Professional Documents
Culture Documents
Session T1J
concrete, they designed both a logic system where these
concepts were included and an architecture for the execution
of BDI based programs [9].
Beliefs are related to everything that the agent knows
(both related to the environment and to its internal state) and
will be stored in the belief base of the agent. Goals allow
describing what the agent must to achieve, but they do not
include information about the actions the agents must
exactly carry out to do that. Finally, plans are composed by a
set of instructions which allow the agent to carry out several
actions to try achieving the previously set goals. That is, if
the current situation (that the agent knows) is not the desired
one (i.e. the situation specified by the goals) then the agent
will execute the plan or plans to achieve this desired
situation. The relation between goals and plans will be
carried out by a reasoning engine, which will decide what
plan will be executed to try satisfying a specific goal.
The use of the words to try is not casual, because the
achievement of the goal is not guaranteed. This is so because
the BDI agent is designed to reason under certain conditions,
which can change while the plan is being executed (or while
the reasoning engine is taking a decision). Taking this in
account, it is not out of mind to find alternative plans.
This representation of the behavior using mental notions
has several benefits, such as to avoid the low level
abstraction (because only specific goals to achieve and
several plans to do that are presented) and easily understand
the autonomous behavior of the agent (even its prediction).
I. BDI platforms
Inspired by the design of Rao and Georgeff, several
programming platforms have been arisen. Jadex [10] and
JACK [11] can be considered as two of the most important
ones.
The Jadex project is been developed by the Distributed
Systems and Information Systems group of the University of
Hamburg, and it is distributed under GNU LGPL [12]
license. Nowadays it has several tools which can be grouped
in three sets:
Sessiion T1J
control the participattion of the sttudents, and tto evaluate
them.
LCMS
II. Funcctionalities as L
An LCM
MS is used to ddesign, create, and manage thhe contents
of a learrning program
m (usually, conttents belong to courses).
As an LCMS, INE
ES provides fuunctionalities such as:
FIGURE 1
BLOCK DIAG
GRAM OF INES.
Administraation
and
managementt
(registratio
on, directory
y of userss,
qualificatio
ons, etc.).
Distributio
on of contents.
Collaborative work tools.
Control an
nd assessment of
o users.
Design of personalized
p
leearning plans.
of
courrses
consult of
In short, our
o platform offers
o
to the instructors,
i
as an
L
LMS, function
nalities to distribute learniing contents, to
9978-1-4244-6262-9/10/$26.00
0 2010 IEEE
E
Occtober 27 - 30, 2010, Washington, DC
40th ASE
EE/IEEE Fron
ntiers in Educcation Confereence
T1J-3
Sessiion T1J
uld see beforee, the commun
nication with the
As we cou
sstudents is carrry out by a chaatterbot (CHAR
RLIE CHAttteR
L
Learning Interrface Entity), which acts as an interfa
face
bbetween a student and th
he system. More
M
concreteely,
C
CHARLIE wiill communicaate with a messenger
m
ag
gent
((EMMA Events and Messsages Manageer Agent), wh
hich
w
will be the interrmediary between it and ISM
MAEL (Figure 2).
2
II. EMMA Agent
T
The objective of
o this agent iss acting as a kind
k
of messenger
oor intermediary
y between the chatterbot an
nd the BDI ag
gent
IISMAEL. It is
i in charge of managing the events and
a
m
messages that can arise betw
ween them (thiis agent has beeen
im
mplemented ass a BDI agent too).
t
When the system
s
start up
p, an instance of
o EMMA willl be
ccreated, which will be runnin
ng waiting conn
nections from the
cchatterbot. Wh
hen a connectiion arises, it will
w be created
d a
thhread which will be in ch
harge of readiing the receiv
ved
m
messages. If a login messag
ge is received
d, then a stud
dent
w
wants to log in
nto the system
m, and so this messenger ag
gent
w
will launch an
n instance of ISMAEL (ob
bviously, when
n a
sstudent log offf the system, th
he messenger agent
a
will destrroy
thhe correspondiing instance off ISMAEL prev
viously created
d).
When the connection between the ch
hatterbot and the
nnew instance of
o ISMAEL is on, EMMA
A will format the
m
messages from
m the first one to
t a format thaat the second one
o
ccan understand
d, and then it will
w send these messages to this
t
laast one. Likew
wise, when IS
SMAEL wants to communiccate
w
with the chattterbot, the inv
verse process will take plaace
((Figure 2).
II
II. ISMAEL Ageent
IISMAEL is an
a agent whiich running under
u
the Jad
dex
pplatform, and so it is defineed by a XML file called ADF
((Agent Definitiion File) and a set of Java classes which will
w
eexecute its plaans. As we haave told beforee, the objectivee of
thhis agent is to
o guide the stu
udent through
h several learning
taasks, commun
nicating with him/her
h
throug
gh the chatterb
bot
((and the messen
nger agent).
Following,, we will brieefly present so
ome of its more
im
mportant BDI characteristicss:
Beliefs: Th
hey will be maainly compound
ded by a seriess of
references to: instances to
t access the ontology
o
modu
ule,
instance of the messeng
ger agent to co
ommunicate with
w
the chatterrbot, identifierr of the studeent who is being
tutorized, identifiers of the tasks the student is doiing,
temporizattions, etc. In short,
s
they aree a whole set of
parameterss (information
n) the agentt need to taake
decisions related to th
he goals to achieve
a
at every
moment, and
a the actions which will be carried out to try
reach thosee goals.
