Predictive Analytical Framework based on Formal Method to Enhance

Mobile and Pervasive Learning Experience

Manel BenSassi, Mona Laroussi and Henda BenGhezala

Riadi Laboratory, National School of Computer Science, Manouba University, Tunisia

Keywords:

Learning Experience Design, Mobile Learning Scenario, Reliability, Predictive Analytics, Formal Method.

Abstract:

In this paper, we present a predictive analytical framework for mobile and ubiquitous learning environment

based on three main dimensions: learner, contextualized activity and space. The main objective of this proposal

is to assist pedagogical designer in developing engaging and effective courses by focusing on the learner’s

experience and his learning environment. To do that, a solid structure is essential to organize correlations

between different activities and to assess their reliability with the learner’s context. The strengths of our

proposal lie in the fact that, in a formal manner through friendly graphical interfaces, it allows pedagogical

designers:(1) to specify, model, simulate, analyse and verify different types of context-aware and adaptive

learning activities and their related contexts, (2) to assess the reliability of the indoor and outdoor learning

spaces within pervasive environment through factual cases and to experiment various learning scenarios, (3)

To simulate and to verify interactions and co- adaptability rules between learner, contextualized activity and

space.

1 INTRODUCTION

Mobile and ubiquitous technologies have emerged as

facilitators in learning process, both in formal and

informal educational contexts, offering new ways to

access and use learning resources and drawing the

mobile learning features. These technologies, com-

bined with cloud services, foster learner interactions

based on access to rich content across locations and

any times using portable equipments (such as wireless

laptops, personal digital assistants and smart-phone)

which, in turn, might enable new self-regulated lear-

ning experiences (Moser, 2017). Then, students are

oftenly asked to use their life experiences to make me-

aning of material introduced in classes (Kuh, 1996)

and the question is not whether mobile devices are

suitable as learning tools or not, but ”How can we

improve learners’ experience in mobile learning con-

text?”

With the emergence of adaptive mobile and ubiqui-

tous learning environment and the improvement of

teaching methodology, straightforward mediums are

evolving into wealthy and interesting learning expe-

riences. But in the other hand, learning scenario de-

sign gains importance and complexity and it becomes

a complex process (Clark and Mayer, 2016).

While an authoring tool is useful and necessary, there

is a real temptation to start using it straight away, be-

fore a clear idea of content and structure has been

developed. However, once the scenario is broken

up into its separate activities in the authoring tool,

it is difﬁcult to analyse the correlation between these

brick and its convenience with the deployment’s con-

text. If the scenario has not been well-analysed at

early stage, the learning environment could be ina-

dequate and it does not tie to the pedagogical out-

comes. Consequently, learners could ﬁnd difﬁcult to

navigate through. Without an analytical framework,

designing mobile and context-aware lesson become a

complex process and focus on pedagogical objectives

rather than the learning experience (Westera, 2011).

In this sense, we introduce an analytical framework

for mobile and context-aware learning scenario based

on formal methods and consider three main dimensi-

ons: learner, contextualized activity and space. The

strengths of our proposal lie in the fact that, in a for-

mal manner through friendly graphical interfaces, it

allows designers to analyse and assess the reliability

of adaptive mobile learning activities at early stage.

This paper is organized as follows. In section 2,

we are going to present some context-aware learning

challenges that should be considered in order to en-

hance mobile learning experience. Section 3 descri-

bes key dimensions and features of our analysis fra-

BenSassi, M., Laroussi, M. and BenGhezala, H.

Predictive Analytical Framework based on Formal Method to Enhance Mobile and Pervasive Learning Experience.

DOI: 10.5220/0006935802310238

In Proceedings of the 14th International Conference on Web Information Systems and Technologies (WEBIST 2018), pages 231-238

ISBN: 978-989-758-324-7

231

mework. We outline how this abstract deﬁnition of

different learning scenario’s components can improve

our understanding and the current design practice.

Section 4 focuses on our conceptual architecture de-

ﬁnition while its supporting tools and modules tests

are described in section 5. Section 6 concludes this

article with perspective and future works.

2 MOBILE LEARNING

CHALLENGES

The relevant literature identiﬁed different require-

ments and several technologies prerequisites for qua-

lity of mobile and ubiquitous learning.

