PAGE 10

MERITS TRAINING SYSTEM

Using Virtual Worlds for Simulation-based Training

David Chodos, Eleni Stroulia, Pawel Kuras

Department of Computing Science, University of Alberta, Edmonton, Alberta, Canada

Michael Carbonaro, Sharla King

Department of Educational Psychology, University of Alberta, Edmonton, Alberta, Canada

Keywords: Simulation-based Training, Health Education.

Abstract: Virtual worlds offer a new application development platform, and are particularly appealing for creating

new types of educational training programs. However, in order to enable the adoption of this platform by in-

structors, special-purpose authoring tools are necessary to enable domain experts to create and maintain

their lessons plans. In this paper, we propose a framework for virtual world-based training, which uses the

BPEL workflow language to organize educational content. The framework uses a web services-based ap-

proach to connect the content, workflows, and virtual world, thus avoiding dependence on a particular vir-

tual world. Finally, we present a case study, currently in progress, designed to assess the utility of the

framework.

1 INTRODUCTION

Virtual worlds have emerged over the last several

years as a means for users to experience immersive

environments and interact with each other in new

and exciting ways. While much of the activity in

virtual worlds has been of a social, unstructured

nature, there have been efforts to make effective use

of virtual worlds for business and, of particular

interest here, education. However, much of this

effort has focused on re-creating existing

educational settings (campuses, classrooms) in a

virtual environment, and has not been broadly

applicable across educational contexts or

institutions.

Meanwhile, workflows have been used for

decades to specify and drive business processes, and

have recently seen increasing use in structuring

composite online processes through the use of

SOAP (www.w3.org/TR/soap/), a standard for

enabling online service interoperability, and BPEL

(www.oasis-open.org/committees/tc_home.php?

wg_abbrev=wsBPEL), an executable language for

specifying workflows.

In our work, we combine the immersive,

collaborative potential of virtual worlds with BPEL-

based process specification to enable instructors to

specify educational scenarios, and to allow students

to experience those scenarios in a realistic,

interactive manner.

The immediate use of our framework is in

simulation-based training of health professional

students. It is in these terms that we evaluate our

framework in this paper. In the longer run, we are

interested in developing a suite of methods for

analyzing the modeled processes, identifying

opportunities of improvement, and better managing

them.

The rest of this paper is organized as follows.

Section 2 describes the MERITS framework for

modeling and simulating training scenarios. Section

3 reviews our on-going case studies with MERITS.

Section 4 discusses related work. Finally, Section 5

concludes with a summary of our work to date and

outlines plans for the future.

2 THE MERITS FRAMEWORK

The MixEd Reality Integrated Training System

Chodos D., Stroulia E., Kuras P., Carbonaro M. and King S. (2010).

MERITS TRAINING SYSTEM - Using Virtual Worlds for Simulation-based Training.

In Proceedings of the 2nd International Conference on Computer Supported Education, pages 54-61

 SciTePress

PAGE 10

(MERITS) combines a workflow description

component, SOAP-based web services, BPEL

workflows, and virtual-world components to

support instructors in creating simulation scenarios,

through which students can experience situations

similar to those that they will have to deal with in

their future professional life. The major elements of

the MERITS software architecture are shown in

Figure 1.

The MERITS architecture mimics the three-

tiered structure of traditional web-based

applications, with a virtual world as the user

interface, a BPEL orchestrated set of software

services as the application logic (that is, the

software implementing the automated activities of

the service-delivery process), and a resource

repository maintaining a record of the archival data

of the organization and the transient data of each

service-delivery process. For more about the

workflow description process, see Section 2.2.

Figure 1: MERITS framework.

Instructors can specify relevant educational

entities by updating the resource repository through

web-based forms, accessing REST APIs of the

repository. The BPEL-specified workflows that

specify the behaviors of people and objects in the

scenario may, in principle, be created using

graphical, web-based tools. However, there are

conceptual challenges involved in the specification

of a BPEL workflow that make merely providing a

graphical interface insufficient for removing

implementation barriers for non-technical users. For

more about this issue, see Section 3.2. At run time,

the BPEL workflows are enacted through the

interactions of people and objects in the virtual

world and through the behaviors of underlying

automated software systems. When a student

performs an action through his or her avatar, a

behavior script is executed in the virtual world. The

execution of this script may (a) change the state of

the virtual world and (b) change the state of the

corresponding workflow, shown in the second tier

in the diagram in Figure 1. In our implementation,

the workflow server interprets the action in the

context of the overall process workflow to

determine how the scenario should proceed in

response to the action. The BPEL workflow can

also be connected to external devices, thus allowing

the simulation to extend beyond the boundaries of a

particular educational institution. For example, in a

healthcare education context, the system may be

connected to a web service that provides simulated

patient data.

