beginning of primary school,some children have not achieved all thenecessary acquisitions to basis tasks. Forexample, some have difficulties to place inorder a set of labels in order to constructsimple sentences. And this may have mul-tiple causes; the child has not well memo-rized the oral sentence, is not able to realisea correspondence between the oral sen-

Journal of Distance Education Technologies, 2(4), 18-40, Oct-Dec 2004 19


tence and the written one, or, is not able tocoordinate the spatial distribution from leftto right with temporal succession of alreadypronounced words. In order to avoid theinitial difficulties to accumulate and to pro-duce a child’s discouragement, part of theactivity in primary schools is organisedaround small groups of children, rather thanindividual work. More, when we talk aboutlearning at childhood age we are usuallyusing a story metaphor, because narrativelends itself to active exploration of a do-main through challenging and enjoyableproblem-solving activities, which is essen-tial for learning.

On the other hand, the informal con-text in which the child learns reading haswell evolved in the past three decades.New technologies such as Internet, multi-media, and virtual reality are now parts ofour children’s everyday life. For this rea-son, it is not surprising that educators growtheir interest in distance learning and dis-tributed education, and try to use these typesof media in their lessons.

Our paper presents a distributed non-immersive virtual environment (VE), calledEVE: Environnements Virtuels pour Enfants(Virtual Environments for Children), whichhelps primary school children learn to read.Based on a story reconstruction, it offerschildren a pleasant approach to learning bymeans of two games that implement emu-lation and cooperation.

In the following, after a brief state-of-the-art of educational computer aidedenvironments for children, we present theEVE project from a pedagogical perspec-tive. Next, we will make some consider-ations from an architectural point of viewand we give some insights concerning theimplementation of our project. Finally, wewill make some conclusions based on theexperience obtained during the project andwe end with perspectives of our work.

BACKGROUND

With ever-increasing computationalprocessing power, the rapid growth of theWorld Wide Web, and the ongoing construc-tion of a digital communications infrastruc-ture, the creation of distributed (immersive),multi-participant VEs running on theInternet starts to prove its usefulness in childeducation.

These learning experiences maycome in many forms. Educational quiz-likeor puzzle-like kids’ games for ABC activi-ties such as reading, reading comprehen-sion, math, writing, and so forth, and digitallibraries, such as QueryKids (Druin et al.,2003), are some examples of multimedia-supported learning environments. The cre-ativity, encouragement, and great motiva-tion supported by interactivity in a socialcontext permits passive children to becomeactive. More, these experiences realisecultural enrichment of a child’s knowledge(Pantelidis, 1995). Dedicated Web sites, like“J@rdin des jeunes branchés” (J@ardin,2004), provide online learning activities thatare combined with games and informationfor elementary school projects as a help tonavigate on the main Web sites.

Much of the appeal for applying VRin education is derived from the observa-tions of educational theorists (Bruner, 1986;Piaget, 1929) who have stressed the valueof actualizing learning through making itmore real for students. Learning is facili-tated through the construction of conceptsbuilt from the intuitions that arise out of theirdirect experience of the environment. Theopportunities for young users to visit placesand interact with events that distance, time,or safety concerns would normally prohibit,the greater understanding of conceptsthrough the creation of multi-modal meta-phors or representations and the ability to



scale and manipulate these representations(Winn, 1993; Youngblut, 1998) are the ma-jor benefits of applying VR in educationalenvironments.

In the early 90s, narrative-based sys-tems such as Oz (Bates, 1992; Mateas,1997) or Virtual Theatre (Hayes-Roth &vanGent, 1997) have been mainly devel-oped for the purpose of entertainment. Theypropose story contents asking the users,particularly children, to accomplish sometasks.

CityGame (Volbracht, Domik, Backe-Neuwaldand & Rinkens, 1998) studieschildren’s’ ability to find physically andmentally one’s way in the 2D and 3D envi-ronments. Round Earth (Johnson, Moher,Ohlssoon & Gillingham, 1999) andQuickWorlds projects (Johnson, Moher,Leigh & Lin, 2000) look at the issues in-volved in the use of projection based VRsystems with children, specifically in thewell known ImmersaDesk (Johnson et al.,2002) or CAVE VR theatres. The effec-tiveness of multiple representations in en-hancing education in elementary school isinvestigated.

NICE (Roussos et al., 1999) imple-ments a persistent virtual garden in whichchildren may collaboratively plant and har-vest fruits and vegetables, cull weeds, andposition light and water sources to differ-entially affect the growth rate of plants.Their interactions constitute the story of thegarden.

KidsRoom (Bobick et al., 2000) isanother interactive narrative for children.Using images, lighting, sound, and computervision action recognition technology, achild’s room was transformed into a story-based world for play. Objects in the roombecame characters in an adventure, andthe room itself actively participated in thestory, guiding and reacting to the children’schoices and actions. Through voice, sound,

and image the KidsRoom entertained andprovoked the mind of the child.

Systems as Storykit (Montemayor etal., 2000), Teatrix (Prada, Machado &Paiva, 2000), Puppet (Marshall, Rogers &Scaife, 2002) or GhostWriter (Robertson,2002) go further and evaluate the use ofVE and computer mediated communica-tion technology in the domain of teachingchildren story writing skills. Based on im-provisation as a social and perceptual ex-perience, the children are expected to in-teract with each other, playing the roles ofdifferent characters, as in GhostWriter. InPuppet, children are allowed to play mul-tiple roles in an interactive narrative: audi-ence, actor, scriptwriter, and editor. Theyare able to both “dive in,” taking on therole of a character in the drama, and to“step out,” reasoning about a character’semotional states and goals, as well as re-flecting upon their own character dialoguerecorded while playing with the system asscriptwriter. While in Storykit children cre-ate physical immersive story environmentsusing low-tech construction materials suchas cardboard, glue, paper and ink pens, aswell as high-tech smart objects, in Teatrixthey are doing the same using a set of pre-defined scenes and a set of predefinedcharacters. These characters may act onbehalf of the children or autonomously.Each child will expect the story to evolvein reaction to her character’s actions. So,their characters must act in a believableway, in order for the story creation envi-ronment to engage the children in an en-tertaining experience, which can meet thechild’s cognitive needs to interpret, under-stand and interact with the world in termsof stories.

As we can see, there are several VRenvironments that are applied specificallyto elementary children learning. Based ona story metaphor, they encourage children



to be co-constructors of narration, promot-ing in this way the deep, connection-build-ing, meaning-making activities that defineconstructivist learning; or suppose thechildren’s participation as actors.

