An Architecture for Dynamic and Adaptive User Activity Planning
Systems
Jaime Pavlich-Mariscal
1
, Yolima Uribe
1
, Luisa Fernanda Barrera León
1
, Nadia Alejandra
Mejia-Molina
1
, Angela Carrillo-Ramos
1
, Alexandra Pomares Quimbaya
1
, Rosa Maria Vicari
2
,
Ramon Fabregat
3
and Silvia Baldiris
3
1
Departamento de Ingeniería de Sistemas, Pontificia Universidad Javeriana, Bogotá, Colombia
2
Departamento de Informática Teórica, Instituto de Informática,
Universidade Federal do Rio Grande do Sul, Porto Alegre, RS, Brasil
3
Dept. of Computer Technology and Architecture (ATC), University of Girona, Girona, Spain
Keywords:
Adaptation, Personalization, Dynamic Planning, Software Design.
Abstract:
A plan is an organized set of activities that are performed to achieve certain goals. Changing environments
have multiple requirements: users have diverse needs and preferences, the context may be different for each
user, depending on the time, location, and access devices (e.g., mobile devices, desktop computers). Above all,
these types of environments have a key element in common: they require the creation of dynamic and adaptive
plans, which can address different situations and provide the best set of activities for each user and context.
This paper proposes an architecture for the creation of dynamic and adaptive planning systems that can address
specific user needs and contexts. This specification encompasses the main components of a planning process
and can be translated into more concrete implementations. As part of the validation of this approach, this
paper describes the prototyping effort and an ongoing case study of an educational web application.
1 INTRODUCTION
A plan is “a detailed proposal for doing or achiev-
ing something" (Oxford-University, 2014), while
dynamic is something “characterized by constant
change, activity, or progress" (Oxford-University,
2014). A third important concept is adaptation, the
ability of a system to automatically change its behav-
ior and appearance to better satisfy the specific needs
of each user and the context in which the user interacts
with the system (Stewart et al., 2008; Casteleyn et al.,
2009). This research unifies all of the above concepts
into dynamic and adaptive planning, to create plans
especially suited for the requirementsof each user and
context and that is able to evolve over time.
In today’s world, there are many situations that re-
quire dynamic and adaptive planning. For instance, in
the classroom students are diverse, they have different
learning needs and preferences (Brusilovsky and Mil-
lán, 2007). They also have different access devices
and location. Therefore, it is necessary to especially
tailor learning plans for each of them (Sangineto et al.,
2008). Moreover, as students progress in their learn-
ing, their characteristics change, which may require
the modification of their learning plans. Another sim-
ilar situation occurs in physical training. There is also
a diversity of people who require different training ac-
tivities and as they progress they require change in
their plans to accommodate to their physical improve-
ment.
Overall, multi-user systems that guide users in the
execution of a plan may significantly benefit from dy-
namic and adaptive planning. This paper proposes
ASHYI, an architecture for dynamic and adaptive
planning systems. The word ASHYI in Quechua lan-
guage means to search, to investigate. ASHYI de-
fines the most important structures and processes re-
quired for these kinds of systems and provides a way
to translate those specifications into a more detailed
and concrete design. To illustrate the applicability of
this approach, this paper also describes the design and
prototyping effort of an educational application and
an ongoing case study that utilizes this application in
a university course.
The remainder of this paper describes ASHYI and
its validation. Section 2 discusses related work. Sec-
228
Pavlich-Mariscal J., Uribe Y., Barrera-León L., Mejia-Molina N., Carrillo-Ramos A., Pomares Quimbaya A., Vicari R., Fabregat R. and Baldiris S..
An Architecture for Dynamic and Adaptive User Activity Planning Systems.
DOI: 10.5220/0005430302280235
In Proceedings of the 11th International Conference on Web Information Systems and Technologies (WEBIST-2015), pages 228-235
ISBN: 978-989-758-106-9
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
tion 3 explains the main components of ASHYI. Sec-
tion 4 describes the design of ASHYI-EDU, an instan-
tiation of ASHYI for the educational domain. Section
5 describes an ongoing case study to validate ASHYI-
EDU in a university course. Section 6 concludes and
discusses future work.