2.1 Workflow Integration

Workflows play an essential role in modeling any

process. In the MERITS framework, we define four

different types of workflows, each of which serves a

specific purpose in the overall process:

1. The process workflow defines the various

process paths that may be taken when executing

the process.

2. An object workflow defines the behavior of an

interactive object involved in the process, such

as a piece of equipment.

3. A character workflow defines the behavior of

automated characters that play some part in the

process but are not simulated by students, such

as non-responsive patients or bystanders.

4. Finally, normative workflows may be used to

define the set of actions (and their control

dependencies) that learners are supposed to

follow, as they play their parts in the process.

While the process and normative learner

workflows may seem similar, they serve very

different purposes. The process workflow describes

all possible paths – including those that are

incorrect or sub-optimal – that may be taken in

executing the process. A normative learner

workflow, on the other hand, describes the sequence

that should be followed by the learner. In this way

the system is able to follow the state of the process

even when it is not going in the prescribed manner,

MERITS TRAINING SYSTEM - Using Virtual Worlds for Simulation-based Training

PAGE 10

and provide feedback on the discrepancies between

the learners’ actual and desired behaviors.

For example, a student's execution of the

emergency rescue process could be checked for the

required actions – checking blood pressure, moving

the patient into the ambulance – and might be

invalidated if any erroneous actions – such as

administering an improper medication – occur.

However, there may be sequences of actions that are

correct, according to the normative user process, but

quite unconventional; for instance, the student

might check the victim’s pulse, go to the ambulance

for a piece of equipment, and then return to the

patient and assess the victim’s level of

consciousness.

2.2 BPEL Implementation

The workflows used in the MERITS system are

specified in BPEL (WS-BPEL, 2009) and consists

of three types of constructs: web service

connections, program control constructs, and

exception and error handling constructs. Using

BPEL, while simpler than a programming language

such as C++ or Java, is, most likely, too technically

challenging for most content experts. While

graphical tools for designing BPEL workflows – of

which there are several – make some of the details

easier to manage, these tools rely on the user

understanding the underlying programming

concepts. Thus, the workflow specification process

currently works as follows.

First, one or more context experts describe the

workflow using storyboard diagrams and

descriptive sentences, and record this description on

a wiki. Second, these descriptions are then analyzed

by technical experts who elicit relevant entities and

actions. Third, the results of this analysis and

modeling decisions are recorded on the wiki, so that

the elicited information can be validated by the

content experts. Finally, based on the elicited,

validated information, the technical experts create

well-defined abstract workflows, which may also be

shared and validated using the wiki. Thus, through

this process, the responsibility for specifying the

workflow in BPEL is delegated to technical users,

although the content experts still retain control over

the meaning of the workflow.

3 EMT TRAINING SCENARIO

To assess the utility of the MERITS system, we

have undertaken several case studies. These case

studies are drawn from a variety of contexts, and are

qualitatively quite different from each other. This

contextual diversity is intended to assess the

flexibility and robustness of the system.

The first version of MERITS was evaluated in

the context of modeling a simple interview process

(Chodos, 2009). This version did not use a formal

specification for the process workflows; we have

since extended the framework with the BPEL

specifications and the corresponding introduction of

a BPEL execution engine to orchestrate the

workflow instances at run time.

We are currently working with colleagues from

the health sciences who want to use this framework

to model complex processes in their field in order to

develop simulation-based training scenarios for

their students. The most mature process model to

date is that of the handoff scenario between

emergency medical technicians (EMT) and

emergency room (ER) personnel when a victim is

being transferred by ambulance from the scene of

the accident to the hospital ER.