In the environment we present in thefollowing, we have adopted a story-basedapproach too. Unlike the mentionedprojects, we ask the children to discover aproposed story, based on a set of imagesthat are distributed in a virtual school. Eachimage is described by a sentence. At itsturn, the sentence is constructed using fivelabels. In order to discover the story, thechildren have to reconstruct the sentencesand after that to cooperate in order to pro-pose a logical order of discovered images.This way, they learn reading by playing.

EVE PROJECT:GENERAL DESCRIPTION

The EVE project has started in 1998.Initially funded by the Fonds Francophonedes Inforoutes (Francophone InformationHighway Fund), EVE has involved ninepartners (universities, primary schools andSMEs) from three countries: France, Mo-rocco and Romania. Since 2000, EVE isused by elementary schools from partners’countries as a supplementary tool in teach-ing children to read. More, in November2003 an experience between France andJapan was realised. Children from the pri-mary school Kroas Saliou (Plouzané,France) cooperated with children fromOotsukadai elementary school (Yokosuka,Japan).

The target of the project is twofold:1. To implement new cooperative working

environments.2. To initiate new products development

such as pedagogical software for pri-mary school children.

On a pedagogical point of view, themain goal is teamwork. Children from dif-ferent classrooms and countries are in-volved in cooperative work. They have toachieve a common task together, hopingthat this will encourage curiosity and re-spect in a multicultural framework, at achildren’s level, and not only.

On a technical point of view, the EVEproject implements distributed virtual real-ity technologies. Software had been devel-oped using Virtual Reality Modelling Lan-guage (VRML) and Java languages.

From both perspectives, we can de-fine our project as a NICE-like one, beingnarrative, constructionist and collaborative.Even if it does not consume a lot of re-sources, like KidsRoom, it can offer to itsyoung users great satisfaction because oftheir capability to transfer the virtual expe-riences in the real world and vice-versa.

Actual use of virtual environment isput into the context where social patternsoutside the desktop are considered. In ourcase the social dynamics and physical pres-ence even meant that the kids achieved anefficient cooperation in the classroom whenthey come to learn how to use the soft-ware and interact.

Pedagogical Aspects

The primary pedagogical goal is toinstil in children the capacity to read andunderstand a text. If for the reading andphonemes association from six to 18 monthsare necessary, for text understanding thechild needs a much longer period of time.The latter understanding may become moreprofound and subtle.

Very often children become able toread without knowing what they are read-ing. And this represents a problem in alllearning processes. How can we resolve a



problem in math if we are not able to readand understand its content?

If the children are enthusiasts at thebeginning, this state of spirit is calmed downwith the time passing because of the longtime interval necessary to learn, by the dif-ficulties met and by the mixture betweenthem. It is strongly necessary that teach-ers adopt a proper strategy in order to di-versify and refresh children’s interest. De-spite the diversity of existed and used edu-cational tools such as books, libraries, mov-ies, theatres, educational games, and soforth, the learning process seems to be torepetitive. And this takes place in the child’slife, full of other activities, such as sports,playing an instrument, TV, video games,holidays…

The child need of diversity does notstop at the school’s doors, and our dayschools are plugged into this evolution.

Two Games

The EVE application has been devel-oped in order to help primary school chil-dren to learn reading. It offers children apleasant approach to learning by means oftwo games that implement emulation andcooperation. This way, EVE is involved intothe VR-based education current, being asource of educational diversity and rich-ness.

The Labels’ GameThis first game is a self-training step

according to a global learning method ofreading, based on the label’s game. Thetarget of this first game is to build sentences.A picture illustrates each sentence, in or-der to suggest to the children the messageof the sentence (Figure 1). The child mustdiscover the sentence using a disorderedset of words and moving words to the rightplaces. In case of doubt, the child can hearthe sentence by clicking on the appropriateicon. When the sentence is correct the childwins the picture. The child must win threepictures before being able to join the sec-ond game. Three children are working con-currently in three different virtual rooms,as orange, green and blue ones (Figure 2).The order in which the children finish theirrooms is the same order they will have theright to express themselves in the next step.More, the first one will be allowed to startthe second game. And this motivates themmore.

Let’s Find Out the StorySeveral studies suggest that coopera-

tion between children in virtual environ-ments has a positive effect on learning(Johnson et al., 1999). It is not very un-usual that performances that were expectedwithout results in a child-adult context may

Figure 1: The Labels Game

Figure 2: The Virtual Environment



appear in a children’s cooperative context.This is why collaborative learning is one ofthe most important requirements of our tech-nology. Within a group of children, eachchild contributes with its knowledge in thegroup’s knowledge, more or less easily ac-ceptable by the others participants in itsgroup.

In the framework of the second game,the spirit of the concurrence adds to theteam spirit because of the cooperative con-text we propose to children. This time thechildren will meet each other in the com-mon virtual room (Figure 2). Each childowns a personal avatar. The team of threechildren must now build a story. For this,they have to place the previously won pic-tures in a logical order. According to theorder they have finished their rooms, theyare allowed to move and propose positionsfor images. By viewing the moving of theselected images by the other children, thechildren are sure that they cooperate andthey all have the same goal; even if theyare in a neighbouring room, school or in aforeign country.

A special mechanism is needed inorder to avoid conflicts: for example if twochildren want to move the same picture totwo different positions. We have chosen toimplement a voting mechanism (Figure 3):

• A child moves a picture to a chosen po-sition and requests a vote from the oth-ers.

• The two other children tell him if theyagree or not. They are voting (greenstands for yes, red for no).

• According to the vote result, the chosenposition for the picture will be acceptedor not.

When the team has finished buildingthe story two cases may appear:1. Pictures order is wrong. Badly placed

pictures are removed and the team muststart again to build the story.

2. Pictures order is right. An agent comesin the virtual common room and tells thestory to the children.

This way, the success or the failureof the second game becomes a team re-sult.

Interactivity

EVE supposes the ability of a child tofind physically and mentally one’s way inthe 3D space. Spatial orientation is an im-portant ability for understanding, interpret-ing and developing the world in which thechildren live. Children are given tasks of

Figure 3: The Second Game —Voting Mechanism

Figure 4: Avatars’ Behaviour



navigation and orientation inside the envi-ronment through intersecting paths in or-der to make them interact. The experiencedchildren with three-dimensional computergames have certainly influenced thebehaviour of the subjects.

One important result was that chil-dren often focused on the navigation tools(mouse and direction buttons) rather thanon their tasks. More, they love to play “hide-and-seek” and are eager to see what chil-dren in other rooms are doing. The appli-cation lets them to perform this task, be-cause emulation concept is the basis of thefirst game. According to this fact, twomechanisms have been implemented:• a scoreboard, which indicates the num-

ber of pictures that each child has won,• a teleporting device, which enables the

child to view what happens in otherrooms.