From pedagogical point of view, the m-learning sce-

nario should have clearly explicit pedagogical design

principals with high interactivity appropriate to lear-

ner preference, needs and context which enables his

mutual feedback with tutor and assists him in the

identiﬁcation of knowledge gaps.

The purely pedagogical aspects of quality in m-

learning system are certainly important, but coupled

with equally important technical aspect of quality. In

fact, digital devices play an important role. Among

these technologies, mobiles devices (laptops, smartp-

hones, tablets, etc) are crucial for learning in mo-

bile and context aware learning environment. They

should fulﬁll several criteria, at least to a certain de-

gree. These include mobility, accessibility, conve-

nience, interactivity, study auxiliary, scalability and

application costs (Moser, 2017).

M-learning technologies that are reliable, technically

ﬂexible and contain available resources are those are

most likely to be successful. In this context, there are

a number of aspects of M-learning quality that can be

assessed from a technical perspective. Mobile lear-

ning scenario are tiny closed to several location-based

services in order to enable learner to interact with

real-life objects and surrounding computing node.

Therefore, location-based applications and environ-

ments allow learners to transcend the boundaries of

mobile devices.

A signiﬁcant aspect of mobility is quality of service in

terms of the reliability and speed of wireless connecti-

ons. Although some learning content can be downlo-

aded to a mobile device and used locally, the limita-

tions on storage mean that network connectivity is an

essential component of most mobile learning environ-

ments. The reliability and speed of such connections

can inﬂuence which media types can be used in an M-

learning system, for example video streaming is only

feasible over a high speed connection.

Another technical aspect of M-learning quality is

the limitation of energy and battery consumption on

many mobile devices, with certain mobile device ope-

rating systems and software platforms supporting dif-

ferent types of display.For instance, communication,

collaboration and interactivity issues can be affected

by more than one of the learning contexts community,

facilities, time and location activity and learner. Thus,

we conclude that a quality technical assessment for

M-learning environment should be encompass both

functional and contextual aspects.

From learning experience (LX) designing point of

view, the LX designer has complicated, challen-

ging and versatile responsibilities with several duties

that mostly concerns the improvement and adapta-

tion of learning scenario to the development of digi-

tal technology. Based on user experiences theories,

learner experience design is an iterative process that

combines several activities such as: (a)Content and

planning designing, (b) prototyping and Testing furt-

her improvements and development, and (c) Execu-

tion and Analytic (Davidson-Shivers et al., 2018).

Because the focus of learning is no longer on the sim-

ple acquisition of information but rather on actively

interacting with the current context that encompass

the information, the challenge of the learning expe-

rience design is to establish a connection between bu-

siness aims and user’s needs and expectations by tes-

ting and understanding the speciﬁcations of both sides

in the process.

3 KEY DIMENSIONS FOR OUR

PREDICTIVE ANALYTICAL

FRAMEWORK

Our starting point has been a previous mobile learning

assessment position that there are two dimensions to

take into account in modelling and evaluating perva-

sive learning scenarios: Learner and learning scenario

and their contexts (BenSassi et al., 2014). We propose

in this work to widen the initial approach and to take

into account Space as a key dimension in the mobile

environment engineering, and not only as a contex-

tual element related to learner. Our Framework leads

us to consider mobile and pervasive learning systems

in terms of three key dimensions: Learner, Learning

scenario and Space.

The essence of our approach to the learning evalua-

tion approach is the effective integration of space, by

allowing teachers and pedagogical designers to assess

the reliability of educational mobile application and

its convenience to the physical space and to embed

digital and smart components within it. Our approach

WEBIST 2018 - 14th International Conference on Web Information Systems and Technologies

232

Figure 1: Key Dimensions of our predictive analytical Fra-

mework ReStart-Me.

has the potential of being a scenario driven analytical

framework for the engineering and re-engineering of

mobile and pervasive learning systems.

This framework models all possible interactions and

constraints between Learner, Scenario and Space be-

cause they inﬂuence each others in learning processes.

