Adaptive and Blended Learning for Electrical Operators Training
With Virtual Reality Systems
Yasmín Hernández and Miguel Pérez Ramírez
Gerencia de Tecnologías de la Información, Instituto de Investigaciones Eléctricas, Reforma 113, Cuernavaca, Mexico
Keywords: Empathic Agents, Bayesian Networks, Blended Learning, Student Model, Virtual Reality.
Abstract: Due to the danger involved in the electrical field, qualified electricians are required. Traditionally, training
has been based on classroom courses and camp training, but it is costly and students need to spend a long time
to develop their competences. We propose to complement traditional training with an intelligent training
system composing a blended training model. The blended model enables adaptive training through a student
model which represents the affective and knowledge states of the trainees. The affect is recognized taking
into account a theoretical model of emotions. The knowledge of the student is updated as he interacts with the
system. The instruction is presented in a virtual reality environment by an empathic agent. The virtual reality
system enables practicing in a controlled and safe environment. In this paper, the general proposal for the
blended training model is presented.
1 INTRODUCTION
The electrical domain requires efficient and well
trained electricians because a manoeuvre badly
performed can result on accidents that can injure
people or damage costly equipment. However,
training personnel confronts problematic situations
such as the limited availability of electrical
installations that trainees need for practicing in real
environments, therefore the trainees have to help as
assistant of electricians for a long time, and due to
danger included, they first only observe the
manoeuvre. This limited opportunity to practice in
real environments makes training to take a long time
besides being costly.
In order to advance in the solution of this problem,
we are working in composing a blended learning
model. A definition states that blended learning is
learning that is facilitated by the effective
combination of different modes of delivery, models
of teaching and styles of learning, and founded on
transparent communication amongst all parties
involved with a course (Heinze and Procter, 2004).
In our proposal, trainees still attend classroom
courses but they complement learning and practice
aided by an intelligent training system. The intelligent
training system integrated virtual reality systems
which allow having a virtual representation of the
electrical environment. The virtual reality is the
electronic representation (partial or complete) of a
real or fictitious environment. Such representation
can include 3D graphics and/or images, has the
property of being interactive and might or might not
be immersive (Pérez and Ontiveros, 2009).
Another component is a trainee model
representing the knowledge and affect states of the
trainee. The trainee model enables adaptive training
as in an intelligent tutoring system (Woolf, 2009).
The trainee model is represented by Bayesian
networks. On the knowledge side the model includes
the topics the trainee has already learnt and the topics
the trainee does not know yet. On the affect side the
model includes what the trainee feels according to the
OCC model (Ortony, Clore and Collins, 1988) and to
the basic emotions proposed by Ekman and Friesen
(1978).
The blended learning model also includes an
empathic agent to be the face of the intelligent
training system.
In this way, we have a blended learning model
which enables adaptive and intelligent training where
the individual state of trainees is considered. The
training scenarios are presented as virtual
environments enabling valuable practice before going
to real electrical installations, and the learning is
facilitate by an animated agent.
Regarding the integration of the technological
pair: virtual reality and intelligent tutoring systems, it
Hernández, Y. and Ramírez, M.
Adaptive and Blended Learning for Electrical Operators Training - With Virtual Reality Systems.
In Proceedings of the 8th International Conference on Computer Supported Education (CSEDU 2016) - Volume 1, pages 519-524
ISBN: 978-989-758-179-3
Copyright
c
2016 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
519
is not a new proposal. In the late 90s, Steve the
animated agent who played the role of an instructor
was representative (Johnson et al., 2000). At that time
(Lane and Johnson, 2008) pointed out that there were
still a number of unanswered key questions in the
literature of this technological pair, some of these are:
How distracting is explicit feedback? and What are
the risks of stealth guidance and experience
manipulation on learners with respect to confidence,
self-efficacy, and help-seeking skills. Virtual reality
carried out in its evolution and Burdea and Coiffet
(2003) consider that training is one of the main fields
for VR application.
Nowadays virtual reality technology enables
developers to offer more realistic environments.
Virtual environments are helpful for users to visualize
how physical activities should be realized.
Interactivity has improved but is still under research.
Augmented reality, another strand of virtual reality is
becoming more useful as training support in industry
(Carson, 2015). Wagner (2015) states virtual reality
will make online tutoring as common place as one-
on-one tutor over the next years; he even envisages
that online degrees are more affordable that the
traditional ones.