A first idea when developing the EVEproject was to implement a chat that en-ables children to discuss and explain theirchoices, especially during the second game.Previous experiments demonstrate that thischoice was not suitable for the inexperi-enced in reading and writing learners. Ac-tually, teachers are the main users of thechat!

By the mean of an avatar’sbehaviours (Figure 4) we increase the re-alism of the virtual representations of chil-dren playing in the virtual world and inducea better communication between children.We have chosen to implement behavioursfor avatar actions as Walk, Standby, andused in children communication as“Hello!”, “Help me!”, “Thank you!”, “Iagree,” “I disagree,” Laughing, Smiling,and “Good Bye!” These previous lists arenot exhaustive and should be easily com-pleted according to current experimentswith teachers and children. The child may

select behaviour, which will be broadcasted(gesture and sound) to other childrenthrough it’s the child’s personal avatar.

In order to increase the interactivitylevel, some of the subjects suggested popu-lating EVE with more elements to interactwith.

Agents in EVE

In the very first stages of our projectwe have limited the number of children ableto participate in a working session to three,according to the number of partner coun-tries involved into the project. This was astrong constraint for children. Even theteacher proposed a certain story to play,children love some stories more than oth-ers, and usually there are more than threechildren that want to play the same story inthe same time. So, in order to give to chil-dren the freedom to use the environmentwe have chosen two solutions.

The Local VersionIn the local version of EVE, the task

is basically the same except that there areno other children to participate in the vot-ing mechanism. This means that the childhas to complete all the three rooms in or-der to have access in the common, butempty, second room. Walking through theenvironment enables children to get use ofthe VRML plug-in functionalities beforestarting a distributed working session withother children. This way, a story may beused by as many children as possible.

Using Virtual AgentsCooperation is an important concept

of our technology. Suppose that the chil-dren are currently playing in EVE and forsome unexpected cause a child is deter-mined to quit the game. In such situations,the game cannot continue because normally



there is nobody to take the released place.In order to avoid this type of blocking, wehave introduced a new user type, the vir-tual agent, which coexists within the envi-ronment and acts in the absence of a child.Doing so, the child can quit the game with-out affecting its flow, its role being switchedto the corresponding virtual agent.

Based on virtual agents the numberof users restriction falls. In other words,there are still only three children able tomove labels and to vote in the same story,but we do not block the participation of otherchildren as before. This way, the child isable to start alone a distributed workingsession, based on the cooperation with theexisting agents. When there is another childwho wants to participate, it is accepted asthe second child, and the two children willcontinue the game. Finally, when the thirdchild arrives in the environment it is ac-cepted and the team is completed. Fromnow on, all the other children will be ac-cepted just as simple participants; they can-not express themselves by voting but theyare allowed to cooperate (i.e., move labels)with others.

More, as in real life, the teacher andparents are represented in the environ-ments, based on their special agents. Theyare the main users of the chat. While the

parent’s associated agent is not differentfrom the child’s ones from a behaviouralpoint of view, the teacher’s one is. As wewill explain more in detail later, theteacher’s agent is able to assist the child inits game. And he or she will do this by tell-ing the sentence or even the whole story.

Cultural Exchange

Another dimension of EVE project isthe cultural one. The proposed story-basedmetaphor facilitates the possibilities both forteaching familiar material in new ways, andpresenting completely new material. So theyare potentially of great value to education.

In order to facilitate and so to encour-age the multi-cultural exchanges, we haveextended EVE to support multi-languageworking sessions. This way, children canstart learning foreign languages, or at leasthaving a first contact with them, using theproposed stories. For the moment, in EVEwe have stories in French, English, Roma-nian and even in Japanese (Figure 5).

Children & Teachers Together intoan Interdisciplinary Team

Children can play an important rolein creating new technologies for children

Figure 5: Cultural Exchange through Language



(Druin, 1999). Therefore, we have estab-lished a geographically distributed team ofcomputer scientists, educational research-ers, visual artists, primary school childrenand their teachers. During discussionsamong teams’ members, children andteachers gave a strong feedback concern-ing stories’ subjects, their graphic repre-sentation and audio support. More, the chil-dren were involved, at their demand, inaudio and graphical support of stories’ cre-ation process. This way, we were able notonly to understand the children’s impactonto designing process, but we have alsounderstood how the new technologies canimpact children as users, and even as cre-ators of these technologies.

In addition, the integration of primaryschool teachers as environmental and peda-gogical designers has permitted us torealise a balance between technical andpedagogical aspects of our project.

DESIGNINGCONSIDERATIONS

For the development of EVE as edu-cational application, we have consideredtechnological, domain specific, pedagogi-cal and psychological aspects. First andforemost we have considered that in orderto develop an application for real use inclassrooms, we need an extendable sys-tem as a platform. And there is no singletechnology that fits all the application andeducational needs. More, pedagogic anddidactic skills were needed to adapt theapplication to the user requirements. Be-cause the primary school children are themost part of our users, we have reduced atminimum the necessary skills to be trainedor enhanced, like spatial abilities, but wehave considered a number of psychologi-cal aspects. All these considerations have

influenced content design, user interface’sdesign and concepts for evaluations.

The Environment

The VE consists of a set of roomsthat creates a simple maze (Figure 2). Theenvironment wall’s textures are based onreal images from the Kroas Saliou primaryschool; wall paintings made by children, oron traditional French, Romanian or mo-rocco images. Doing so, children havefound the environment familiar, when evenfrom architectural point of view it was notso. Visibility and accessibility aspects weretaken into account also. As the three chil-dren have chosen one colour to log in EVE,orange, green or blue, we have use thesame login colour in order to highlight therooms through which they have to pass.More, we have matched the rooms’ dimen-sions with the children’s avatar ones.

Cooperation is basically supported bydirectly embodying the users in the VE us-ing different avatars, and providing themwith different inter-user communicationfacilities such as a set of predefinedbehaviours.

As constructivism underlines(Mantovani, 2001), learning takes placewhen learners can build conceptual mod-els that are both consistent with what theyalready understand and with the new con-tent. In order to ensure the successful ad-aptation of old knowledge to new experi-ence, we have provided some flexible learn-ing directions. The first one is the integra-tion of different types of information andeducational support other than 3D repre-sentation (such as audio annotations andimages). More, we have defined specifictasks to the children through interaction.Participants are able to move their avatararound the rooms using the arrow keysand/or using the mouse. They are able to



pick up objects (labels, images or speak-ers) in the VE by click-and-drag or justclick them. Participants can communicatewith each other using a chat channel.