2 RELATED WORK
Table 1 summarizes the analysis of the related work.
The (+) symbol means that the approach fully sup-
ports the given criterion. The (-) symbol means that
the approach does not support it. The (#) symbol
means that the criterion is supported to a certain de-
gree. The (?) symbol means that it is not possible to
determine whether that criterion is supported or not.
There are several works in dynamic activity plan-
ning. The first group includes proposals for plan-
ning activities in specific contexts. Nejad el al.
(Nejad et al., 2011) (1) propose a dynamic activity
planning system for a manufacturing process. Us-
ing multi-agents and negotiations based on decision-
rules, the system manages resources and resolves con-
flicts among elements of the plan. Finally, Auld and
Mohammadian developed a system to plan travel ac-
tivities through a multi-agent system and an activity-
based model (Auld and Mohammadian, 2012) (2). In
this work, horizontal planning is performed to char-
acterize each activity and its environment to predict
activities to perform in the short term. The plan
adapts to external constraints, past experiences, and
user needs.
The second group of proposals define generic sys-
tems or frameworks to create dynamic plans. Wang
(Wang, 2006) (3) presents a dynamic activity plan-
ner that integrates a general plan and plans associated
with different roles who share resources. Planning is
resource-centered, so it uses negotiation to organize
activities. Another project created by Cavazza et al.
(Cavazza et al., 2008) (4) proposes a dynamic plan-
ner for daily activities. A physical device interacts
with the user to determine his/her opinion and behav-
ior. Using that information, together with the user,
preferences, activities, and context profiles, the sys-
tem adapts and updates the user’s daily activity plan.
In the same way, Nijland et al. (Nijland et al.,
2009) (5) propose an application to plan daily activ-
ities utilizing a multi-agent system and a user needs
model, which includes social, personal, home needs,
among others). The system proposes and plans activ-
ities and events. The utilization of needs and context
information assists the prediction of an event and to
re-plan the user activities, according to their useful-
ness for the user.
The last group includes proposals specifically de-
signed for learning environments. The approach of
(Rytikova and Boicu, 2014) (6) utilizes a set of re-
sources provided by a teacher to create a course. Stu-
dents perform an initial test to measure competences.
This test is utilized by the system to convey differ-
ent information to students, based on their specific
characteristics. Lecomp5 (Limongelli et al., 2008) (7)
is a system to plan student activities, based on their
knowledge (initial and goal), learning style, course
progress, and learning objects and their prerequisites.
Lecomp5 provides tools to manage content, realize
different teaching strategies, and provide learning se-
quences to students. Finally, Capuano et al. (Capuano
et al., 2008) (8) propose LIA, an intelligent e-learning
system to provide personalized planning to students.
A generation algorithm utilizes teaching and learn-
ing preferences and cost constraints to select a set of
learning activities that address all of the concepts that
the student has not learned until that moment. The
work of (Bahmani et al., 2011) (9) is a system for
course recommendation to students that utilizes some
student attributes and context.
Table 1: Related Work Comparison.
Criterion (1) (2) (3) (4) (5) (6) (7) (8) (9)
Activ. Analysis - + # + + + # # -
Res. Mgmt. + - + - - # - - #
User Adapt. - # - + + + + + +
Plan Feedback - + - + - + ? - ?
Role Mgmt. + - + - ? ? - - +
Re-planning + + + + + - + + ?
Most approaches do not take into account detailed
user adaptation aspects that can affect activity execu-
tion. Only two works provide feedback to the plan
from the user, which limits their ability to improvethe
plan. In addition, several of them do not take into ac-
count activity resources and the way these resources
affect activity execution. All of the related works
were developed for specific environments, which lim-
its the user and context characteristics.
Taking into account the above analysis, the contri-
bution of our approach is two-fold:
1. Unlike other planning approaches, ASHYI is both
dynamic and adaptive. ASHYI is adaptive, since
is able to change create specific plans for each
user, based on context and user characteristics. It
is dynamic, since it is able to analyze its interac-
tion with users, determine the way users and con-
text change over time, and perform re-planning if
necessary. In addition ASHYI can also dynami-
cally assign resources, based on availability.
