INTEGRATING KNOWLEDGE FROM VIRTUAL REALITY

ENVIRONMENTS TO LEARNING SCENARIO MODELS

A Meta-modeling Approach

Nicolas Marion, Ronan Querrec and Pierre Chevaillier

LISyC, ENIB, UEB

CERV, 29280 Plouzan´e, France

Keywords:

Learning scenario, VRLE, IMS-LD, Meta-modeling.

Abstract:

This paper focuses on learning scenario modeling for Virtual Reality Learning Environments (VRLE). Learn-

ing scenario models used in computer-supported learning environments are usually not able to describe ed-

ucational activities implying interaction of learners with a virtual environment. In this paper, we propose

an IMS-LD extension making it possible to describe executable educational activities taking place in virtual

worlds. Moreover, the model described in this paper is generic in the sense that it can create scenarios regard-

less of the nature of virtual environments or application domain.

1 INTRODUCTION

A learning scenario describes educational activities

taking place during a training session, and is usually

written by the trainer. This paper focuses on learning

scenarios modeling for sessions where some learning

activities use virtual reality learning environments

(VRLE). A VRLE is a computer-supported learning

environment (CSLE) that uses virtual reality (VR)

technology in order to immerse learners in a virtual

environment. In the context of learning scenarios

modeling, the speciﬁcity of VRLE in comparison to

classic CSLE is related to the nature of activities that

take place in the scenario, that imply actions of actors

in a virtual environment.

Educational modeling languages that describe learn-

ing scenarios usually deﬁne ﬁve types of information

(Koper, 2001):

• Prerequisites: prerequisites describe knowledge

or skills the learner should have in order to take

advantage of the learning scenario.

• Learning objectives: learning objectives are

knowledge or skills to be gained by learners that

achieve the learning scenario.

• Activities: it consists in the description of activi-

ties that can be performed in the environment by

the different actors, learners or teachers, and their

scheduling.

• Roles: they describe the involvement of the users

in the learning scenario, and activities they have

to perform.

• Environments: an environment describes the

context of execution of educational activities. It

contains necessary resources to the execution of

activities.

Existing learning scenario models in the domain of

CSLE (Koper et al., 2003; Rodr´ıguez-Artacho and

Verdejo Ma´ıllo, 2004), generally consider educational

activities as “black boxes”, described by a textual de-

scription. Activities are considered atomic, in the way

that their execution is not described in the scenario.

Only their inputs (resources) and outputs (outcomes)

are taken into account.

On the other side, most VRLE include an author-

ing tool that allows the description of learning activi-

ties execution in a virtual environment (Munro, 2003;

Gerbaud et al., 2008). Learning scenario models in-

cluded in these authoring tools usually have two prob-

lems. 1) They don’t describe the integration of the

activity in virtual environment into a more global ed-

ucational process. 2) They are generally not reusable

because of their speciﬁcity in relation to a particular

254

Marion N., Querrec R. and Chevaillier P. (2009).

INTEGRATING KNOWLEDGE FROM VIRTUAL REALITY ENVIRONMENTS TO LEARNING SCENARIO MODELS - A Meta-modeling Approach.

In Proceedings of the First International Conference on Computer Supported Education, pages 253-258

DOI: 10.5220/0001976102530258

 SciTePress

domain, like maths (Koedinger et al., 2004), to a par-

ticular task, like assembly operations (Brough et al.,

2007), or to a particular pedagogical strategy, like dis-

covery learning (Van Joolingen and de Jong, 2003).

Our learning scenario model proposition is

grounded on the observation made by (Gu´eraud and

Cagnat, 2006) that learning scenarios describing situ-

ations in which learners interact strongly with interac-

tive educational tools (like virtual environmentsin our

context) must provide information on the internal pro-

gression of educational activities. However, (Burgos

et al., 2007) pointed out the difﬁculties encountered

when trying to integrate simulations or games in an

executable learning scenario. A full integration of the

simulation in the learning ﬂow should make it possi-

ble to receive information from the learning scenario

(initial conﬁguration, environment modiﬁcation...)

and to send information generated during its execu-

tion (actions performed, state of the world...). The

constraint it brings is that a speciﬁc wrapper, linking

the simulation and the learning scenario, should be

created for each simulation. Then, evenif the learning

scenario model is as generic as classical CSLE mod-

els, we encounter the same problem as with VRLE

authoring tools because the integration of the virtual

environment requires speciﬁc development.