An analysis of virtual reality evidences its
potential for tutoring activities, since it allows the
integration of features such as interactivity,
visualization and audio, among others, to stimulate
different human learning channels (Pérez and
Ontiveros, 2009), which are propitious and helpful to
enhance the learning process.
Both technologies are still under research but
already mature enough to answer some of the
questions posed in the past.
This paper presents our general proposal for the
blended training model and describes their main
components.
2 BLENDED TRAINING MODEL
Training is a strategic activity in corporations as it is
recognized that the efficiency of organizations
depends directly on human capital, which in turn may
depend on adequate training. High productivity in
part is the result of efficient training, which becomes
even more valuable when there is risk of accidents
that harm people. Such is the case of electrical field,
which involve risk of electric shock, arc flash and
other hazards for people; also there may be potential
damage to equipment in electrical installations.
The training programs on the electrical field are
very detailed and strict. A trainee has to accredit the
classroom courses but he also must have camp
practice with the close supervision of an instructor. In
this traditional training method, the trainees spend a
lot of time and also the training becomes costly.
With these elements, we are developing a blended
training model to support traditional training as it can
be seen in Figure 1. In this training model, the trainees
learn through three elements: i) an instructor in
classroom courses, ii) an intelligent training system
and iii) camp practice.
Figure 1: Blended training model.
The aim of the blended training model is to have
efficient, fast and safe training, and also to reduce
training costs. The base of the blended training model
is the traditional training which in turn is based on a
plan including theoretical lessons and practice.
In a face-to-face interaction, the instructor teaches
trainees in classroom (first component). These
classroom classes are supported by the intelligent
training system (second component) where trainees
can reinforce the theoretical topics by executing
practices in a virtual environment. Separately,
trainees can learn and practice with the intelligent
training system as much as they want; this is in a self-
learning modality.
When trainees have attended the appropriate
courses they have to serve as auxiliary electricians to
have camp practice (third component) in a real
electrical installation.
The training course is planned by the instructor;
he decides which topics will be included in the course
and designs the course in the intelligent training
system. In classroom, the instructor explains
theoretical concepts and shares his experience in the
performance of electrical manoeuvres.
CSEDU 2016 - 8th International Conference on Computer Supported Education
520
This model, which supports self-learning, is
looking for adaptive training, where particular
trainee’ needs are considered. We have established a
road map with several phases to achieving such
training model (Hernández and Pérez, 2014).
3 INTELLIGENT TRAINING
SYSTEM
The intelligent training system is the component
which allows adapting the training to particular needs
of each trainee. It includes elements from intelligent
tutoring systems such adaption to particular needs by
means of a student model (Sottilare, 2013). Figure 2
shows a diagram of the intelligent training system.
Figure 2: Intelligent training system.
Besides attending the course, trainees practice the
electrical topics included in the course by using the
intelligent training system with an animated agent.
A key element of this intelligent system is the
trainee model that is built based on the interaction
trainee-system. Also the trainee model is updated by
the instructor considering the progress of the trainee
in class and his performance in camp.
The trainee model represents the knowledge and
affective states of trainees and their profile. The
information in this model is useful for instructors to
adapt the instruction in classroom, to plan the camp
practice, to recommend attending other training
courses or finally to grant a certification. The model
can be useful to design new courses and new testing
materials and even to redesign training material.
The trainee interacts with the intelligent training
system via a virtual reality system. This system
presents to trainees a virtual representation of the
electrical environment enabling the practice before
going to the real electrical installation. Also the
system allows a safe training since the trainees can
practice as much as they want without risk to injure
them or damage costly equipment.
Depending on the specific electrical topic, the
instruction and the practice in the intelligent training
system can be adapted to the progress and knowledge
of the trainee. In specific cases, it is difficult to adapt
instruction because the electrical manoeuvre has to be
performed sequentially. However, the intelligent
training system can suggest to reviewing specific steps
or to studying certain topics.
Another component is an empathic animated
agent which uses the trainee affect to present the
instruction properly. We are using the characteristics
of the operators for developing the agent, such as
wearing the uniform and safety helmet, among other
features. We believe that by representing the tutor as
an electrician, operators will accept better the training
environment. Empathy is the ability to perceive,
understand and experience others’ emotions, in other
words, to step into the shoes of another. This
construct has been incorporated in animated agents
with the aim to achieve credibility, social interaction
and user engagement (Hone, 2006).