AnArchitectureforDynamicandAdaptiveUserActivityPlanningSystems
229
2. Above all, ASHYI is conceived as a generic ar-
chitecture for dynamic adaptive planning systems.
This means that ASHYI has the potential to be
seamlessly reused in different projects that require
to dynamically plan according to specific user and
context characteristics. In this sense, our contri-
bution is to clearly identify the most common fea-
tures of these types of adaptive systems, which fa-
cilitates the understanding of the overall planning
problem in adaptation.
3 THE ASHYI ARCHITECTURE
ASHYI was conceived to facilitate the design of sys-
tems that require dynamic and adaptive planning. As
such, it provides a high-level specification of all of the
main processes and entities required to perform adap-
tation of a plan and the ways to dynamically adjust
them over time.
Figure 1 is an overview of the main processes of
ASHYI and data flows between them. The following
sections explain this figure in detail.
(c) Activities
(g) Select
cantidate
activities
(h) Create
meta-plan
(j) Create
adapted
plan
(i) Meta-plans
(l) Adapted
plans
(d) Resources
(f) Planner
(m) Execute plan
(p) Update
user profile
(n) Executor
(o) Execution
history
(e) User
profiles
(k) Context
plan goals
activities
resources
selected
activities and
resources
meta-plan
meta-plan
user
profile
context
information
adapted
plan
activity of
the plan
activity
input data
activity
execution
results
activity execution results
updated
user
profile
(b) Configure
Environment
(a) Administrator
activities,
resources,
user
information
activities
resources
user
profiles
Figure 1: Processes and Data Flows of ASHYI.
The Configure environment (b) process of Figure
1 focuses on defining the specific planning elements.
In particular, the Administrator (a) specifies: activi-
ties, resources, and user characteristics.
The specified activities include: a description of
the actions to perform as part of the activity, precon-
ditions (assertions that must be true in order to prop-
erly execute the activity), and postconditions (asser-
tions that are expected to be true after the successful
execution of the activity).
Each activity may require resources to be per-
formed. The Administrator also specifies the avail-
able resources for each activity and the context con-
ditions that determine their availability. For instance,
if the user needs to access a journal paper that is ac-
cessible only within the network of a university that
is subscribed to that journal, the paper may be un-
available if the user tries to access it from an outside
network.
The Administrator can also specify the user infor-
mation, i.e., the individual user characteristics that are
used by the system to decide which activities are the
most suitable for each user. For instance, based on
information collected directly from users, the admin-
istrator may specify which users have some specific
competencies. This information may also be collected
by automated surveys and tests (not shown in the fig-
ure).
An important principle of ASHYI is that every
plan must achieve a set of predefined goals, which
means that each goal must be satisfied by one or more
activities of that plan.When the Planner (f) user asks
the system to create a plan, he/she must specify the
goals of that plan. Based on that information, the sys-
tem selects (g) those activities and resources that can
satisfy the given goals when combined together.
Using the selected activities and resources, this
process (h) creates a meta-plan, a structure from
which all of the possible different plans can be ex-
tracted. For instance, if one wants to create plans with
strict pre and post condition precedence, one can uti-
lize the GraphPlan algorithm (Blum and Furst, 1997),
which yields the meta-plan in the form of a graph that
condenses all of the possible plans that could be ob-
tained from the input activities. Of course, this is not
the only option. The prototype described in this paper
(see Section 4) utilizes a different structure to store a
meta-plan.
The meta-plan is the base to create specific plans
adapted to each user. This process (j) navigates
through the meta-plan to extract an adapted plan.
This plan is an instantiation of parts of the meta-plan
that includes activities specially tailored for a given
type of user. The set of activities in the adapted
plan must satisfy the goals given by the Planner (f)
user. Overall, the selected activities and resources in
this plan are the best suited for that particular user,
based on his/her profile and context, and according to
domain-dependent criteria.
ASHYI provides to the Executor (n) user the activ-
ities that he/she must execute to accomplish the plan.