There is a need to provide models and tools allow-

ing trainers to create learning scenarios integrating ac-

tivities in virtual environment, regardless of the nature

of those environments, and directly executable on a

virtual reality platform. The learning scenario model

we propose is based upon IMS-LD (Koper et al.,

2003), a speciﬁcation that has been normalized and

widely accepted in the ﬁeld of CSLE. IMS-LD allows

to write scenarios describing the ﬁve types of infor-

mation presented in introduction (prerequisites, learn-

ing objectives, activities, roles and environments).

In order to fully integrate a virtual environment

in the learning ﬂow, a both-way communication is

necessary so that: 1) the virtual environment can be

adapted to the proﬁle of learners and activities they

previously performed and 2) the learning ﬂow can be

modiﬁed based on what happens in the virtual en-

vironment in real time. It is thus necessary to ex-

tend IMS-LD so that the learning ﬂow can take into

account learners activity in the virtual environment.

In section 2 we present the abstraction level of the

model, and more precisely the virtual environment

meta-model that it uses. In section 3 we show how

IMS-LD can be extended to integrate activities in vir-

tual environments in learning scenarios.

2 MODEL’S ABSTRACTION

LEVEL

As previously said, the integration of a virtual en-

vironment in the learning ﬂow requires a both-way

communication between the learning ﬂow and the vir-

tual environment. To enable this communication, the

learning scenario must be able to reference concepts

used in the virtual environment. To do so, an approach

consists in describing all those concepts in a domain-

speciﬁc model of the application. This domain-

speciﬁc model gathers information about types of en-

tities that can be found in the environment, their prop-

erties, behaviors, etc. Thus, a learning scenario linked

to such a model can express properties about the en-

vironment using domain-speciﬁc concepts. At this

point, two abstraction levels can be identiﬁed: the vir-

tual environment (level 0) and the model of this envi-

ronment (level 1). In order for the learning scenario

model not to be speciﬁc to one domain, the concepts

of this domain (level 1) must be considered as inter-

changeable input data of the learning scenario model.

For the learning scenario model to be able to han-

dle these data, it uses a meta-model (level 2) that de-

scribes concepts used in domain-speciﬁc model.

To sum up, a learning scenario takes place in a vir-

tual environment (level 0) and references the model

of this environment (level 1). The scenario is ex-

pressed using a learning scenario model that uses the

virtual environment meta-model (level 2). Figure 1

shows the links between learning scenarios, learning

scenario model and the different virtual environments

modeling level.

Figure 1: Representation of links between learning scenar-

ios, learning scenario model and the different virtual envi-

ronments modeling level.

This meta-modeling approach can be compared with

the content layer of the PALO model (Rodr´ıguez-

Artacho and Verdejo Ma´ıllo, 2004). Unfortunately,

the meta-model they use contains only two concepts

(entity and relationship), and is not expressiveenough

to describe complex virtual environments. For exam-

INTEGRATING KNOWLEDGE FROM VIRTUAL REALITY ENVIRONMENTS TO LEARNING SCENARIO

MODELS - A Meta-modeling Approach

255

ple, it cannot describe the dynamic behavior of enti-

ties, or actions that can be performed by users in a

virtual environment.

The MASCARET (Buche et al., 2004) (Multi-

Agent System for Collaborative, Adaptive and Re-

alistic Environment for Training) project is intended

to design realistic VRLEs. MASCARET contains sev-

eral models that describes different parts of VRLE. In

this paper, we describe two models that are used by

the learning scenario model: the virtual environment

meta-model and the organisational model.

Virtual Environment Meta-model: VEHA

VEHA (Marion et al., 2007) is a meta-model designed

to deﬁne virtual environments, providing a semantics

allowing artiﬁcial or human agents to build a repre-

sentation of it.VEHA meta-model is based upon UML

2.1

. Figure 2 presents a overview of VEHA.

Figure 2: A part of VEHA meta-model.

Concepts of domain-speciﬁc models are represented

by instances of the

Class

class. The structural

(

Property

) and Behavioral (

BehavioralFeature

)

properties of classes are associated with

Class

via the

Feature

class. Instances of these classes (entities ac-

tually evolving in the environment) are deﬁned by the

InstanceSpecification

class.

To explicit knowledge about virtual reality, VEHA

extends UML meta-model adding classes making it

Uniﬁed Modeling Language 2.1:

http://www.omg.org/docs/formal/07-11-01.pdf

possible to model objects that are geometrically rep-

resented in the virtual environment, and which are

therefore controllable by the user. A VEHA environ-

ment is a set of objects with graphical representations.