Three types of technology- driven interaction are eva-

luated (see ﬁg. 1) (1) Learner-Space Interaction (2)

Scenario-Space Interaction (3) Scenario-Learner In-

teraction. The question arises how to analyse, to eva-

luate and to assess the reliability of the conceived le-

arning scenario at early stage in order to enhance lear-

ner’s experience with mobile environment. Based on

the quality aspects of Mobile Active learning research

previously outlined, we have developed an analytical

framework, named ReStart-Me, for M-learning based

on a combination of design issues, dimensions of le-

arning context and criteria. We consider that these fe-

atures, described below, should be considered in the

analysis process:

• Time Constraint: The time in the predictive ana-

lytical framework points out two aspects: the date

elements (begin and duration) and the learning

progress. Some of the learning resources are with

date-duration constraint, which means accessing

the learning contents and activities depends on

when they are available. For example, the con-

tents in a remote laboratory or museum would be

accessible only when these venues are open . On

the other hand, learner’s progress is considered as

a time sensitive factor because it can be used as a

constraint for providing up to date learning con-

tents to the learner.

• Space:The location in our analytical framework

indicates the learner’s current geographic posi-

tion. The global position system of the smart de-

vice (GPS) can be employed to sense the current

site of the learner. Therefore, space based lear-

ning scenarios can be implemented to enhance the

contextual interaction for learners. In this sense, a

signiﬁcant aspect of Space-mobility is quality of

service in terms of the reliability and speed of wi-

reless connections. Although some learning con-

tent can be downloaded to a mobile device and

used locally, the limitations on storage mean that

network connectivity is an essential component of

most mobile learning environments. The reliabi-

lity and speed of such connections can inﬂuence

which media types can be used in an M-learning

system, for example video streaming is only fea-

sible over a high speed connection.

• Device Constraint: It refers to the learner’s smart

device that is used to manage the learning process.

The device’s features are also considered as a con-

textual constraint for the mobile learning compa-

red with other computer assisted learning scena-

rios. In fact, smart devices are heterogeneous with

multiple operating platforms. They, also, have dif-

ferent and limited user- device interaction capabi-

lities (energy and battery consumption) (Tan and

Kinshuk, 2009) with certain mobile device ope-

rating systems and software platforms supporting

different types of display. A client software appli-

cation needs to run on the smart device to access

its native hardware and features and provides con-

textual information to the learning management

system. Therefore, it is essential to provide an

adapted format (text, interactive presentation, or

video) of learning contents to the smart device in

order to properly render the contents and to en-

hance the capability for the learner to interact with

the learning scenarios.

• Scenario Constraint: The learning scenario in-

clude learning objects, learning activities, and le-

aning instruction. The learning scenario can be

basic learning resources or smooth adaptive le-

arning activities stored in the learning contents

repository of the learning management system.

The learning scenario can be designed and re-

trieved based on pedagogical objectives and out-

comes. For instance, communication, collabora-

tion and interactivity issues can be affected by

more than one of the learning contexts commu-

nity, facilities, time and location activity and lear-

ner. Thus, different learning activities and its cor-

relation should be properly described and tagged

so that they could be easily designed and retrieved

by the adaptation mechanism.

• Learner: The learner is the main actor who

plays learning activities through smart device in

the contextual learning environment. The lear-

ner’s information contains static and dynamic data

mainly including the learner’s learning progress,

learning behaviours, and learning assessment re-

Predictive Analytical Framework based on Formal Method to Enhance Mobile and Pervasive Learning Experience

233

sults. The data is either manually entered into or

automatically collected from the learning mana-

gement system before or while the learner plays

his/her learning activities in the pervasive environ-

ment.

All these aspects are equally important, as the lear-

ning activity has to be simultaneously pedagogically

and technically sound.

4 ReStart-Me FRAMEWORK

4.1 Conceptual Framework

We have built a conceptual framework of learning en-

vironment analyse that structures different steps and

assists designers in this process. ReStart-Me (Re-

engineering of the Educational Scenario based on Ti-

med Automata and Tracks Treatment for Malleable

Learning Environments) is a modular architecture that

is based on two fundamental components :