Depending on the planning algorithm utilized in pro-
cesses (h) and (j), these activities may be provided in
a strict sequence, or the user may have the freedom to
WEBIST2015-11thInternationalConferenceonWebInformationSystemsandTechnologies
230
choose the execution order of some or all of them.
Each activity executed by the user provides spe-
cific results that are stored by the system in a history
repository (o) to later provide feedback to the user.
This process (p) utilizes the activity execution log
of a specific user to update the user profile. This up-
dated information is further utilized by ASHYI to re-
create the adapted plan for that user, if necessary.
For instance, if the Executor user performs a learn-
ing activity, he/she may gain new specific compe-
tences or, if the activity is not successful, he/she may
only gain a subset of the expected competences that
are postconditions of the learning activity. If the latter
happens, the current adapted plan for that user may
not be the optimal, since that plan may have been cre-
ated under the assumption that the user successfully
executed all of his/her activities. From the updated
user profile information, ASHYI can infer that the
user needs a new plan, so it executes process (j) again
and yields a new adapted plan.
4 ASHYI-EDU
The ASHYI Architecture provides the basis to cre-
ate dynamic adaptive systems. To validate our pro-
posed approach, we instantiated ASHYI into a spe-
cific design of a system called ASHYI-EDU. ASHYI-
EDU is a software component that enhances web-
based Learning Management Systems (LMS) with
dynamic adaptive planning. ASHYI-EDU can be
used to specify learning activities within a course and
it can generate adaptive plans for each student, based
on their specific competences, skills, and learning
styles. ASHYI-EDU can automatically modify plans
when students have not correctly learned the required
competences and skills of each activity, and can sug-
gest remedial activities to facilitate future learning
within the course.
This section focuses on illustrating the process of
instantiating the ASHYI architecture into the design
of a dynamic adaptive system. For more details about
the implementation of ASHYI-EDU, the reader can
refer to (Jaime Pavlich-Mariscal and Martin, 2015)
ASHYI-EDU’s design is based on ASHYI, thus
each of the relevant roles, processes and repositories
of Figure 1 are concretely defined for the educational
domain.
ASHYI-EDU defines a specific component model
to realize the ASHYI architecture, as shown in Fig-
ure 2. The Logic component contains all of ASHYI-
EDU’s business Logic. The Environment Configura-
tor component lets the Teacher to specify the activi-
ties, resources, and students of a course . The selec-
Figure 2: ASHYI-EDU Component Model.
tion of candidate activities is performed by PUMAS-
Lite. The Planner utilizes PUMAS-Lite to retrieve ac-
tivities and resources and create the meta-plan and the
adapted plans. Plan execution and profile updating
are performed by several components. The Executor
assists the students in the execution of their assigned
activities. The Event broker connects the Executor
and the Monitor using a publish-subscribe approach.
The Monitor supervises the events generated by ac-
tivity executions and utilizes the Notifier to trigger a
re-planning process in the Planner component, which
also may delegate to the Environment Configurator to
update the profiles.
The overall functionality of ASHYI-EDU is de-
scribed as follows.
4.1 Roles
In the ASHYI-EDU, the ASHYI roles are specialized
for the educational domain. The roles in ASHYI-
EDU are the following:
Teacher. This role combines the roles Administrator
(a) and Planner (f) from Figure 1. The Teacher can
specify activities, resources, and students who partic-
ipate in the course, and can also indicate the goals that
ASHYI-EDU requires to create adapted plans for stu-
dents.
Student. This role is equivalent to the Executor (n)
role of Figure 1.
AnArchitectureforDynamicandAdaptiveUserActivityPlanningSystems
231
4.2 Plan and Execution
The instantiations of processes (g), (h), and (j) are
highly interrelated in ASHYI-EDU, thus they are ex-
plained as a whole.
For process (g), the module of ASHYI-EDU that
selects activities for a meta-plan is called PUMAS
LITE. This is a simplified version of PUMAS
(Carrillo-Ramos et al., 2007), a multi-agent frame-
work for nomadic users. PUMAS provides users with
information adapted to the specific user and context
characteristics, regardless of the access device (mo-
bile or not). PUMAS provides a method to query
different sources of information (e.g., servers, plain
text files in clients, etc.). In addition, PUMAS can
automatically act as a broker between mobile de-
vices and servers, depending on the query informa-
tion; the agents can autonomously connect and dis-
connect from PUMAS.