The Story

A story is defined by an ordered setof nine sentences. As previously mentioned,a picture illustrates each sentence. A soundfile is also linked to the sentence in order toenable children to hear the sentence. Asentence is defined by an ordered set ofwords. As it is presented, (Figure 1), a pic-ture is associated to each word. The storydata model is presented hereafter (Figure6). According to this data model, which isreally simple, we have chosen to implementthis model using a simple directory struc-ture.

The Users’ Interfaces

Another challenge was to provideteachers with a user-friendly interface (Fig-ure 7) that enables them to build their ownstories. Using this tool, teachers can cre-ate stories that fit children pedagogical re-quirements all along the school year. Teach-ers can also involve children in designingnew stories for their friends.

For each sentence, the teacher willchoose: a gif file (the picture) and a wavfile (the sound), and write the text of thesentence.

When this job ended, the teachersaves the story on the server. Before pro-ceeding to data transfer, the applicationautomatically:• Generates a gif file for each word tak-

ing into account word length in order touse the best character font size;

• Transforms each wav file into an mp3file in order to reduce the amount of data.

The teacher can clear the interface,and load or delete a story from the serverpreviously created.

A new virtual world that implementsthe new story is generated when theteacher decides to publish its story. On theserver side, symbolic links to the virtualworld are created in the new story direc-

Figure 6: Story Data Model

Figure 7: Story Interface — TOP (the left), BOTTOM (the right)



tory structure. At the same time, an e-mailis sent to a moderator.

A special interface (Figure 8, left) hasbeen implemented that enables the mod-erator to view the story and to hear sen-tences. The moderator will read the newstory and decide whether he or she acceptsit or not according to its content.

The last interface (Figure 8, right)enables children or teachers to choose avirtual world. The list of all published sto-ries is shown on the left upper corner ofthe page. The user:• Chooses a story in the list;• May click on the picture to view all pic-

tures;• Can press a button to read sentences—

sentences disappear when the mouse isreleased.

This interface enables the teacher toread the story before he or she starts work-ing with the children.

VIRTUALENVIRONMENT’SARCHITECTURE

In EVE, we consider that the virtualenvironment is populated by virtual entitiesthat correspond to objects populating thereal world. In our vision, the “meaning” ofeach virtual entity is its associated object

in the real world. As we know, in order tomodel an object in the real world some es-sential properties of the object should beused. They will make up the object’s mean-ing from the modeler’s point of view. Inthe virtual environment, for each propertyof the real object the modeler considersessential will be an attribute of the virtualentity with its informational shape, so thereal object’s property is for us the meaningof the attribute associated shape.

In the following, a virtual entity,briefly called entity, is the set of all infor-mational shapes, as introduced in Popovici,Serbanati and Harrouet (2003), which com-plete its meaning (Figure 10). By using vari-ous criteria, we can structure the set ofentities within the virtual environment. Forexample, the entities can be specialised inreceptors and effectors. A receptor is astimuli detector in an informational space,while an effector realises shape modifica-tions by means of the entity’s actions in thevirtual environment. The entities may begrouped in order to produce complex ag-gregations, themselves entities. In the fol-lowing section the agent as an aggregationof entities is presented.

Agent’s Architecture

In our approach, a virtual agent is acomplex entity in the virtual environmentable to perceive, decide, and react based

Figure 8: Moderator’s Interface (the left) and the Child’s One (the right)



on its profile, internal structure, and tasks.Users’ avatars are particular cases of vir-tual agents.

In order to realise the agent’s settingin situation into the virtual environment wehave followed the immersion, the interac-tion and the autonomy principles (Tisseau& Harrouet, 2003). For this, we were in-spired by the human perception mecha-nisms (Hefco, 1997), as Herrero uses(Herrero & Antonio, 2002). Our approachuses perception and emission fields of theentities, which are generalisations of me-dium, nimbus, aura, and awareness notionsas they were introduced in Benford andFahln (1993). The entities are supposed topopulate a multi-dimensional informationalspace. This space is called virtual envi-ronment and it is viewed as the union of allentities’ fields.

The interactions between agents areof cause-effect type: any change of theagent’s state represents a possible causeand may be followed by a stimulus emis-sion. In our approach, the stimulus containsinformation regarding the agent state(Popovici, Serbanati & Gerval, 2003). Allthe agent’s actions are followed by stimuliemissions.

Our virtual agents are autonomous inthe measure of their adaptability to theirdynamic environment. For this, they haveto be able to perceive the environmentchanges, to decide and to react accordingly(Figure 9). A more detailed view of theagent’s architecture is given in Figure 10.

During the virtual agent’s life, its stateis given by the values of its attributes thatare the generators of its structure. Thestructure’s variations are produced by theeffectors and are perceived by means ofthe receptors, under the form of stimuli.These modifications may be triggered bythe reception of an external stimulus such

as a change in environment followed byemission of internal stimuli.

The receptors from perception mod-ule generate perceptions based on thesestimuli. The resulting perceptions will befurther used by the decision component,which is responsible for effectors’ activa-tion. In their turn, effectors, founded in theaction module, may operate structural modi-fications followed by stimuli emissions, ormay trigger themselves specific stimuli.This way, the agent’s life cycle is com-pleted.

Information concerning the agent’sobjectives and abilities, as well as its worldmodel, is stored in its knowledge base. Thisis the only component which is modifiablenot by the means of the agent’s effectorsbut the decider one.

The problem that arises concerns thedynamical aspect of the agent’s environ-ment. In such an environment, the agent’sdecision based upon its actions may becomeobsolete between the moment the decisionis made and the moment the action is ap-plied. And this is due to the fact that in thistime interval the agent’s action selectionmechanism does not update the activatedaction context. In order to eliminate thisinconsistency we use a dynamical FCM-like action planning mechanism, based onagent’s perception, situated at decisionlevel.

Figure 9. Agent’s architecture.

Figure 9: Agent’s Architecture



Agent’s Decider ModuleUsing fuzzy cognitive maps (FCM),

instead of the hierarchical task networks(HTN) (Cavazza, Charles & Mead, 2002),we are able to ensure the agent’s adapt-ability to environment changes. A FCM isan influence graph having as nodes ele-ments of a set of concepts (Kosko, 1997).Concepts may be sensorial concepts (if theyexpress perception values), internal con-cepts (for knowledge or decisional values),or driving concepts (for actions/objectivesvalues) that the agent possesses(Parenthoen, Reignier & Tisseau, 2001).

The environment perturbations de-tected through the agent’s receptors willrepresent imposed values of sensorial con-cepts. This way, each time the FCM is com-puted, the agent’s behavioural (re)activationis realised.