• Trace is a tool-kit of post-evaluation process that

helps designers to build for each actor his ef-

fective learning scenario by organizing its events

and actions. Trace tool-kit incorporate two com-

ponents: the ﬁrst one ScanUp is a mobile applica-

tion that captures different states of device’s sen-

sor. The second one is program daemon that col-

lects the data from several databases and commits

it to a uniformed and structured ﬁle. In order to

do so, it needs (1) the Document type deﬁnition

of the destination ﬁle (in which to commit the raw

trace) (2) the Document type deﬁnition of the raw

data ﬁle that describes its structure and its diffe-

rent ﬁelds. This ﬁle is used by the data decoder

in order to reconstruct the learning scenario for

each actor as it was when the trace started. Trace

allows designers through a friendly graphical in-

terface: (1) to conﬁgure a trace session and alter

the retrieved events (2) to specify weather some

data or events (ﬁelds) are eventually omitted (3)

to set a range of IP address for data collection and

treatment.

• Evals is a graphical tool-kit through which the de-

signer can:(1) Create the formal model for each

activity (specify the title, learning objectives, pre-

requisites and outcomes of learning scenario); (2)

Model graphically the correlation between diffe-

rent activities(scenario structure). Thus designer

should specify temporal cartography and space

constraints; (3) Generate a contextualized Scena-

rio Model as shown in ﬁgure 2.

Figure 2: ReStart-Me interfaces.

4.2 Analysis Process

In this section, we present our proposed predictive

analytical process ReStart-Me. It is a conceptual fra-

mework based on timed automata modelling and for-

mal veriﬁcation. The evaluation process contains es-

sentially these three steps:

1. Step 1: Modelling M-learning scenario by se-

lecting activities’ template which are deﬁned by

Timed automata model expanded with global and

local clocks, in the ﬁrst step. By creating a for-

mal speciﬁcation, designers are guided to make

and to deﬁne a detailed learning scenario analy-

sis at early stage before its deployment into the

mobile and pervasive learning environment. In

the second step, designers had to specify temporal

correlation between different activities and to des-

cribe contextual constraints that reﬂect the beha-

viours of different entities in and with the learning

environment.

2. Step 2: Tracks simulation through which we can

observe all possible interactions between corre-

sponding to different entities involved in the le-

arning scenario. This step generates simulated

tracks that facilitate errors detection.

3. Step 3: Properties veriﬁcation with formal veri-

ﬁcation through which we check the reliability of

the designed learning scenario. This process aims

to build a remedial scenario and to help designer

in the re-engineering process by providing errors

and warning reports. In the scope of our study, we

tried to check the following properties:

• Deadlock is a situation in which two or more

active entities (for example students or groups)

are each waiting for the other to ﬁnish, and thus

neither ever does.

WEBIST 2018 - 14th International Conference on Web Information Systems and Technologies

234

• Live-lock is similar to a deadlock, except that

the states of the processes involved in the live-

lock constantly change with regard to one anot-

her, none progressing

• Starvation describes a situation where a sub-

group is unable to gain regular access to shared

resources (for example helps of the coach) and

is unable to make progress. This happens when

shared resources are made unavailable for long

periods by ”greedy” actors. We want to be sure

that each Subgroup will do his activity in time.

4. Step 4: Data collect and processing with Trace

and Scan-Up tools in order to enrich the abstract

model with different contextual value. This step

is deployed when the learning scenario prototype

is implemented.

This predictive analysis based on simulations helps

designers to assess an evaluate the reliability of their

designed learning scenario through the examination

of possible dynamic executions. Thus, it provides an

inexpensive mean of errors and fault detection that

avoids costly and time- consuming scenarios deploy-

ment and ensures quality in scenario engineering to

enhance learning experience.

4.3 From Framework to Reliability

Analysis Tool: Evals

Having presented the component model and the tailo-

ring platform, the question arises how to evaluate and

assess the reliability of the conceived learning scena-

rio at early stage in order to enhance learner’s expe-

rience with mobile environment based on the key di-

mensions previously deﬁned.

To go beyond the theoretical base and to operationa-

lize the framework in a form that designers can readily

use, we develop an analysis assistant tool ReStart-Me

based on the model checker Uppaal (Bengtsson et al.,

1996). To implement those extensions proposed we

used:

• Domain speciﬁc modelling Eclipse Tools (EMF

and GMF) for developing the graphical modeller;

• Template method pattern which is a behavioural

design pattern that’s let us deﬁne the skeleton of

activity’s model. At the same time, designers have

the possibilities to redeﬁne certain options wit-

hout changing its global behaviours. In order to

facilitate the modelling step, designers should:

– Select Activity, in the ﬁrst step. We have de-

ﬁned several model of activities: synchronous

activity (for example chat), asynchronous acti-

vity (such as discussing in forum)...