PUMAS LITE focuses on selecting activities and
resources, based on given goals for specific plans.
The result of this selection is a set of activities and re-
sources that satisfy the goals provided by the Teacher.
For process (h), the meta-plan may be structured
in several ways. One option that was analyzed was
GraphPlan (Blum and Furst, 1997), which models ac-
tivities with hard pre and post conditions and is able
to chain them into plans that satisfy specific goals.
However, in the specific context of ASHYI-EDU, it
is expected that some students should be able to exe-
cute latter learning activities even if they fail previous
ones. Therefore, pre and post conditions should not
be hard, but flexible.
Instead, the meta-plan is structured as a multipar-
tite graph, as shown in Figure 3. Each node repre-
sents an activity with its corresponding resource set.
Columns surrounded by an ellipse correspond to the
main activities, i.e., activities that can be used to ac-
complish a specific learning goal in a course. Each of
these columns is aligned to a particular course goal.
For instance, nodes A
i
in Figure 3 are all of the dif-
ferent activities that can satisfy goal A. Columns sur-
rounded by a rectangle correspond to remedial activi-
ties that the student may execute between main activ-
ities.
All of the main activities in each column are con-
nected with all of the remedial activities of the adja-
cent column on the right and also to all of the main
activities of the column adjacent to the right of the
remedial activities (the latter are not shown for space
reasons). For instance, in Figure 3 each node A
i
is
connected to all of the nodes R
j
of the contiguous col-
umn and is also connected to all of the nodes B
k
.
The process (j), to create the adapted plan,
Figure 3: ASHYI-EDU Meta-Plan and Adapted Plan Ex-
ample.
searches for the shortest path in the graph, which will
include at least one main activity per learning objec-
tive and zero or more remedial activities. Figure 3
shows an example adapted plan as nodes and connec-
tions with bold outlines. The weight of each graph
connection is given by a similarity measurement be-
tween the profile of a given student and the activity
to which the connection is pointing. For instance, as-
sume that –after successfully executing activity A
1
–
the similarity measure between the student profile and
activity B
4
is 5.23. Then, in figure 3, the weight of
the connection between activity A
1
and Activity B
4
is 5.23. The similarity formula is outside the scope
of this paper and it is described in (Jaime Pavlich-
Mariscal and Martin, 2015).
4.3 Matching Students to Activities
To determine the similarity measure between users
and activities, both have a common profile structure,
shown in Figure 4. The structure is a vector of nu-
meric values between 0 and 1. Each number repre-
sents the degree to which a certain attribute is present
in a student or whether it can be satisfied or required
by activities.
Figure 4: Student and Activity Vectors.
The attributes include 16 competences, 19 skills,
4 learning styles, and 16 personality traits. If a user
profile has a specific attribute value greater than zero,
it means that the student has that attribute, i.e., the
user has a particular competence, skill, learning style,
or personality trait.
Activities have two of those vectors, one for pre-
conditions and another for postconditions. If an at-
tribute has a value greater than zero in the precondi-
tions vector of the activity profile, it means it is de-
sirable that a student has that attribute to execute the
activitysuccessfully. If an attribute has a value greater
WEBIST2015-11thInternationalConferenceonWebInformationSystemsandTechnologies
232
than zero in the postconditions vector of the activity
profile, it means that the successful execution of the
activity will enrich the student on that attribute.
The similarity measure between a student and an
activity is given by a function that combines differ-
ent distance metrics between the student vector and
the activity preconditions vector. This measure rep-
resents how appropiate is the activity for the student,
according to his/her profile. The lower the distance,
the more recommendable is the activity for the stu-
dent.
4.4 Context
Whenever the context makes some activities unavail-
able, the ASHYI-EDU automatically creates a new
plan. For instance, if a given resource in an activity
is not available, ASHYI-EDU automatically creates a
new plan that may include that same activity, but us-
ing an available resource.