Let us consider the following FCMs,associated to the agent that plays the roleof the tutor in EVE (Figure 11). Here, thesets of FCM concepts are “obstacle left,”“obstacle right,” “child far,” “child close,”“good answer” and “bad answer” as sen-sorial concepts, “motivation to approach”and “need to assist” as internal concepts,

and “turn left,” “turn right,” “approach” and“assist” as driving concepts. “Advice,”“valid sentence” and “valid story” are alsodriving concepts but, as we shall see in thenext section, they are used in “assist” ac-tion plan. This means that they are not in-

{ 1..* receptor and 1..* effector }

Group(f rom Structure)

Entity(f rom Structure)

1..*

+component

1..*

Attribute(f rom Structure)

Shape(f rom Structure)

1..*

+shape

1..* 1..*

+attribute

1..*

Stimulus(f rom Environment)

Action(f rom Action)

Effector(f rom Action)

1 ..*

0..*

1 ..*

0..*

modify

11launch

Receptor(f rom Perception)

0..*0..*

receive

FCM(f rom Decision)

Knowledge(f rom Knowledge)

0..* +capability0..*

AnaliserDecidor(from Decision)

commandlisten

1..*1..*

0..*0..*

use/update

Objective(f rom Action)

Agent

0..*0..*

0..*0..*

1..* 1..*1..* 1..*

Avatar

T

T

T

T

T

T

Figure 10: Inside View of the Agent’s Architecture

Figure 11: Parts of the Tutor’s FCM



volved in the decisional process but thebehavioural one.

In our model, not only the agent’sobjectives are viewed as fuzzy goals (El-Nasr, Yen & Ioerger, 2000), but also theaction plans. This is because they all havecorresponding driving concepts in thedecider’s fuzzy maps. An action with itscorresponding driving concept greater thana specific value is called active; otherwisethe action is called inactive.

An action’s plan may include differ-ent “routes” the agent can follow in orderto complete the action. Even if it is unique,the plan’s execution can provide differentsolutions for the corresponding action, de-pending on the current context.

For example, in order to assist thechild, the tutor can tell a sentence, or recalla grammar rule, or highlight the error. An-other way to assist the child is to tell a sen-tence and highlight the error, and then torecall a grammar rule, depending on thechild responses. We use three behaviouraloperators, “all,” “first of” and “sequence”in an action plan expression. To briefly de-scribe them, let us denote by wait the ac-tion with any arbitrary effect (i.e., any setof resulting stimuli) on the agent’s state/structure and which is by default accom-plished, and by none the action that is neveraccomplished. We will denote by A the setof an agent’s actions, by A* the set A-{wait}, and by Time a linear temporal struc-ture assumed to be discrete for the sake ofsimplicity.

We define the ALL pattern by meansof the binary operator “all” ⊗:A2→A withthe following semantic: we say that an ac-tion A

res=A

1⊗A

2 is completed, and so the

associated context is validated if∃t

1>t

0∈Time for which both A

1 and A

2 are

completed in the moment t1. Here t

0 de-

notes the moment of parallel activation of

the actions. Its associated FCM is given inFigure 12.

Here, the set of FCM’s concepts willcorrespond to the action’s components,actions themselves. For each of the agent’seffectors there will be one action that theeffector controls. This means that an acti-vated action will not activate its influencedactions (i.e., the actions activated by thecurrent one in the FCM) until the corre-

+1 +1 +1/2 +1/2

A1 A2

Ares started

effectori

effectorj

Ares completed

Figure 12: ALL Pattern’s FCM

+1 +1 +1 +1

A1 A2

Ares started

effectori

effectorj

Ares completed

Figure 13: FOF Pattern’s FCM

+1 +1

+1

A1

A2

Ares started

effectori

effectorj

Ares completed

Figure 14: SEQ Pattern’s FCM



sponding effector completes its job. In Fig-ure 12, we have associated the effector

i

to the action A1 and the effector

j to the

action A2. With ALL pattern we can ex-

press parallel cooperative actions (non-se-quential).

Using the same FCM structure butwith different influence values we obtainanother binary operator, the “first of”⊕:A*2→A, which gives us the FOFbehavioural pattern. Its semantic is the fol-lowing: we say that an action A

res=A

1⊕A

2

is completed if ∃t1>t

0∈Time and ∃j=1,2

such that Aj is completed at the moment t

1

and ∀k=1,2, and Ak is not completed be-

fore, that is, at any moment t, t0<t<t

1 (Fig-

ure 13). In this case, the context consistsin the first accomplished action’s effect.We use the FOF pattern when parallel con-current actions (non-sequential) areneeded.

When an order between the plan’sactions is needed, we can use yet anotherpattern, the “sequence” one (SEQ), de-fined by the operator Θ:A2→A. We say thatan action A

res=A

1ΘA

2 is completed if

∀j=1,2 ∃tj>t

0∈Time and t

j>t

j-1 with the

property that Aj is completed starting with

the tj and A

j+1 is activated at t

j+1 (Figure

14). Here t0 represents the moment of the

activation of the action A1. In other words,

the actions are activated and completed inthe indicated sequence.

By constraining the action(s) comple-tion into a time interval, we obtain thebounded versions of the above defined ALL,FOF, and SEQ operators, as vanLamsweerde suggests (Lamsweerde &Letier, 2000). We have found useful, in theaction’s plan expression, the use of n-aryextensions of these binary operators.

The action’s activation correspondswith forcing A

res started concept to 1 in the

action’s plan. This means that the execu-tion of the action’s plan produces the acti-

vation of component actions. In their turn,the corresponding effectors will be trig-gered.

If a component action fails, the wholeplan fails in the case of ALL and SEQ pat-terns. By contrary, if the failed action par-ticipates in a FOF pattern, the plan is stillactive, and another action is waiting to suc-ceed.

In the case of use of bounded ver-sion of above-mentioned operators, theaction plan is ended if the plan is accom-plished in the specified time interval, orstopped otherwise. More, at any moment,if the action’s context became non-valid,the action is stopped.

Based on the action’s plan, we areable to obtain different evaluations of itsaccomplishment level, either by countingits completed actions or by the means ofthe objective completed concept value.The greater this number, the greater maybecame the objective/action priority fromthe agent’s perspective.

The action’s corresponding FCMconvergence to a fixed-point attractor as-sures the action’s accomplishment or fail-ure, depending on the value of completedconcept at this point.

In our notation, we can express ourtutor agent’s global objective as O=ALL(AC,F,ASC), where we have noted by ACthe “avoid collision,” by F the “followthe child” and by ASC the “assist thechild” objective. According to the proposedoperators semantic, the tutor will evaluateand then activate the three objectives, basedon their corresponding values in the FCM.

If there are some concurrent actionswith similar effects, then they are let tocooperate; otherwise the action with lowerpriority is made inactive. In our case, theonly concurrent actions may be AC and F.For this, the value of the corresponding driv-ing concept for AC, respectively for F, is



given by the inverse of the distance to theobject the tutor will try to avoid, respec-tively the child the tutor is following. Afterthe currently active action will be accom-plished, the rest of the actions will be re-evaluated and activated accordingly.