The

Entity

class can represent such objects. Enti-

ties are instances of a particular class:

EntityClass

It owns not only

Features

, but also information re-

lated to the shape, geometry, positioning, etc.(not rep-

resented on the ﬁgure).

VEHA provides an informed virtual environment

meta-model that explicits characteristics of environ-

ments and entities that compose them. In addition, it

allows the introspection of the domain-speciﬁc model

(level 1), as well as the virtual environment (level 0).

This meta-model can be used by a learning scenario

model to express a learning scenario that references

concepts of the domain. Thus, the scenario can de-

scribe properties of the environment that have to be

observed by the learning ﬂow during learners activity,

as well as modiﬁcations to bring to the environment,

based on the progress of the learning ﬂow execution.

MASCARET’s Organisational Model

MASCARET allows the creation of virtual environ-

ments in which humans and virtual agents can in-

teract by playing roles within organisations. This is

done via an Agent–Role–Organisation structure, de-

scribed in (Buche et al., 2004). It is important for the

learning scenario to reference organisational entities

that exist in a virtual environment, in order to make

it possible for users of the learning scenario to play

speciﬁc roles in the virtual environment. The learn-

ing scenario model proposed in this article uses MAS-

CARET, more precisely its virtual environment meta-

model (VEHA) and its organisational model. The

scenario model allows to create scenarios describing

learning activities in virtual environments, as long

as those virtual environments are described in MAS-

CARET. MASCARET has already been used to create

virtual environments for learning like GASPAR (Mar-

ion et al., 2007), S

ECUR

EVI (Querrec et al., 2003)

and a physics lab work (Baudouin et al., 2008).

3 LEARNING SCENARIO MODEL

In this section, we present how IMS-LD can be ex-

tended to integrate activities taking place in virtual

environments. This section contains three parts, de-

scribing three aspects of learning scenarios that need

to be extended to integrate activities in virtual envi-

ronments: the pedagogical organisation model (3.1),

the property model (3.2) and the environment model

(3.3).

CSEDU 2009 - International Conference on Computer Supported Education

256

3.1 Pedagogical Organisation Model

This part focuses on the organisation of the different

roles that perform activities in the scenario. In IMS-

LD, every information of the scenario is contained in

the

learning design

. A learning design contains

a set of roles. Two types of role exist: learner roles

and staff roles. In the scenario, roles are associated

to activities that represent activities that users play-

ing that role have to perform. Proposed model adds

the notion of pedagogical organisation. This notions

is based on MASCARET’S organisational model. Fig-

ure 3 presents the pedagogical organisation model we

propose, as well as the part of MASCARET on which

it is based.

Figure 3: Pedagogical organisation class model.

Every

LearningDesign

contains one pedagogical or-

ganisation. Pedagogical organisations, are deﬁned

using the notion of organisation described in MAS-

CARET. A pedagogical organisation is composed of

pedagogical roles. In comparison to a MASCARET

role, a pedagogical role adds two information:

• a link to one or many roles of the virtual environ-

ment (

rolesInWorld

). This linked can be used

by the trainer to specify that a user will play one

or several roles of the virtual environment (in ad-

dition to his pedagogicalroles) in order to perform

a collaborative activity.

• a link to one or several instances of the

Person

class, that represent the users of the scenario

that will play this role at run-time, and perform

associated activities. These persons can be

humans or virtual agents.

This extension of IMS-LD organisational model al-

lows the learning scenario to take into account the or-

ganisational structure of the virtual environment. The

trainer can associate a pedagogical role to one or sev-

eral roles in the virtual environmentso that users play-

ing that pedagogical role are automatically associated

to the corresponding role in the virtual environment

for a speciﬁc activity. an trainer or an intelligent tu-

toring sustem (ITS).

3.2 Properties Model

The principle of properties is that the trainer can de-

scribe variables on which tests can be done, and de-

scribe the progress of the scenario based on the results

of those tests. This mechanism adds ﬂexibility for the

trainer to describe the scheduling of the scenario.

In IMS-LD speciﬁcation, a property represents a

variable that can have a type and a value. Properties

deﬁned by IMS-LD can have different scopes: local

(same value for all users in a scenario), global (same

value for every user in every scenario), personal (dif-

ferent value for every user) or role (same value for

every user playing the same role).

In order to make it possible for the trainer to take

into account the virtual environment in the learning

scenario, it is necessary to extend the deﬁnition of

property made by IMS-LD. In the learning scenario

model, a property can not only represent a variable

as in IMS-LD, but also an entity’s property, an en-

tity’s state-machine or the fact that an action has been

performed in the virtual environment. The properties

model is based on UML’s properties model, already

implemented in VEHA. Figure 4 represents the class

diagram of properties model.