Currently, ASHYI-EDU determines resource
availability, based on two variables: access device and
location. Resources that cannot be adequately opened
in mobile devices are marked as unavailableif the stu-
dent tries to access them from a limited mobile device.
Similarly, location also determines if a resource can
be used or not. For instance, if a given resource is an
URL that can only be accessed from the University’s
internal network, it cannot be used if the student is
accessing from outside the campus.
Overall, ASHYI-EDU detects the access device
and IP of the device and marks the corresponding re-
sources as unavailable. Based on this information,
ASHYI-EDU creates a new adapted plan for the stu-
dent in which activities that use those resources are
excluded from the planning process.
5 CASE STUDY
ASHYI-EDU is currently being validated in a uni-
versity course for students of the career of Primary
School Teacher. The first part of the validation is an
undergrad elective course called "Learning to Learn
in the Web", offered during the Fall semester of 2014.
The course utilizes blended learning and its goal is
to give the student tools to learn in the web. The
study involvedall of the course 29 participants: 6 men
and 23 women. The students are from different back-
grounds (Education, Accounting, Information Sci-
ences, Psychology, Software Engineering, Business
Administration, and Microbiology).
To effectively utilize ASHYI-EDU in the class-
room, this component was integrated into Sakai
(Apereo-Foundation, 2014), a web-based LMS. In
particular, we modified the lesson builder component
to include ASHYI-EDU’s dynamic planning facili-
ties.
The course is divided into several learning units,
each covering coherent groups of topics. Each learn-
ing unit has its own dynamic adaptive plan. Teach-
ers specify learning goals for each learning unit. The
PUMAS LITE component automatically selects the
activities that match those goals and are taken as in-
put by ASHYI-EDU to create the multipartite plan-
ning graph. Figure 5 shows a planning graph as it is
generated by the tool for the teacher. Main activities
are colored orange, while remedial activities are col-
ored blue.
In addition, student profiles were initially built
from surveys completed by the students. These sur-
veys are based on the Chaea questionnaire (Alonso
et al., 2009) and can be used to determine the student
learning styles.
Figure 5: Planning Graph Example in ASHYI-EDU.
Using this information, ASHYI-EDU utilizes the
process indicated in Section 4.2 to automatically cre-
ate the best plan for each student to complete the goals
of each learning unit.
To illustrate the capabilities of ASHYI-EDU, con-
sider two students with different profiles:
Student A. Skills: Comprehension, relating reality
with the environment, interpreting and analyzing in-
formation, observation, interest and initiative to learn,
agility and adaptability, interpersonal relations and
managing emotions and feelings. Personality: ex-
troversion, intuition, thinking, judgment. Learning
styles (max. score: 20): Active - level: 13, Pragmatic
- level: 11, Reflexive - level: 12, Theoretical - level:
12.
Student B. Skills: Comprehension, expression of
ideas, relating reality with the environment, interpret-
ing and analyzing information, generating new ideas,
raising critical reflection questions, answering based
on knowledge and experience, interest and initiative
to learn, agility and adaptability, interpersonal rela-
tions, and managing emotions and feelings. Personal-
ity: extroversion, intuition, feeling, judgment. Learn-
ing styles: Active - level: 8, Pragmatic - level: 6, Re-
flexive - level: 14, Theoretical - level: 16.
AnArchitectureforDynamicandAdaptiveUserActivityPlanningSystems
233
Figure 6: Adapted Plan for Students A and B.
When students access a learning unit in ASHYI-
EDU, a different plan is shown for each of them. Fig-
ure 6 shows the adapted plans for Student A and B.
Each student plan has some main activities and, in
some cases, remedial activities. These adapted plans
depend on the student’s characteristics so they differ
for both students. For instance, the first activity of
Student A is designed for people with active and prag-
matic learning styles, while the activity for Student B
addresses individuals with theoretical learning style.
In both cases the activities address the main learning
style of each student.
Remedial activities are included when the student
needs to strengthen some attribute that is required by
a main activity. For Student A, the system plans a
remedial activity to strengthen critical thinking before
the last class activity (see Figure 6). For Student B,
the plan includes, before the first activity, a remedial
activity to strengthen search and selection skills.