We have to remind that the associ-ated effectors control the values propaga-tion through a fuzzy plan. This makes themdifferent than but as simple as classicalFCMs.

IMPLEMENTATION

Virtual environment and its compo-nents have been developed with VirtualReality Modeling Language (VRML) ISOstandard. Users may view 3D contentswith a Web browser and a VRML plug-in.The actual implementation is using CortonaVRML plug-in from ParallelGraphics. Theenvironment distribution is based on theDeepMatrix software (Reitmayr, Carroll,Reitmeyer & Wagner, 1998) fromGEOMETREK. This software enablesusers to enter 3D Web sites where theycan interact with other users and objects.DeepMatrix implements client-server ar-chitecture.

On the server side, all messages arebroadcasted in the same order to all cli-ents. We have refined the proposed imple-mentation from GEOMETREK, by intro-ducing a filtering and pseudo-dead reckon-ing mechanism, which permit a morefriendly and flexible connection of youngusers.

On the client side, interfaces havebeen implemented in Java language with aspecial attention due to the fact that usersare viewing the HTML pages with InternetExplorer and running the Microsoft VirtualMachine. The use of External AuthoringInterface (EAI) permits the client toachieve complex tasks by connecting the

VRML Web browser plug-in with a Javaapplet within the same Web page (Figure15). The Java applet loads VRML contentinto the plug-in and adds avatar represen-tation to the virtual world. The plug-in up-dates the Java applet about users’ positionand orientation in the virtual world.

There is one more type of informa-tion that is published between all the cli-ents of a virtual world on DeepMatrixserver, the other shared information, likepictures, sounds, and strings. Generally,everything is declared as shared node inVRML files. By doing this, we can reachan optimised solution for a distributed vir-tual environment, with non-shared informa-tion downloaded locally.

VR Agents

The virtual agents were designed onan approximate-body approach (Thalmann,2000), which frequently provides positionand orientation information to remote hosts,taking into account a minimal set of jointpoints. Their geometrical structure com-ponents were implemented in VRML us-ing a PROTO structure. Behaviours weredesigned with the 3DStudioMax software,and then exported to VRML format. Theresulting code was finally inserted into theagent’s PROTO structure.

The environment dynamics is per-ceived by the agent itself through its Per-ception module (Figure 9). Based onVRML sensor-like components, this mod-ule uses VRML eventIn mechanism in or-der to transmit the detected stimuli to theagent’s decisional module. And because theagent itself has VRML components in theVRML environment, using these sensorsit is able to realise the exteroception as wellas the interoception. Once the decision ismade by the agent’s Decisional module,which was implemented using Java



classes, it sends execution commands toits Action module. This time, we are usingthe VRML eventOut mechanism in orderto activate the agent’s behaviours (Figure15).

The implemented behaviours arewalk, turn left, turn right, standby, look at,laughing, drag/drop object, greetings suchas “Hello!”, “Good Bye!”, “Help me!”,“Thank you!”, “I agree” and “I disagree”.These VR agent’s atomic actions formagent’s capabilities and are used in the ac-tion plan realisation. The child may selecta behaviour, which will be broadcasted (ges-ture and sound) to other children throughits personal avatar.

In order to assure the EVE’s agentsadaptability to their dynamic environmentwe are using the following plans:

AgentGoal=ALL(Play,AvoidCollision,Help)Play=SEQ(Room

1Completed,Room

2Completed)

Room1Completed=ALL(Sentence

1,Sentence

2,Sentence

3)

Sentencei=SEQ(ALL(Label

1,…,Label

5),Valid_Sentence

i)

Labelk=action to place the Label

k into a label

socket from agent’s walls.Valid_Sentence

i=is an example of requested

action (which depends on other agents’ ac-tions), in our case on tutor decision on pro-posed Sentence

i, which is part of tutor’s Assist

plan.

Figure 15. Client-server architecture

Figure 15: Client-Server Architecture

Figure 16: Virtual Tutor in EVE Environment



Room2Completed=SEQ(ALL(Img1,…Img

9),Valid_Story)

Imgi= action to place the Image

i into an image

socket from common room green wall.Valid_Story=is another tutor action which par-ticipate as requested action in a fuzzy plan.AvoidCollision is considered as an atomicaction. For its corresponding FCM and use ofother atomic actions (Figure 11).

Help=FOF(Ask_for,Give)Ask_for = explicit demand of other agents tohelp the current one in its task.Give = SEQ(Approach,Label) means that theagent will try to approach to the agent who hassent the Ask_for demand. After that, it will try toplace a Label into the right socket. This time,the action plan components are demandedactions for all the other agents, except the agentasking for help.

This is the most frequent situation inwhich the cooperation between agents ap-pears. Once an agent sends the Ask_forhelp demand, all the agents in its environ-ment will receive this demand under theform of an external stimulus and will acti-vate their Give action plan. Due to theenvironment’s dynamic (agents positions,access paths, obstacles, etc.), only one ofthese agents will succeed to accomplish thistask. The other agents will be informedabout this and they will skip the Give planand will decide on the next action accord-ing to environment context. This way, wehave eliminated the obsolete agent’s ac-tions due to environment changes duringthe time interval between agent’s actiondecision and action completion.

More, using virtual agents we are ableto ensure the persistence of the game. Letus suppose that there is a child who wantsto quit the game before its end. In this case,without a virtual agent who takes the child’splace, the game will be blocked. But usinga virtual agent, the child can quit the game

without affecting its flow, its role beingswitched to the corresponding virtual agent.

Virtual Tutor

Because of the number and distribu-tion of children that are playing the game,the teacher is unable to assist them all thetime. More, some children may need addi-tional assistance from their teacher. Forthis, we have introduced the virtual tutor inEVE environment (Figure 16). The role ofvirtual tutor is not to substitute the realteacher but to encourage the child to testthe environment even in the absence of itsreal teacher.

A virtual tutor is basically a virtualagent which has as objective to assist thechild in its training. He or she can do thisby recalling one sentence the child has toreconstruct or the whole story. He or shecan do all this at user’s demand or he/shecan decide to act based on user actions(label’s erroneous movements). More, heis able to move, avoid collisions, salute, con-firm or infirm user’s choices.

The tutor’s plans are extension ofthose described in the previous subsection:

TutorGoal=ALL(Goal,AvoidCollision,Assist)Assist=SEQ(Approach,FOF(GiveAdvice,Valid_Sentence,Valid_Story)), (Fig.11).GiveAdvice=FOF(Say_sentence, Say_story),depending on the assisted agent actions, thetutor will give some suggestions.