The knowledge about the electrical field
composed by teaching and testing material are
designed and developed by a team of experts.
3.1 Pedagogical Trainee Model
The pedagogical trainee model represents the
trainee’s knowledge about electrical topics included
in the course. The model is updated when the trainee
practices the electrical manoeuvres and when he
solves theoretical exams. The model consists of a
Bayesian network (Sucar, 2015). The Bayesian
network is built when the instructor designs a course.
Figure 3 shows an example of a Bayesian network for
a course with five electrical topics. In turn each topic
is composed by a sequence of subtopics.
The Bayesian network is composed by a node for
each electrical topic included in the course. In turn,
each node of the Bayesian network representing a
topic is a Bayesian network composed by topics and
subtopics.
Initially, the nodes representing topics have two
possible values: learnt and not learnt and their
probabilities are conditionally dependent on the
probabilities of learning the subtopics nodes.
Course nodes also have two values: acquired and
not acquired and their probabilities are conditionally
dependent on the probabilities of knowing the topic
and subtopics nodes.
We are working on including a node for a
theoretical exam also represented by a Bayesian
network composed by a number of items. The causal
relationships between items and conditional
probabilities for each node will be established when
the exam is designed by the instructor. For the time
Adaptive and Blended Learning for Electrical Operators Training - With Virtual Reality Systems
521
being, we have not defined the complete structure and
values of this Bayesian network. However we want to
model trainee’ guesses and slips on the basis of the
relationships between the items and the evidence of
the answers to questions. Figure 4 shows an exam
with 8 items as a preliminary example.
Figure 3: Bayesian network for a course.
Figure 4: Initial Bayesian network for an exam.
3.2 Affective Trainee Model
The affective trainee model uses the OCC model
(Ortony, Clore and Collins, 1988) to provide a causal
assessment of emotions based on contextual
information. The OCC model defines emotional state
as the outcome of the cognitive appraisal of the
current situation with respect to one’s goals. The
trainee model consists of a dynamic Bayesian
network that probabilistically relates personality,
goals and interaction events with affective states.
Figure 5 shows a high level representation of the
model, where each node in the network is actually a
set of nodes in the detailed model. The model is based
on the proposal by Conati and Mclaren (2009) and in
our previous work (Hernández, Sucar and Arroyo,
2012).The dynamic Bayesian network models the
dynamic nature of emotions. To infer the affective
state, it considers the trainee’s knowledge,
personality, and the tutorial situation at that time, as
well as the previous trainee affective state. The
tutorial situation is defined based on the results of the
trainee actions.
The trainee’s appraisal of the current situation
given his goal is represented by the relation between
Figure 5: Dynamic Bayesian network for the affective
trainee model.
the goals and the tutorial situation nodes through the
satisfied goals node. The influence of the appraisal
process on the trainee’s affect is represented by the
link between the satisfied goals node and the affective
state node. From the complete set of emotions
proposed by the OCC model, the affective model only
includes six emotions: joy, distress, pride, shame,
admiration and reproach. We use only these emotions
because they are related to the events we want to
evaluate: the emotions joy and distress are reactions
by the individual to an event in the training session.
The emotions pride and shame emerge as a
consequence of the trainee’s action. The emotions
admiration and reproach emerge as a consequence of
the tutor’s action.
According to the OCC model, one’s goals are
fundamental to determine one’s affective state, but
asking the trainees to express these goals during
training would be too intrusive. Consequently, the
goals in our network are inferred from personality and
trainee’s knowledge.
3.3 Animated Agent
Training activities are presented to trainees through
an animated pedagogical agent. These agents
represent a major trend to have a more natural human-
computer interaction (Breese and Ball, 2008,
Johnson, Rickel and Lester, 2000). Animated
pedagogical agents interact face to face with the
students through facial expressions, gaze, emotions
and deictic gestures; and cohabit with the students
learning environments. Animated pedagogical agents
have a significant impact on training systems as they
give the impression that someone is on the other side
(Sagae et al, 2012); thus the trainee perceives a very
different behavior from a traditional system and more
alike to human behavior. Among the behaviors of an
animated pedagogical agent are those typical of
intelligent tutoring systems, but there are some
Course
To p ic 1
Topi c 2
Topi c 3
Topi c 4
Sub
topic 1
Sub
topic 2
Sub
topic 3
Sub
topic 1
Sub
topic 2
Sub
topic 3
Sub
topic 4
Sub
topic 1
Sub
topic 2
Sub
topic 3
Sub
topic 4
Sub
topic 7
Sub
topic 5
Sub
topic 6
Sub
topic 1
Sub
topic 2
Sub
topic 3
Exam 1
Item
1
Item
2
Item
3
Item
4
Item
7
Item
5
Item
6
Item
8
Satisfied Goals
Knowledge
State
Goals
Personality traits
Tutorial
Situation
Affective state
Satisfied Goals
Goals
Personality traits
Affective state
t
n
t
n+1
Knowledge
State
Tutorial
Situation
CSEDU 2016 - 8th International Conference on Computer Supported Education
522
particular of these characters, such as demonstrations
of complex tasks, observe and assist the trainee to
perform their tasks, in addition to guiding trainees in
virtual spaces (Wang et al, 2008).