Figure 4: Properties model of proposed learning scenario

model.

This ﬁgure shows four types of property:

• Properties as deﬁned by IMS-LD are represented

by the

Variable

class. A variable contains a

VEHA

Property

, that deﬁnes the data type and

initial value of the variable. During run-time, one

or many slots (property instances) are instantiated

INTEGRATING KNOWLEDGE FROM VIRTUAL REALITY ENVIRONMENTS TO LEARNING SCENARIO

MODELS - A Meta-modeling Approach

257

for each variable (depending on the type of the

variable; for example, for a personal value will be

instantiated one slot by user).

• An

EntityPropertyRef

represents a link to

a property of an entity in a virtual environ-

ment. The difference with a

Variable

is that an

EntityPropertyRef

references an existing slot,

that belongs to an instantiated entity of the virtual

environment.

• A

StateMachineRef

represents a link to a state-

machine of an entity in a virtual environment. The

value of such a property is the current state of the

referenced state machine.

• An

ActionProperty

allows to know if a speciﬁc

action has been performed in the virtual environ-

ment.

Actions

are deﬁned in the domain-speciﬁc

model and describe the set of actions that can be

performed by actors (human or virtual) in the vir-

tual environment.

ActionProperty

model is not

detailed in this paper.

Those four types of property behave the same way

in that that we can retrieve their value or modify

their value, with the exception of

ActionProperties

from which we can only retrieve the value.

The properties deﬁned by the trainer are used to

deﬁne conditions in the scenario. These conditions

associate an expression (deﬁned by a set of proper-

ties and their values combined with boolean opera-

tors) and a pedagogical action.During run time, con-

ditions are tested by the platform, and pedagogical

actions associated with true conditions are triggered.

As in IMS-LD, pedagogical actions can be: change

the value of a property, send a notiﬁcation or change

the visibility of a component of the scenario (make

available or unavailable activities, environment, re-

sources, ...). In addition, we added a new action

called

VRAction

. The

VRAction

class represents a set

of pedagogical actions that can be applied to a virtual

environment, given a certain context. The various ac-

tions are based on the list of pedagogical assistances

deﬁned by (Lourdeaux et al., 2002).

Properties, expressions, conditions and pedagogical

actions make it possible for the trainer to add run

time individualization to the scenario based on learn-

ers’ actions. Our proposition adds three new types

of property that allow the description of learners’ ex-

pected activity in virtual environments, and pedagog-

ical actions to perform.

3.3 Environment’s Model

Every activity takes place in an environment. Figure

5 represents corresponding class model.

Figure 5: Environment’s class model.

In IMS-LD, an environment contains learning objects

and services. In the context of VRLE, the notion of

environment is a bit different. In addition to learn-

ing objects and services, an environment contains a

virtual world in which learners are immersed and act.

To take this speciﬁcity into account, proposed model

adds a virtual world to the environment via the

World

class deﬁned in VEHA. A world represents an in-

formed virtual environment created with VEHA and

contains two main information:

• A domain-speciﬁc model for the application

(

Model

, level 1). This model contains all the con-

cepts described by a domain expert (as classes, as-

sociations, etc.), types of entities composing the

environment, their properties, etc. the whole be-

ing gathered into packages.

• A virtual environment (

Environment

, level 0).

This environment describes an actual instantia-

tion of the domain-speciﬁc model and contains

InstanceSpecification

objects (cf. section 2).

Every information about the virtual environment can

be referenced by the scenario and properties about en-

tities of this environment can be deﬁned (cf. 3.2).

Proposed model adds a link between a learning ob-

ject and VEHA elements. This association makes it

possible to express links that can exist between vir-

tual environment entities and learning objects outside

the virtual world. For example, it can describe the as-

sociation between a virtual tool and its instructions of

use in PDF format. Then, this link can be used dur-

ing simulation by an trainer or a pedagogical agent to

provide struggling learners an appropriate resource.