In the tutor’s Goal plan, the actionslike RoomCompleted, Sentence or Labelare requested actions for the existing agents.The teacher may use the virtual tutor’savatar and so it will be able to realise re-mote assistance for the children using EVE.In this case, the virtual tutor’s behaviourwill be much more complex from a deci-sional point of view, because there will be



a real person who interacts, but will stillfollow the actions plan.

CONCLUSION &FUTURE WORK

In this paper we have presented avirtual environment specially dedicated topedagogical purposes. This virtual environ-ment is accessible with a standard Webbrowser and VRML plug-in without anyspecific or additional software. The appli-cation is successfully in use as a comple-ment of traditional learning exercises inprimary schools from France, Morocco,Romania and Japan. Such exercises alreadyexist using scissors and paper. But the dif-ference here is twofold:• Children get an immediate validation of

their works; for example an animal ap-pears when the sentence is correct;

• Children do not realise that they areworking; in fact they feel as if they wereplaying together.

Experiments have pointed out thatgames increase children’s motivation, andnew technologies such as virtual reality in-crease children’s autonomy. More, they areable to take over the virtual world and tomove into the virtual space. For example,after a short time of practice, children havetransposed into the virtual world one of theirreal games: “hide-and-seek”. The coopera-tive work demonstrates that children in amulticultural framework are able to worktogether if they comply with common rules.It is important to remember that a com-puter is only a machine, which is pro-grammed to provide positive encourage-ments and congratulations so that childrencan consider it as a friend. On the one hand,it encourages children to be more active sothat they dare answer something; it enables

children to make the exercise as manytimes as necessary. On the other hand, itenables children to go faster and then in-crease the number of exercises.

The second game (Figure 3) is meantto help with text understanding and logic.Generally children have difficulties read-ing and to correcting themselves. They of-ten make the same mistakes. Here, it isimpossible because they have to changesomething if they are wrong; they have tobuild a new reasoning. This points out thatour application contributes to increasing theability of children to reason by themselves.

On a pedagogical point of view,planned for the primary courses, EVE isnow adaptable for older children. Theteacher may pass from a text comprehen-sion to grammar or syntax studies. More,new applications should be developed inother domains of interest such as: foreignlanguages, geometry, algebra, and others.A special benefit of the authoring tools liesin the fact that the teacher can easily cre-ate new exercises. Especially, the teachercan design exercises that fit with children’sinterest. Consequently, it strengthenschildren’s attention to exercises contents.As stories are shared between classroomsor schools, it increases the quantity of ex-ercises teachers can propose to children.

On a technical point of view, currentexperiments will help us to design andimplement new avatar behaviours accord-ing to end users’ needs. We are also work-ing on a more sophisticated mechanism,which is the integration of streaming audioand video into the virtual world. This shouldincrease the performance of children’s co-operative work. A new direction would beto use a platform independent interface,ARéVi. Doing so, we could reach a totalindependence together with a higher per-formance in 3D rendering. More, we areinterested in emotional and social aspects



of agents’ behaviour. Validation of criticalaction plans, like cyclic ones, is one of ourpriorities. For short term, we intend to letthe EVE’s tutor assist more children in thesame time, to participate into the votingmechanism, or even to make some mis-takes in order to supplementary test thechildren, as a higher level of children’spedagogical evaluation.

To conclude, we believe that virtualreality is likely to enable the developmentof new products that will help children andteachers in their tasks of learning and teach-ing through friendly interfaces. Finally, co-operative experiments pointed out that suchcooperative work introduces a kind of sub-liminal target, which is: “training children inlearning democracy”.

ACKNOWLEDGMENTS

Thanks are due to the Intergovern-mental Agency of the Francophonie, whichfunded the EVE project in the frameworkof the Francophone Information HighwayFund, the Brittany Regional Council, whichfunded this work in the framework of theMEGALIS Project, and the Hewlett-Packard Company. Thanks are also due toprimary schools Kroas Saliou, Plouzané,France, Dan Barbilian, Constanta, Roma-nia, El Khalil, Casablanca, Morocco, andOotsukadai elementary school, Yokosuka,Japan, and companies Virtualys, Brest,France, Impromex, Constanta, Romania,and TBE MAROC, Rabat, Morocco in-volved in the project.

REFERENCES

Bates, J. (1992). Virtual reality, art andentertainment. PRESENCE: Tele-op-erators and Virtual Environments,1(1), 133-138.

Benford, S.D., & Fahln, L.E. (1993). Aspatial model of interaction in large vir-tual environments. Proc. ECSCW’93(pp. 109-124). Milano, Italy: KluwerAcademic.

Bobick, A., Intille, S., Davis, J., Baird, F.,Pinhanez, C., Campbell, L., Ivanov, Y.,Schutte, A., & Wilson, A. (2000). TheKidsRoom - A perceptually-based inter-active and immersive story environment.Presence: Teleoperators and VirtualEnvironments, 8(4), 367-391.

Bruner, J. (1986). Actual Minds, PossibleWorlds. Cambridge, MA: Harvard Uni-versity Press.

Cavazza, M., Charles, F., & Mead, S.J.(2002). Interacting with virtual charac-ters in interactive storytelling. ACM JointConference on Autonomous Agentsand Multi-Agent Systems (pp. 318-325).Bologna, Italy.

Druin, A. (1999). Cooperative inquiry: De-veloping new technologies for childrenwith children. Proceedings of HumanFactors in Computer Systems (chi99)(pp. 223-230). ACM Press.

Druin, A., Revelle, G., Bederson, B.,Hourcade, J., Farber, A., Lee, J., &Campbell, D. (2003). A collaborativedigital library for children: A descriptivestudy of children’s collaborative behav-iors and dialogue. Journal of Computer-Assisted Learning, 19(2), 239-248.

El-Nasr, M.S., Yen, J., & Ioerger, T. (2000).Flame - A fuzzy logic adaptive model ofemotions. Autonomous Agents andMulti-Agents Systems, 3, 219-257.

Hayes-Roth, B., & vanGent, R. (1997).Story-making with improvisional pup-pets. Proceedings of First Interna-tional Conference on AutonomousAgents (pp. 1-7).

Hefco, V. (1997). The Physiology of Ani-mals and Humans (in Romanian).Bucharest, Romania: Ed.D.P.



Herrero, P., & Antonio, A.D. (2002). Ahuman based perception model for co-operative intelligent virtual agents.(COOPIS’02) 10th International Con-ference on Cooperative InformationSystems (pp. 195-212). Irvine, CA,USA.