The EMT and ER personnel need to acquire and

apply a diversity of basic and complex knowledge

and skills to provide the best patient care. Medical

and procedural knowledge is utilized to quickly

transfer the patient safely to the ER. To coordinate

activities with co-workers, hospital staff and

victim’s families, effective communication within

and across disciplines is critical. The environment is

unpredictable and highly stressful, thus emphasizing

the need for the integration of medical knowledge,

procedural and communication skills prior to

entering clinical practice.

Basic skills and medical knowledge are typically

conveyed in a classroom setting. Procedural skills

are often gained through conducting training

scenarios with students, who interact with

professional actors playing the roles of accident

victims, emergency workers, and hospital staff.

While these training scenarios offer lifelike

experiences for the students, there are a number of

limitations: a) restrictions on the number of students

who can participate at any one time; b) it is

expensive to set up the scenarios because of the

need for equipment, space, and people; c) distance

education students are entirely excluded from this

type of training. Furthermore, developing

interprofessional communication skills across

disciplines is often ignored and students are

expected to gain these skills on the job.

CSEDU 2010 - 2nd International Conference on Computer Supported Education

PAGE 10

Figure 2: Screenshot of EMT training scene.

Figure 3: Wiki page - description and entities.

Figure 4: High-level workflow.

Figure 5: Check pulse workflow.

Figure 6: 3D model of ambulance.

Thus, given the shortcomings identified above, a

MERITS-based training tool presents advantages

over existing systems in several key areas. One is

that a workflow-based system, with independent

objects and characters, enables flexible scenarios.

Thus, the student doesn't just learn a rigid sequence

of steps; rather, the student can interact with active

objects in any order, determine his or her path

through the scenario (subject to constraints imposed

by the context and the workflow), and thus engage

in self-directed learning.

Another advantage of the proposed tool is that it

offers a blend of character types – students,

instructors and automated characters – which

creates a variety of communication possibilities, and

means that the system can offer several types of

student-instructor involvement. First, students can

interact with other students, either in the EMT field

or in other, complementary disciplines. This allows

MERITS TRAINING SYSTEM - Using Virtual Worlds for Simulation-based Training

PAGE 10

students from many disciplines to not only get

experience with the processes relevant to their own

area, but also to get experience communicating with

professionals from a variety of other areas. Second,

students can interact with a mix of other students,

instructors, and automated characters. Thus, a role

in a scenario with minimal interactivity (such as an

unconscious patient) can be simulated by a

workflow-based automated character. More

complex roles, meanwhile, can be played either by

students or instructors (serving as “puppet masters”

moving the scenario along). Finally, in cases where

the behavior all of the external characters may be

modeled, students can interact entirely with

automated characters. This situation offers

maximum accessibility for the student, who can

train using the scenario whenever he or she chooses.

A prototype system for EMT training is

currently under development, in collaboration with

the Department of Computing Science, Faculty of

Education, and the Interdisciplinary Health

Education Partnership conducted by the Health

Sciences Education and Research Commons. In this

training program, EMT students encounter a victim

at an accident scene, and must determine which

actions to take in order to transport the victim to a

hospital. A screenshot of this scene, implemented in

a Second Life-based prototype system, is shown in

Figure 2.

In creating a virtual world-based training

scenario, we followed the process described in

Section 2.2. First, experts in the field were asked to

describe the scenes that make up the scenario. For

each scene, the artifacts and actions were identified

and transformed into workflow diagrams. A

screenshot showing the wiki page for one such

scene is shown in Figure 3.

The MERITS system stores and accesses content

as needed, and coordinates user input with the

various interrelated workflows. The actions taken

by the student include using medical diagnostic

equipment and interacting with the accident victim.

These actions are interpreted by workflows, and the

results are conveyed through a variety of artifacts.

Most of the artifacts in this scenario are pieces of

medical equipment – a spine board, for example, or

the two-way radio in an ambulance – which are

used in treating the victim, and provide immediate

feedback. Other, larger scale, artifacts include the

victim’s vehicle and the EMT ambulance. It should

be noted that, in this case, the victim is also

considered an artifact, since it helps convey the

results of actions through its condition and location.

The implementation of each of these components in

the EMT training prototype will be briefly discussed

in detail in the following paragraphs.

At the core of the MERITS system, we have

workflows which specify a) the victim treatment

process, and b) the actions associated with each

artifact (gloves, spine board, radio, victim) in the

scenario. The high-level workflow for the victim

treatment process is shown in Figure 5, and

described in the folllowing paragraph.