CSEDU 2009 - International Conference on Computer Supported Education

258

4 CONCLUSIONS AND

PROSPECTS

This paper describes a learning scenario model able

to integrate VRLE in the learning process. The main

advantage of this model, based on a meta-modeling

approach is that it is generic, in that virtual envi-

ronments can be integrated, regardless of their na-

ture or domain. This model extends IMS-LD on sev-

eral aspects. First, it makes it possible to take into

account the organisational aspect of virtual environ-

ments. Then, three types of properties have been

added (

EntityPropertyRef

StateMachineRef

and

ActionProperty

) so that conditions about virtual en-

vironments’ state can be written. Finally, the model

integrates virtual worlds in IMS-LD environments,

and thus makes it possible to create links between

learning objects and elements of the virtual environ-

ment. The learning scenario model described in this

paper has been used to create a learning scenario for

ECUR

EVI (Querrec et al., 2003), a MASCARET-

based virtual reality application designed to train ﬁre-

ﬁghters. The main prospects of this work focus on the

evaluation of this model by didactics expert of differ-

ent domains. The goal of this evaluation is to check

that the scenario model is ﬂexible enough to take into

account characteristics speciﬁc to different domains

as well as characteristics speciﬁc to different peda-

gogical strategies.

ACKNOWLEDGEMENTS

This study has been carried out with the ﬁnancial sup-

port from R´egion Bretagne.

REFERENCES

Baudouin, C., Chevaillier, P., Le Pallec, A., and Beney, M.

(2008). Feedback on design and use of a virtual envi-

ronment for practical lab work. In Proceedings of the

Virtual Reality International Conference, VRIC 2008,

pages 117–125.

Brough, J., Schwartz, M., Gupta, S., Anand, D., Kavetsky,

R., and Pettersen, R. (2007). Towards the develop-

ment of a virtual environment-based training system

for mechanical assembly operations. Virtual Reality,

11(4):189–206.

Buche, C., Querrec, R., De Loor, P., and Chevaillier, P.

(2004). Mascaret : A pedagogical multi-agent sys-

tem for virtual environment for training. Journal of

Distance Education Technologies, 2(4):41–61.

Burgos, D., Tattersall, C., and Koper, R. (2007). Re-

purposing existing generic games and simulations

for e-learning. Computers in Human Behavior,

23(6):2656–2667.

Gerbaud, S., Mollet, N., Ganier, F., Arnaldi, B., and Tis-

seau, J. (2008). GVT: a platform to create virtual en-

vironments for procedural training. In Proceedings of

IEEE VR, pages 225–232.

Gu´eraud, V. and Cagnat, J.-M. (2006). Automatic seman-

tic activity monitoring of distance learners guided by

pedagogical scenarios. In Innovative Approaches for

Learning and Knowledge Sharing, pages 476–481.

Springer.

Koedinger, K., Aleven, V., Heffernan, N., McLaren, B., and

Hockenberry, M. (2004). Opening the Door to Non-

programmers: Authoring Intelligent Tutor Behavior

by Demonstration. LECTURE NOTES IN COM-

PUTER SCIENCE, pages 162–174.

Koper, R. (2001). Modeling units of study from a peda-

gogical perspective. Educational Technology Exper-

tise Centre, Open University of the Netherlands, First

Draft, version, 2.

Koper, R., Olivier, B., and Anderson, T. (2003). IMS Learn-

ing Design Information Model. IMS Global Learning

Consortium.

Lourdeaux, D., Burkhardt, J.-M., Bernard, F., and Fuchs,

P. (2002). Relevance of an intelligent agent for

virtual reality training. International Journal of

Continuous Engineering and Life-long Learning,

12(1/2/3/4):131–143.

Marion, N., Septseault, C., Boudinot, A., and Querrec, R.

(2007). Gaspar : Aviation management on an aircraft

carrier using virtual reality. In Cyberworlds 2007 pro-

ceedings, pages 15–22.

Munro, A. (2003). Authoring simulation-centered learning

environments with RIDES and VIVIDS. In Authoring

tools for advanced technology learning environments,

pages 61–91. Kluwer Academic Publishers.

Querrec, R., Buche, C., Maffre, E., and Chevaillier, P.

(2003). S´ecuR´eVi : virtual environment for ﬁre-

ﬁghting training. In Richir, S., Richard, P., and Tar-

avel, B., editors, 5

virtual reality international con-

ference (VRIC’03), pages 169–175, Laval, France.

ISBN: 2-9515730-2-2.

Rodr´ıguez-Artacho, M. and Verdejo Ma´ıllo, F. (2004).

Modeling educational content: The cognitive ap-

proach of the palo language. Educational Technology

& Society, 7(3):124–137.

Van Joolingen, W. and de Jong, T. (2003). SimQuest,

authoring educational simulations. In Authoring

tools for advanced technology learning environments,

pages 1–31. Kluwer Academic Publishers.

INTEGRATING KNOWLEDGE FROM VIRTUAL REALITY ENVIRONMENTS TO LEARNING SCENARIO

MODELS - A Meta-modeling Approach

259