Figure 3: Template Method for learning scenario modelling.

Figure 4: From theory to application tool.

– Deﬁne temporal correlation intra-activity in the

second step and the inter-activities in the last

step of mlearning scenario modelling (see ﬁ-

gure 4).

Our Design tool, in a formal manner through friendly

graphical interfaces, allows pedagogical designers

and teachers: (a) to specify, model formally, gene-

rate and simulate different types of context-aware and

adaptive learning activities and their related contexts.

(b) to evaluate indoor and outdoor spaces within per-

vasive environment (c) to simulate and to verify inte-

ractions rules between Learner, Contextualized Acti-

vity and Space.

5 CASE STUDY

5.1 Learning Scenario Description

The conceptual designers’ meetings have produced a

scenario design document that describes the sequen-

cing of different activities and rules of the pervasive

learning scenario. We used pervasive learning scena-

rio models generated by the authoring tool ContAct-

Me (Malek and Laroussi, 2011). To summarize, the

learning scenario can be described as follows:

During a learning session of 4 hours, learners are in-

structed to carry out a proposed lesson that aims to

consolidate collaborative and competitive works by

enhancing interaction among them. To achieve this,

the scenario is divided into two phases as a serious

game. Each phase is composed of a set of activities to

Predictive Analytical Framework based on Formal Method to Enhance Mobile and Pervasive Learning Experience

235

Figure 5: Learners are working on Dinosaurs’ scenario.

be performed sequentially.

In order to boost intra-group competition, students

were divided in groups under the supervision of their

coach and each group consisted of two pupils. The

ultimate goal behind this clustering is to reinforce te-

amwork and collaboration within the individual sub-

groups and to make it a collaborative and challenging

game that takes place in different locations. Each sub-

group is equipped with a smart phone with a wireless

internet connection. In the ﬁrst phase, learners are

divided into small groups. Additionally, each group

is itself divided into two subgroups. The ﬁrst sub-

group collects information related to part of inquiry

in zones 1 and 2. This trial enables pupils to learn

through factual cases and to experiment various sce-

narios using social media and networks, pervasive and

mobile technologies.

The physical setting of this trial is the scientiﬁc cam-

pus where a mobile and pervasive learning environ-

ment is developed for guiding pupils in their tours.

This campus offers different learning activities in va-

rious disciplines: mathematical, ecological and histo-

rical activities. At the beginning of the ﬁrst stage, the

outdoor subgroup is localized by a localization sensor

and a notiﬁcation is sent to ask students to identify

and take a photo of the QR-code stuck to the dinosaur.

Instantly, a text adapted to the pupils’ level and pictu-

res that visualize and describe the activities to accom-

plish in the current stage is displayed on the screen

of the smart phone (see ﬁg. 5 ). After a pre-deﬁned

time, the subgroup will receive a stage-adapted quiz

via automatic text message. Pupils need to write an

answer using their smart phone and submit it. If the

answer submitted by the group is not correct, the sy-

stem sends an alert to the coach informing him/her

that pupils need some support. The coach should send

to them some hints.

In order to improve the coaching task, tutor decides

that after three wrong attempts, the pupil is guided to

start learning session by using his mobile device. The

e-learning client allow the student to directly mash up

widgets to create lesson structure and add powerful

online test widgets, communication widget (chat, fo-

rum and personal messages), content scheduling wid-

gets, communication tracking, announcements, con-

tent ﬂows, cooperative content building widgets.

The pupil could drag from the widget repository and

drops into the e-learning client UI all the widgets nee-

ded for providing video, audio and other multimedia

content. The duration of the learning session depends

on the intensity of the signal. Else, if the answer is

correct the indoor subgroups will receive the list of

activities of the second stage and will get joined by

their corresponding outdoor subgroups that will hand

over the picked plant samples.