Many of the components of the victim treatment

workflow – Gloves, Pulse, SpineBoard and

CallHospital – are concerned with recognizing and

recording simple actions. The MoveVictim

component ensures that the victim is on the spine

board before he can be moved to the ambulance.

The DriveToHospital component analyzes the

variables, which store previous actions, determines

whether a prerequisite action (e.g., moving the

patient into the ambulance) has been missed, and

returns an appropriate message.

The workflows that define artifact-based actions

in this process fall into two types: simple toggle

(on/off) actions and information retrieval. The

toggle actions, such as putting on gloves, register

the action with the overall process workflow and

instruct the virtual world to make the appropriate

change in the object’s appearance. The second type

of action adds information retrieval to the above

tasks. This information may be retrieved either from

a database or from an external web service. An

example of this type of workflow (for the Check

Pulse action) is shown in Figure 5.

It should be noted that, while the artifact-related

components in the victim treatment workflow may

seem quite similar to the artifact workflows just

described, they are conceptually and functionally

quite different. An artifact workflow describes the

behavior of that artifact, independent of any

processes. An artifact-related component within a

process workflow, meanwhile, describes how that

artifact's behavior – that is, the actions taken on the

artifact – relate to the process as a whole. In this

process, for example, the CheckPulse workflow

describes the behavior of a pulse oximeter (or some

other pulse-checking device). The Pulse component

of the process workflow, on the other hand,

describes the impact that checking the patient's

pulse has on the process being modeled.

Representing an action in a virtual world can

pose a variety of challenges, depending on the

affordances and capabilities provided by the virtual

world. For example, in Second Life, objects can

only be touched, worn, sat on, driven, or taken by a

player. Thus, an action such as carrying an object is

CSEDU 2010 - 2nd International Conference on Computer Supported Education

PAGE 10

difficult to represent, since to have an object follow

an avatar, it must incrementally follow the avatar,

be worn by the avatar, or be driven by the avatar,

each of which have their limitations. It turns out

that, while driving an object seems to be the least

intuitive, it produces the most accurate results.

Finally, the artifacts relevant to the process must

be modeled in the virtual world. For this process,

the artifacts that were modeled included a spine

board, the unconscious victim, a two-way radio, and

the ambulance.

There are two distinct issues here: first, there is

the representation of the artifact in the virtual world.

Depending on the virtual world that is chosen, one

may be able to create this representation using

external 3D modeling software, import 3D models

created by other users, or use in-world 3D modeling

tools. For this process, we used 3D models created

by other users for Second Life. Figure 6 shows a

close-up of an ambulance, one such 3D model.

Second, this artifact must be able to exhibit the

behavior implied by the associated actions. This,

typically, involves the addition of native code to the

virtual world representation of the artifact. See

Section 3.1 for more about this issue. This code

calls web services responsible for interpreting the

result of the action, both in terms of any immediate

changes to the artifact appearance, as well as the

impact of the action on the process workflow.

Finally, the artifact must be able to change its

appearance in an appropriate manner.

At this point, we have just had some initial

feedback with respect to the quality of our scenario

representation by simulation experts from the EMT

field but have not yet conducted any formal

experiment. The first trial for empirically evaluating

the effectiveness of our virtual-world simulation for

training is scheduled for January 2010.

4 RELATED WORK

Using virtual worlds to create scenario-based

training programs touches on several areas of

related work. First, there is a growing body of work

describing and analyzing the use of virtual worlds --

most prominently, Second Life - for post-secondary

education and training. Second, there are several

theories from educational psychology that support

the effectiveness of scenario-based training

programs. Each of these areas will be described in

the following sections.

4.1 Virtual Worlds

The issue of using virtual worlds for education and

training has received an increasing amount of

attention from the academic community in recent

years, as virtual worlds have become better

established in both mainstream culture and in

educational institutions. The following paragraphs

present a sample of this work, which indicates both

the steadily increasing interest in the topic, and the

variety of approaches that are being taken.