We are using the characteristics of the operators
for developing the agent, such as wearing the uniform
and safety helmet, among other features. We believe
that by representing the tutor as an electrician,
instructors and trainees will accept the training
environment. We have conducted a study to evaluate
the design of the empathic agent and gather
knowledge to refine it. We obtained encouraging
results, as the electricians welcomed the agent
(Hernández et al, 2016). The results of the study shed
some light to refine the facial expressions of the agent
and its overall design. Figure 6 shows two instances
of the animated agent.
Figure 6: Animated pedagogical agent.
In this initial phase, the animated agent will
deploy the emotions recognized in the trainee base in
the OCC model as described above. The facial
expressions, consequence of the emotions, adopt the
theory proposed by Ekman and Friesen (1978). We
are trying to accomplish an empathic behaviour in the
animated agent to achieve believability and user
engagement, and in turn to improve learning.
4 VIRTUAL REALITY SYSTEMS
We have developed different non immersive virtual
reality systems for training. ALEn
3D
is one of them
and nowadays is a complementary training tool for
medium tension live-line maintenance. In fact there
are different versions of this system, all devoted to
maintenance of energized lines. Thus, we have
ALEn
3D
MT for medium tension power lines, See
Figure 7, ALEn
3D
AT for high tension power lines
and ALEn
3D
LS for underground power lines.
Besides these systems we have also developed a
virtual reality system for protections maintenance, see
Figure 8, and substation tests. All these systems share
in some degree the same architecture and
functionality within different instructional domains.
They keep track of trainees’ progress; however we are
still working on them to integrate some intelligence,
so that these systems are able to keep fully records of
the model of trainees and even integrate the animated
agents and capability to recognise emotions among
other functionalities proposed in the blended training
model.
Figure 7: Virtual reality training system for medium tension
live line maintenance ALEn
3D
MT.
Figure 8: Virtual reality training system for maintenance
tests to protections.
5 CONCLUSIONS
In this paper we propose a blended model for training
electricians. This model includes an intelligent
training providing adaptive and intelligent training
since it recognizes the affect and knowledge state of
trainees. The instruction is presented in a proper way
by an empathic agent who is a learning companion for
the trainee. The intelligent training system integrates
a virtual reality system.
We have presented the characteristics of the
intelligent training system as a component of the
blended training model, i.e. learning and practice is
part of a course; however trainees also can use the
intelligent training system as distance self-training
tool, practicing any topic at any time.
Adaptive and Blended Learning for Electrical Operators Training - With Virtual Reality Systems
523
Even though we have added different
technologies to our model and training systems in
order to make them efficient, still presence of human
instructors plays a decisive role. These technologies
are helpful tools to support and improve training but
cannot substitute instructor. As in other fields,
training within the electrical field often involves high
risk activities where mistakes are usually fatal.
Thus, the intelligent training system is a helpful
complementary training tool which can be used to
enhance the traditional training but it cannot be used
instead of it.
As future work we are planning to show the
trainee model to trainee as a self-evaluation tool. Self-
assessment is one of the meta-cognitive skills
necessary for effective learning. Trainees and
students, in general, need to be able to critically assess
their knowledge in order to decide what they need to
study (Mitrovic and Brent, 2002). For the time being
the open trainee model is used only by instructors.
REFERENCES
Breese, J. S., Ball, G. (2008) Modeling Emotional State and
Personality for Conversational Agents, Technical
report, MSR-TR-98-41.
Burdea, G. C. and Coiffet. P. (2003) Virtual Reality
Technology, 2
nd
edition, New Brunswick, NJ: Wiley-
IEEE Press.