Desires: They are the go
oals. They con
nsist of managing
to do the student
s
does a series of educcational activitiies,
which hav
ve previously defined by a teacher of the
platform. Usually
U
these activities will be included in
na
semantic schema
s
which
h defines the relations amo
ong
FIGURE 2
W ORKING SCHEMA
A OF CHARLIE, E
EMMA, AND ISMA
AEL
Sessiion T1J
system willl assume that the student haas left his/her taask
without a communication about it. So, the system will
w
proceed to shutdown the session of thiss student (to saave
resources)..
To monito
or the time thatt the students spend with every
educationaal task: The semantic
s
know
wledge about the
learning ob
bjects includess information about the time a
student wo
ould spend wiith them. So, the actual sp
pent
time to fin
nish a task is really
r
importan
nt to know if this
t
expected tiime is properly
y set. Besides, the spent time by
a student to
o do his/her taasks is valuablee data to take in
nto
account in his/her learnin
ng profile.
II
III. Flow Diagrram of the Intelligent Tutoring Module
N
Now we are go
oing to briefly
y comment thee behavior of this
t
m
module when the student is already doing
g a learning taask
((Figure 3):
At last, we must point out that the selection oof the next
goal too achieve is a process whhich implies ttaking into
accountt both the learrning path andd the current ssituation of
the studdent related to his/her learninng tasks (ISMA
AEL is able
to accesss this knowleedge through tthe semantic innformation
stored iin the ontologyy of the system
m). For examplle, the next
selectedd goal could bee one of the folllows:
FIGURE 3
AGRAM OF THE INTELLIGENT TUTORIING MODULE.
FLOW DIA
Session T1J
him/her personalized tasks to achieve an specific learning
goal; and ii) EMMA, which acts as a messenger between
ISMAEL and a conversational agent to communicate with
the students.
ACKNOWLEDGMENT
[2]
[3]
[4]
[5]
[6]
Friedman-Hill, E., J., Jess, The Java Expert System Shell, Available
at http://web.njit.edu/all_topics/Prog_Lang_Docs/html/jess/, 2000.
[7]
[8]
[9]
[10] Braubach, L., Jadex, BDI Agent System, Distributed Systems and
Information Systems Group, University of Hamburg, Available at
http://jadex.informatik.uni-hamburg.de/bin/view/About/Overview,
2009.
[11] Howden, N., Ronnquist, R., Hodgson, A., Lucas, A., JACK
Intelligent Agents. Summary of an Agent Infrastructure, Proceedings
5th International Conference on Autonomous Agents, 2001.
[12] GNU Lesser General Public License, Available at
http://www.gnu.org/licenses/lgpl.html, 2007.
[13] Gosling, J., Joy, B., Steele G. Bracha, G., The Java language
specification, Addison-Wesley, 2000.
[14] Mikic, F. A., Burguillo, J. C., Llamas, M., Fernndez, D., Using
Semantics in INES, an Intelligent Educational System, 39th
Frontiers in Education Conference - FIE 2009 (San Antonio, Texas,
USA), 2009.
[15] Hodgins, W., Duval, E. Learning Object Metadata Standard, IEEE
1484.12.1-2002, 2002.
[16] Norton, M., Treviranus, J. IMS Learner Information Package
Information Model Specification, IMS Technical Report.
AUTHOR INFORMATION
Fernando A. Mikic Fonte received the M.Sc. degree in
Engineering in 2001 from the University of Vigo, Spain. He
is a Ph.D. candidate in Information Technologies at the
University of Vigo, Spain. He is currently an assistant
professor at the Department of Telematic Engineering at the
same university. His research interests include intelligent
agents, distributed artificial intelligence, e-learning, and
telematics services. mikic@det.uvigo.es
Juan C. Burguillo Rial received the M.Sc. degree in 1995
and the Ph.D. degree (cum laude) in 2001 from the
University of Vigo, Spain; both in Telecommunication
Engineering. He has been assistant professor during the
period 1996-2001 at the ETSI of Telecommunications
(Higher Technical School of Telecommunications
Engineering) in the University of Vigo, Spain. He is
currently an associate professor at the Department of
Telematic Engineering at the same university. His research
interests include intelligent agents, distributed artificial
intelligence, distributed optimization, and telematic services.
jrial@det.uvigo.es
Martn Llamas Nistal received the Telecommunications
Engineer degree (1986) and Doctor in Telecommunications
degree (1994), both from the Technical University of
Madrid, Spain. Since 1987 he has been lecturing at the ETSI
of Telecommunications (Higher Technical School of
Telecommunications Engineering) of the University of Vigo,
Spain (where he was assistant head during the period 19941997). He is currently an associate professor of the
Department of Telematic Engineering at the same university.
His areas of interest are protocol engineering, e-learning,
web engineering, and Internet based applications. Since
December 1998 to October 2003 he has been the Head of
ICT Area at the University of Vigo, Spain. He is a senior
member of the IEEE. martin@det.uvigo.es
David Fernndez Hermida is a M.Sc. degree candidate in
Engineering at University of Vigo, Spain. He is currently a
full-time researcher at the Department of Telematic
Engineering at the same university. His research interests
include intelligent agents, distributed artificial intelligence,
and e-learning. davidfh@gist.det.uvigo.es