J@ardin (2004). Jardin Botanique deMontréal. Online: http://www.ville.montreal.qc.ca/jardin/jeunes

Johnson, A., Moher, T., Cho, Y., Lin, Y.,Haas, D., & Kim, J. (2002, March/April). Augmenting elementary schooleducation with vr. IEEE ComputerGraphics and Applications, 6-9.

Johnson, A., Moher, T., Leigh, J., & Lin, Y.(2000). Quickworlds: Teacher driven VRworlds in an elementary school curricu-lum. Proceedings of SIGGRAPH 2000- Educators Program, New Orleans,LA.

Johnson, A., Moher, T., Ohlssoon, S., &Gillingham, M. (1999). The Round EarthProject - Collaborative VR for concep-tual learning. IEEE Computer Graph-ics and Application, Virtual Reality,19(6).

Kosko, B. (1997). Fuzzy Engineering.Prentice Hall.

Lamsweerde, A. van, & Letier, E. (2000).Handling obstacles goal oriented require-ments engineering. IEEE Trans. onSoftware Engineering, 26(10), 978-1005.

Mantovani, F. (2001). VR learning: Poten-tial and challenges for the use of 3Denvironments in education and training.Towards CyberPsychology: Mind,Cognitions and Society in the InternetAge.

Marshall, P., Rogers, Y., & Scaife, M.(2002). Puppet: A virtual environmentfor children to act and direct interactivenarratives. Second InternationalWorkshop on Narrative and Interac-

tive Learning Environments,Edinburgh.

Mateas, M. (1997). An OZ-centric viewof interactive drama and believableagents. (Tech. Rep.), CMU-CS-97-156.

Montemayor, J., Alborzi, H., Druin, A.,Hendler, J., Pollack, D., Porteous, J.,Sherman, L., Afework, A., Best, J.,Hammer, J., Kruskal, A., Lal, A.,Plaisant-Schwenn, T., Sumida, L., &Wagner, R. (2000, April). From pets tostorykit: Creating new technology withan intergenerational design team. Work-shop on Interactive Robotics and En-tertainment (WIRE-2000), Pittsburgh.

Pantelidis, V. (1995). Reasons to use vir-tual reality in education. VR in Schools,1(1), 9.

Parenthoen, M., Reignier, P., & Tisseau, J.(2001). Put fuzzy cognitive maps towork in virtual worlds. Fuzz-IEEE01,Melbourne, Australia.

Piaget, J. (1929). The Child’s Conceptionof the World. New York: Harcourt–Brace.

Popovici, D.M., Serbanati, L.D., & Gerval,J.P. (2003). Agent-based modeling ofvirtual environments. Proceedings ofVRIC 2003 (pp. 149-158). Laval,France.

Popovici, D.M., Serbanati, L.D., &Harrouet, F. (2003). The virtual environ-ment - Another approach. WSCG’2003Posters Proceedings (pp. 109-112).Plzen, Czech Republic.

Prada, R., Machado, I., & Paiva, A. (2000).TEATRIX: Virtual environment for storycreation. Intelligent Tutoring Systems(pp. 464-473).

Reitmayr, G., Carroll, S., Reitmeyer, A., &Wagner, M. (1998, October 30).Deepmatrix - An Open TechnologyBased Virtual Environment System.

Robertson, J. (2002). Computer games -For better or for worse? i3 Magazine,



12, 32-34.Roussos, M., Johnson, A., Moher, T., Leigh,

J., Vasilakis, C., & Barnes, C. (1999).Learning and building together in animmersive virtual world. Presence, 8(3),247-263.

Thalmann, D. (2000). Challenges for theresearch in virtual humans. WorkshopAchieving Human-like Behavior in In-teractive Animated Agent, Barcelona,Spain.

Tisseau, J., & Harrouet, F. (2003).Autonomie des entités virtuelles. In P.Fuchs (Ed.), Le traité de la réalité

virtuelle, 2, 85-124. Les Presses del’Ecole des Mines, Paris.

Volbracht, S., Domik, G., Backe-Neuwaldand, D., & Rinkens, H. (1998).The ¨City Game¨ an example of a vir-tual environment for teaching spatial ori-entation. Journal of Universal Com-puter Science, 4(4), 461-465.

Winn, W. (1993). A conceptual basis foreducational applications of virtualreality. Technical Report TR 93-9.

Youngblut, C. (1998). Educational usesof virtual reality technology. Techni-cal Report IDA Document D-2128.

Dorin-Mircea Popovici is a PhD student in Computer Science at “Politehnica” University atBucharest, and a lecturer at OVIDIUS University (Constanta, Romania). His PhD subject con-cerns the space modeling in virtual universes. He is interested in computer graphics, computa-tional geometry, virtual space modeling, virtual physiology and multi-agent environments.Since 2003, he has been an associated researcher at CERV as manager of EVE project.

Jean-Pierre Gerval obtained his Doctorate in Automation from the University of Valenciennesin France (1987). He has been project manager at the “Institut d’Informatique Industrielle” inBrest (1988-2003) and associate professor of Computer Sciences at the “Ecole Nationaled’Ingénieurs de Brest” (1993-2003). He is currently head of the Computer Sciences Depart-ment at the “Institut Supérieur de l’Electronique et du Numérique” in Brest. His researchinterests include distributed virtual reality and virtual environments especially dedicated topedagogical applications.

Pierre Chevaillier is a professor assistant in Computer Science and works at the CERV. Heworks on the use of the multiagent paradigm to design digital simulations where users have toplay an active and collaborative role. He is the leader of the MASCARET project and focuses itsresearch activities on the interaction and organisational facets of heterogeneous multiagentssystems (where both artificial and human agents are involved).

Jacques Tisseau is a professor in Computer Science and is head of LI2/ENIB and CERV. Hisdomains of interest are entity’s autonomy in virtual environments, and “in virtuo” experimen-tation in the modeling processes.

Luca-Dan Serbanati is a professor of Computer Science at the Ovidius University of Constantaand an IT adviser, teacher, and mentor for industry in Italy. He is the author of the bookIntegrating Tools for Software Development, where multi-facetted models for the software pro-cess are introduced. Serbanati received a PhD in Computer Science from the PolitehnicaUniversity of Bucharest. His research interests address systems model-driven analysis anddesign, formal models, distributed systems, and software development paradigms.



Patrick Gueguen is a primary school teacher. He is the headmaster of Plouzané Primary School,France. One of his main targets is to open his school and encourage children to use newtechnologies. He takes part in the pedagogical aspects of EVE project, being involved in it fromits beginning.

Educative Distributed Virtual Environments for Childrentisseau/pdf/paper/jt-art-int-7.pdf ·...

Documents

Transcript of Educative Distributed Virtual Environments for Childrentisseau/pdf/paper/jt-art-int-7.pdf ·...