ASHYI-EDU supports a simple workflow that
teachers and students can follow when executing an
activity. This process is summarized as follows:
When the Student opens the learning unit, the sys-
tem shows the adapted plan to the student, which
highlights only those activities the Student is allowed
to perform, according to the plan (see Figure 6). Ac-
tivities not yet available are marked with an hour-
glass icon. Activities in progress are marked with a
pen icon. Finished activities are either marked with a
happy face or a sad face, depending on the teacher’s
feedback.
The Student can complete an activity and send
his/her answer to the Teacher, using simple web form.
The teacher receives the activity, and uses the same
form to grade it, send feedback, and indicate which
attributes (competences, skills, learning styles, or per-
sonality traits) the student has acquired after execut-
ing the activity. The system updates the Student pro-
file with the acquired attributes and creates a new
adapted plan, if necessary.
The student can later see the Teacher’s feedback
through that same form, or can choose to perform new
activities that may become available thereafter. This
cycle continues until the student has finished all of the
planned activities for the learning unit.
5.1 Overview of the Results
Students utilized the system an average of 11.6 times
during the semester, which is close to the total amount
of learning goals of the course (there are 11 learn-
ing goals in the course). Students were interviewed
to determine their satisfaction. Overall, the results
were positive. Students felt that the course effectively
adapted to their particular characteristics and chose
the most adequate activities for them.
Teachers, on the other hand, had to invest more
time preparing the course, since they had to create
more than one activity per learning goal, in addition to
the remedial activities. There was an additional effort
to characterize each activity vector (see Section 4.3).
However, this effort should be significantly reduced
as the activity and resources repository is expected to
grow as more teachers utilize the system and share
their materials.
6 CONCLUSIONS AND FUTURE
WORK
This paper presented ASHYI, an architecture to de-
sign dynamic adaptive systems. ASHYI specifies
a set of processes to create adapted plans that are
aligned with the specific user characteristics and
his/her context information to achieve a specific goal.
The structure of ASHYI is sufficiently abstract to
address different types of adaptive planning systems,
but at the same time it is sufficiently specific to pro-
vide the essential design elements to instantiate it into
a specific domain of application.
We illustrated the capabilities of ASHYI through
ASHY-EDU, a component integrated into the Sakai
LMS to provide dynamic adaptive planning for uni-
versity courses, which is currently being utilized in a
university course as a case study. Although the case
study is still in progress, the current tests on the ap-
plication demonstrate that the system is able to cap-
ture the essential student characteristics and provide
an adapted plan according to their specific attributes.
Even though ASHYI-EDU focuses on education,
the overall graph-based plan structure and the plan-
ning process as an optimal route search, have both a
high reuse potential for other problems with clearly-
defined phases, but multiple choices of activities
to perform. Nevertheless, the architecture behind
ASHYI-EDU (i.e., ASHYI) is sufficiently generic to
be applied to other problems in which the design of
ASHYI-EDU may be inadequate.
Future work is to complete the case study and
analyze the strengths and weakness of ASHYI and
WEBIST2015-11thInternationalConferenceonWebInformationSystemsandTechnologies
234
ASHYI-EDU, then execute a new case study the
Spring semester of 2015.
ACKNOWLEDGEMENTS
This paper is part of the project “ASHYI: Plataforma
basada en agentes para la planificación dinámica, in-
teligente y adaptativa de actividades aplicada a la ed-
ucación personalizada”, executed by the ISTAR re-
search group of the Pontificia Universidad Javeriana,
cofinanced by Colciencias, project #1203-569-33545.
REFERENCES
Alonso, C., Gallego, D., and Garcia, J. (Sep-
2009). CHAEA - estilos de aprendizaje.
http://www.estilosdeaprendizaje.es/menuprinc2.htm.
Apereo-Foundation (Sep-2014). Sakai.
https://sakaiproject.org/.
Auld, J. and Mohammadian, A. (2012). Activity plan-
ning processes in the agent-based dynamic activity
planning and travel scheduling (ADAPTS) model.