5.2 Modelisation, Analysis and

Evaluation

Formal Modelisation. Fig. 6 provide an overview

of different automata modelled for the planned lear-

ning activities (outdoor activities, quiz, and lesson)

and corresponding to student. We deﬁne a global va-

riable ”clock named Time that gives idea about the

duration of each activity. The timing constraints asso-

ciated with locations are invariants. It gives a bound

on how long these locations can be active. We also

deﬁne other integer global variable Power to calcu-

late the energy of battery of the smart-phone.

In order to facilitate the learning scenario analysis, we

model each activity separately. The idea is to deﬁne

templates for activities that are instantiated to have

a simulation of the whole scenario. The motivation

for the use of templates is that the understanding, the

share and the reuse of different components of the le-

arning scenario become easier. The whole scenario

is modelled as a parallel composition of timed auto-

mata. An automaton may perform a transition separa-

tely or synchronise with another automaton (channel

synchronisation) or it can be activated after a period

of time through ﬂags. Fig. 6 also shows a timed au-

tomaton modelling the behaviour of the Smart-phone.

This device has several locations (or states):

1. idle: This initial state is activated when the smart-

phone is switched on.

2. InQuiz: The learner is responding to the quiz’s

questions.

3. Localisation: This state is activated when the le-

arner has to do his lesson. The smart-phone had

to be connected to the given wireless network.

4. LessonHigh, LessonMedium and LessonLow: One

of these three states is activated. The learner drags

and drops into the elearning client UI all the wid-

gets needed for providing video, audio and other

multimedia content. We should note that Lesson-

Low is undesirable state because technical studies

demonstrate that in this zone, student couldn’t do-

wnload correctly the lesson.

WEBIST 2018 - 14th International Conference on Web Information Systems and Technologies

236

Figure 6: The automata model of different activities for one actor (Student).

5. FinLesson: The student has ﬁnished the lesson

successfully.

6. BatteryOver: This undesirable State is activated

when the smartphone is switched off because the

battery state is over.

The dependability of the pervasive learning scenario

heavily depends on several contextual constraints, es-

pecially the quality of the network connection and the

energy provided by the mobile phone’s battery. Based

on studies introduced by (Perrucci et al., 2011), de-

signers elaborate a technical report that shows values

of power consumption of different activities that lear-

ners had to realize in the learning session. The power

consumption is quite higher when student download

lesson, and drop into the elearning client all widgets

needed for providing video, son and representation.

Simulation and Formal Veriﬁcation: Based on ti-

med automata presented above, tracks simulations

are generated visualizing all possible interactions be-

tween different actors. A screen dump of the simula-

tion of the designed educational scenario is below (see

Fig. 7). In order to help designers to improve their

educational scenario and to obtain better outcomes,

through the generated simulations, we try to localise

design errors, to answer and to verify the following

questions:

• Does the description of the learning scenario cle-

arly deﬁne time constraints for each activity and

the whole scenario?

• Is there any situation of deadlocks, liveness or

Figure 7: Simulation of the modelled Learning scenario.

Table 1: The simulation’s parameters.

Parameter Value

Duration of the simulation 240 units of time

Number of learners 40

Phone’s energy 100%

starvation?

We ﬁx different parameters of the learning scenario’s

simulation as shown in tab. 1.

A ﬁrst possible situation of deadlock is detected

within tracks simulation of the whole scenario; In

fact, students could have a period of inactivity es-

pecially when they begin their lesson in the lower

quality zone where the network connectivity is lo-

wer and smart-phone’s energy is limited. To check

this property, we are based on reachability property

that is considered as the simplest form of proper-

ties. We ask whether a given state formula, possi-

bly can be satisﬁed by any reachable state. Anot-

Predictive Analytical Framework based on Formal Method to Enhance Mobile and Pervasive Learning Experience

237

her way of stating this is: Does there exist a path

starting at the initial state, such that ”the battery

state” is eventually over along that path? We tra-

duce this property in temporal logic formula: ”E <>

Smartphone.BatteryOver and duration < 240” and

this property is satisﬁed. A second possible situation

of deadlock could happen when student starts lesson

in the Lower zone. To check this property, we are

again, based on reachability property: Does there ex-

ist a path starting at the initial state, such that stu-

dent couldn’t ﬁnish the learning scenario success-

fully?”. We traduce this property in temporal logic

formula: ”E <>Smartphone.LessonLower and du-

ration > 240” and this property is veriﬁed. This ex-

periment shows that 37% percent of students could

be blocked at this stage. They couldn’t progress in

the lesson because the state of energy is low. Then,

we conclude that dependability is not assured and we

had to adapt mobile application to the learning envi-

ronment’s context (if the student is located in Lower

quality zone, the mobile application had to reduce the

energy’s consumption by loading only the necessary

widgets. Also, we propose to notify student whet-

her the quality of signal doesn’t allow to start learning

session. Then, he should to move on.