Hong Cai, of IBM, has taken a broad view of the

issue, examining the potential of virtual worlds for

any kind of training program (Cai, 2008). He

compared several virtual environments - Second

Life, Active Worlds, OpenSim, and the Torque

game engine - in terms of their fitness for

educational activities, and analyzed various

common learning activities with respect to their

implementation in a virtual environment. He also

presented a development lifecycle for creating

virtual learning environments, and analyzed several

virtual learning projects at IBM according to these

analytical tools.

Edward Carpenter has developed a 3D crisis-

communication training tool to provide

communication students with opportunities to

practice what are, in a standard classroom setting,

largely theoretical approaches to dealing with crises

(Carpenter, 2006). Through the immersive tool,

students get hands-on training, and can experience

events, rather than absorbing and interpreting them

through written information. The tool uses a

narrative, storyboard-based technique to deliver the

educational content, where each student is offered a

set of choices at key points in the story. Afterwards,

the students are debriefed and the instructor

analyzes and evaluates their choices. Because the

system uses storyboards to structure the educational

content, a student's interaction with the system is

largely pre-determined, and quite rigid. As well, the

system does not support collaborative learning,

since it is intended for one student at a time.

Victor Vergara and colleagues at the University

of New Mexico have developed a virtual

environment-based tool to teach medical students

about hematomas (Vergara, 2008). They have

developed a 3D, multi-user virtual environment

(MUVE) within which students can interact with a

virtual character, nicknamed “Mr. Toma,” and other

associated objects. Several rigorous studies of the

system's effectiveness have demonstrated that it is

equally effective as conventional, paper-and-pencil

education methods. Furthermore, it offers additional

MERITS TRAINING SYSTEM - Using Virtual Worlds for Simulation-based Training

PAGE 10

advantages, including the chance to collaborate with

geographically dispersed students, and an increased

sense of immersion when using the MUVE system.

A considerable amount of effort was put into

ensuring that the content was presented accurately

and effectively, including consulting with an

interdisciplinary team of subject matter experts.

Finally, Forterra Systems has developed the On-

LIine Virtual Environment (OLIVE) platform,

which allows clients from government, healthcare

and other contexts to create virtual world-based

systems (Armentrout, 2008). One such system was

developed to train first responders to car accidents

on the Interstate 95 Corridor. A prototype system

was developed at the University of Maryland, and

tested with a small number of potential students.

While the preceding projects are quite varied in

terms of the technology used and the context areas

to which the projects were applied, there are several

common characteristics that should be pointed out.

First, each project found that the students’

educational needs were met by the virtual world-

based projects. This offers evidence that this type of

system is effective in a wide range of contexts, and

with a broad range of students. Second, with the

exception of the project undertaken by Forterra

Systems, most current projects are computing-

science initiatives, for use in a single context area.

This indicates that there is a need for a broadly

applicable framework for virtual world-based

training programs. This framework should enable

the creation and maintenance of learning modules

by non-technical content experts.

4.2 Educational Psychology

From an educational psychology standpoint,

simulation-based training is supported by the

situated cognition theory, proposed by Brown et al

(Brown, 1989). According to this theory, knowledge

is not a set of abstract concepts to be absorbed by

the student; instead, it is dependent on the context

and culture in which it is used. Adhering to

situation-cognition principles, Collins et al

developed the cognitive-apprenticeship model of

educational practice, which incorporates the situated

nature of the knowledge being conveyed to students

(Collins, 1991). This model was later found to be

effective within a technologically rich learning

environment (Järvelä, 1995). Another related

concept is constructivism (Duffy, 1992). This

theory sees learning as an active process of

constructing, rather than acquiring knowledge.

Thus, instruction within a constructivist context

focuses on supporting that construction, rather than

conveying knowledge. These theories and studies

support the value of simulation-based training.

Students learn to integrate the knowledge and skills

in order to apply them in a context similar to the

real environment.

5 FUTURE WORK

There are several areas of future work that will be

pursued over the coming months, to address

existing issues, improve the specification process,

explore diverse context areas, and provide empirical

validation of the framework.

Implementing a process in Second Life (SL)

poses platform-specific challenges. The limited

interaction affordances provided by SL make it

quite challenging, if not impossible, to implement

some types of simple actions (such as picking up or

pushing an object) in a natural, realistic way. Thus,

we would like to either a) develop a consistent SL

API which could be used to model processes in a

natural way or b) migrate to another virtual world

platform that provides better native support for the

processes being modeled.