Transportation Research Part A: Policy and Practice,
46(8):1386–1403.
Bahmani, A., Sedigh, S., and Hurson, A. (2011). Context-
aware recommendation algorithms for the percepolis
personalized education platform. In FIE, 2011, pages
F4E–1–F4E–6.
Blum, A. L. and Furst, M. L. (1997). Fast planning through
planning graph analysis. Artificial Intelligence, 90(1-
2):281–300.
Brusilovsky, P. and Millán, E. (2007). User models for
adaptive hypermedia and adaptive educational sys-
tems. In The Adaptive Web, number 4321 in Lec-
ture Notes in Computer Science, pages 3–53. Springer
Berlin Heidelberg.
Capuano, N., Gaeta, M., Miranda, S., Orciuoli, F., and
Ritrovato, P. (2008). LIA: An intelligent advisor for e-
learning. In Lytras, M. D., Carroll, J. M., Damiani, E.,
and Tennyson, R. D., editors, Emerging Technologies
and Information Systems for the Knowledge Society,
number 5288 in Lecture Notes in Computer Science,
pages 187–196. Springer Berlin Heidelberg.
Carrillo-Ramos, A., Villanova-Oliver, M., Gensel, J., and
Martin, H. (2007). Knowledge management for
adapted information retrieval in ubiquitous environ-
ments. In Webist, number 1 in Lecture Notes in Busi-
ness Information Processing, pages 84–96. Springer
Berlin Heidelberg.
Casteleyn, S., Daniel, F., Dolog, P., and Matera, M. (2009).
Engineering Web Applications. Springer Berlin Hei-
delberg, Berlin, Heidelberg.
Cavazza, M., Smith, C., Charlton, D., Zhang, L., Tu-
runen, M., and Hakulinen, J. (2008). A ’com-
panion’ ECA with planning and activity modelling.
pages 1281–1284. International Foundation for Au-
tonomous Agents and Multiagent Systems.
Jaime Pavlich-Mariscal, Yolima Uribe, L. F. B. L. N. A.
M.-M. A. C.-R. A. P. Q. M. B. M. S. R. M. V. and
Martin, H. (2015). ASHYI-EDU: Applying Dynamic
Adaptive Planning in a Virtual Learning Environment.
Submitted to Webist. In Webist.
Limongelli, C., Sciarrone, F., Temperini, M., and Vaste, G.
(2008). Lecomps5: A framework for the automatic
building of personalized learning sequences. In Ly-
tras, M. D., Carroll, J. M., Damiani, E., and Tennyson,
R. D., editors, Emerging Technologies and Informa-
tion Systems for the Knowledge Society, number 5288
in Lecture Notes in Computer Science, pages 296–
303. Springer Berlin Heidelberg.
Nejad, H., Sugimura, N., and Iwamura, K. (2011).
Agent-based dynamic integrated process planning and
scheduling in flexible manufacturing systems. Inter-
national Journal of Production Research, 49(5):1373–
1389.
Nijland, L., Arentze, T., and Timmermans, H. (2009).
Multi-day activity scheduling reactions to planned
activities and future events in a dynamic model of
activity-travel behavior. Journal of Geographical Sys-
tems, pages 1–17.
Oxford-University (Sep-2014). Oxford dictionaries.
http://www.oxforddictionaries.com.
Rytikova, I. and Boicu, M. (2014). A methodology for per-
sonalized competency-based learning in undergradu-
ate courses. SIGITE ’14, pages 81–86, New York,
NY, USA. ACM.
Sangineto, E., Capuano, N., Gaeta, M., and Micarelli, A.
(2008). Adaptive course generation through learning
styles representation. Universal Access in the Infor-
mation Society, 7(1-2):1–23.
Stewart, A., Niederée, C., and Metha, B. (2008). State of the
Art in User Modelling for Personalization in Content,
Service and Interaction.
Wang, R. (2006). Agent-based collaborative plan adap-
tation with resource constraints. PhD thesis, Penn-
sylvania State University, University Park, PA, USA.
AAI3343728.
AnArchitectureforDynamicandAdaptiveUserActivityPlanningSystems
235