Carson, E. (2015) NASA shows the world its 20-year virtual
reality experiment to train astronauts: The inside story.
[Online], Available: http://www.
techrepublic.com/article/nasa-shows-the-world-its-20-
year-vr-experiment-to-train-astronauts/ [8 Feb 2015]
Conati, C. and Mclaren, H. (2009) Empirically Building
and Evaluating a Probabilistic Model of User Affect,
User Modeling and User-Adapted Interaction, vol. 19,
no. 3, pp. 267-303.
Ekman, P. and Friesen, W. (1978) Facial Action Coding
System: A technique for the measurement of facial
movement. Consulting Psychologists Press, Palo Alto.
Heinze, A. and Procter, C. (2004) Reflections on the Use of
Blended Learning, Conference Proceedings, Education
in a Changing Environment Conference.
Hernández, Y., Pérez-Ramírez, M., Zatarain-Cabada, R.,
Barrón-Estrada, L. and Alor-Hernández, G. Designing
empathetic animated agents for a b-learning training
environment within the electrical domain, To appear in
Educational Technology & Society, vol. 19, no. 2
(2016).
Hernández, Y. and Pérez-Ramírez M. (2014). A b-learning
model for training within electrical tests domain,
Intelligent Learning Environments, special issue of
Research in Computing Science, vol. 87, pp. 43-52.
Hernández, Y., Sucar, L. E. and Arroyo-Figueroa, G.
(2012) Affective Modeling for an Intelligent
Educational Environment in Peña, A. (ed.) Intelligent
and Adaptive Educational-Learning Systems:
Achievements and Trends, Heidelberg: Springer.
Hone, K. (2006) Empathic agents to reduce user frustration:
The effects of varying agent characteristics, Interacting
with Computers, vol. 18, no. 2, pp. 227-245.
Johnson, W. L., Rickel, J. W. and Lester, J. C. (2000)
Animated Pedagogical Agents: Face-to-Face
Interaction in Interactive Learning Environment,
International Journal of Artificial Intelligence in
Education, vol.11, no. 1. pp. 47-78.
Lane, H.C. and Johnson, W.L. (2008) Intelligent Tutoring
and Pedagogical Experience Manipulation in Virtual
Learning Environments, in Schmorrow, Cohn and
Nicholson (eds.), The PSI Handbook of Virtual
Environments for Training and Education:
Developments for the Military and Beyond. Praeger
Security International: Westport, CN.
Mitrovic, A. and Brent, M. (2002) Evaluating the Effects of
Open Student Models on Learning, Conference
proceedings, Adaptive Hypermedia and Adaptive Web-
Based Systems, pp. 296-305.
Ortony, A., Clore, G.L. and Collins, A. (1988) The
Cognitive Structure of Emotions, Cambridge University
Press.
Pérez-Ramírez, M. and Ontiveros-Hernández N. J. (2009)
Virtual Reality as a Comprehensive Training Tool, in
Gelbukh, A. (Ed) Artificial Intelligence & Applications,
Mexico: SMIA.
Sagae A., Hobbs, J. R., Wertheim, S., Agar, M., Ho, E. and
Johnson, W.L. (2012) Efficient Cultural Models of
Verbal Behavior for Communicative Agents,
Conference proceedings, Intelligent virtual agents,
Santa Cruz, pp. 523-525.
Sottilare, R., Graesser, A., Hu, X. and Holden, H. (eds.)
(2013) Design Recommendations for Intelligent
Tutoring Systems Volume 1 Learner modeling, Florida:
U.S. Army Research Laboratory.
Sucar, E. (2015) Probabilistic Graphical Models:
Principles and Applications, Springer.
Wang, N., Johnson, W. L., Mayer, R. E., Rizzo, P., Shaw,
E. and Collins H. (2008) The politeness effect:
Pedagogical agents and learning outcomes,
International Journal on Human-Computer Studies,
vol. 66, no. 2 pp. 98-112.
Wagner, D. (2015) Virtual Reality Revolutionizes Online
Tutoring, [Online] Available: https://www.linkedin.
com/pulse/virtual-reality-revolutionizes-online-
tutoring-tutorz-com [8Feb 2016]
Woolf, B. P. (2008) Building Intelligent Interactive Tutors,
Morgan Kaufmann.
CSEDU 2016 - 8th International Conference on Computer Supported Education
524