Another issue is the workflow definition

process. Currently, processes are described by

content experts and then converted into BPEL

workflows by technical experts, supported by a wiki

system. In the future, we would like to allow the

content experts to create workflows using graphical

tools. These tools would give content experts direct

control over the process modeling workflows.

One of the key benefits of the MERITS system

is its applicability to a wide variety of context areas.

In addition to EMT training, we are also

investigating training students in Occupational

Therapy and the use of industrial equipment. By

selecting a range of contexts, we can assess the

utility of the MERITS system along dimensions

such as social interaction and communication,

artifact complexity and precision.

Finally, we are planning on conducting a series

of empirical studies of the effectiveness of the

MERITS system for modeling service delivery

processes, beginning in November 2009. The initial

study will take the form of a pilot of the EMT

training scenario (described in Section 4) for the

Interdisciplinary Health Education Partnership. The

pilot study will followed by a larger-scale study

beginning in January 2010.

CSEDU 2010 - 2nd International Conference on Computer Supported Education

PAGE 10

6 CONCLUSIONS

In this paper, we discussed a software framework

for specifying interactive educational scenarios,

with BPEL and virtual-world elements for

simulation-based training of students across health

disciplines. We have illustrated our framework with

the EMT training scenario, which highlights the

complexity of such scenarios, in general. The EMT

training scenario is realistic – it has been developed

by EMT personnel and ER nurses – and its

simulation will be used for training EMT and ER

students in the near future.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the

generous support of iCORE, NSERC, and IBM. We

are also grateful to Andrew Reid and Ken Brisbin

for their contributions in modeling the EMT/ER

scenario and refining the MERITS framework over

several lively discussions.

REFERENCES

Armentrout, M., 2008. Transportation Incident Manage-

ment: Using 3D Virtual Worlds to Train First Res-

ponders. Published by Forterra Systems.

Bitner, M., Ostrom, A.L., Morgan, F.N., 2007. Service

Blueprinting: A Practical Technique for Service Inno-

vation. Working paper, published by the Center for

Services Leadership, Arizona State University.

Brown, J.S., Collins, A., Duguid, P., 1989. Situated Cog-

nition and the Culture of Learning. In Educational

Researcher, pages 32-42.

Cai, H., Sun, B., Farh, P., Ye, M., 2008. Virtual Learning

Services Over 3D Internet: Patterns and Case Studies.

In IEEE International Conference on Services Com-

puting, volume 2, pages 213-219.

Carpenter E., Kim, I., Arns, L., Dutta-Berman, M.J.,

Madhavan, K., 2006. Developing a 3D Simulated

Bio-terror Crises Communication Training Module.

In VRST ’06: Proceedings of the ACM symposium on

Virtual reality software and technology, pages 342-

345, New York, NY, USA.

Chodos, D., Stroulia, E., Naeimi, P., 2009. An integrated

framework for simulation-based training on video and

in a virtual world. In Journal of Virtual Worlds Re-

search, volume 2, issue 1, April 2009.

Collins, A., 1991. Cognitive Apprenticeship and Instruc-

tional Technology. In Educational Values and Cogni-

tive Instruction: Implications for Reform, L. Idol and

B.F. Jones, editors, pages 121-138, Lawrence Erl-

baum.

Duffy, T.M., Jonassen, D.H., 1992. Constructivism: New

Implications for Instructional Technology. In Con-

structivism and the technology of instruction: a con-

versation, T.M. Duffy and D.H. Jonassen, editors,

Lawrence Erlbaum.

Järvelä, S., 1995. The Cognitive Apprenticeship Model in

a Technologically Rich Environment: Interpreting the

Learning Interaction. In Learning and Instruction, vo-

lume 5, pages 231-259.

Vergara, V., Caudell, T., Goldsmith, T., Panaiotis, Alver-

son, D., 2008. Knowledge-driven Design of Virtual

Patient Simulations. In Innovate: Journal of Online

Education, volume 5.

WS-BPEL, 2009. Web Services Business Process Execu-

tion Language Technical Committee. Retrieved on

June 4, 2009 from

http://www.oasis-

open.org/committees/tc_home.php?wg_abbrev=wsbp

MERITS TRAINING SYSTEM - Using Virtual Worlds for Simulation-based Training