6 CONCLUSION

ReStart-Me is a predictive analytical framework for

mobile and pervasive learning scenarios. It aims at

supporting pedagogical designers to assess reliability

of their scenarios at an early stage through automata

modelling, tracks simulations and properties veriﬁca-

tion. The proposed analysis process and supporting

tools provide an inexpensive mean of errors and fault

detection before starting implementation (thus retur-

ning on the learning scenario design phase to add mis-

sing elements and restraining development cost). Our

contributions are three-folds as explained below. Fir-

stly, we propose a framework to formally model and

evaluate the system design and the environment in-

puts. Important key dimensions of mobile and perva-

sive computing systems such as contextual activities,

actors and space interaction are discussed. Modelling

patterns for these features are provided and illustrated

with examples. Secondly, we identify critical proper-

ties of mobile and pervasive computing systems and

provide their speciﬁcation patterns in corresponding

logics. According to the stakeholders (designers, en-

gineers and users of these systems), reliability requi-

rements are essential to pervasive computing systems.

In our work, we classify the important requirements

into reachability properties. Furthermore, formal spe-

ciﬁcation patterns of these properties are proposed.

We verify formal properties against the system model

by using Uppaal model checker. Hence, design incon-

sistencies can be detected at the early design stage.

Thirdly, to demonstrate our approach, we present a

case study of applying the evaluation framework to

mobile and pervasive learning environment. Finally,

we will attempt in future works to deepen our pro-

posal on learning scenario re-engineering process in

such a way that a comparative report concerning dif-

ferent iterations of analysis can be generated automa-

tically.

REFERENCES

Bengtsson, J., Larsen, K. G., Larsson, F., Pettersson, P., and

Yi, W. (1996). UPPAAL a tool suite for automatic

veriﬁcation of real-time systems. Springer.

BenSassi, M., Laroussi, M., and BenGhezala, H. (2014).

Layered Evaluation Based on Timed Automata for

Educational Scenarios: Case Study on Blended Lear-

ning. The 6th International Conference on Computer

Supported Education,CSEDU.

Clark, R. C. and Mayer, R. E. (2016). E-learning and the

science of instruction: Proven guidelines for consu-

mers and designers of multimedia learning. John Wi-

ley & Sons.

Davidson-Shivers, G. V., Rasmussen, K. L., and Lowent-

hal, P. R. (2018). Web-Based Learning: Design, Im-

plementation and Evaluation. Springer International

Publishing, 2 edition.

Kuh, G. D. (1996). Guiding principles for creating seamless

learning environments for undergraduates. Journal of

college student development, 37(2):135–48.

Malek, J. and Laroussi, M. (2011). A model-driven appro-

ach for modelling and simulating new malleable lear-

ning scenarios for digital native generations. Interna-

tional Journal of Learning Technology, 6(3):288–319.

Moser, S. (2017). Linking Virtual and Real-life Environ-

ments: Scrutinizing Ubiquitous Learning Scenarios.

Digital Tools for Seamless Learning, pages 214–239.

Perrucci, G. P., Fitzek, F. H., and Widmer, J. (2011). Sur-

vey on energy consumption entities on the smartphone

platform. In IEEE 73rd Vehicular Technology Confe-

rence (VTC Spring), 2011, pages 1–6. IEEE.

Tan, Q. and Kinshuk, K. (2009). Client Mobile Software

Design Principles for Mobile Learning Systems. In-

ternational Journal of Interactive Mobile Technolo-

gies (iJIM), 3(1).

Westera, W. (2011). On the Changing Nature of Lear-

ning Context: Anticipating the Virtual Extensions

of the World. Educational Technology & Society,

14(2):201–212.

WEBIST 2018 - 14th International Conference on Web Information Systems and Technologies

238