On the Support of Multi-perspective Process Models Variability for

Smart Environments

Aitor Murguzur

, Xabier de Carlos

, Salvador Trujillo

and Goiuria Sagardui

Information Technologies Area, IK4-Ikerlan Research Center, Arrasate-Mondrag

on, Spain

Software and Web Engineering Group, Mondragon University, Arrasate-Mondrag

on, Spain

Keywords:

Variability Modeling, Process Conﬁguration, Smart Buildings.

Abstract:

Cloud service-based applications are to be adapted to serve multiple platforms and stakeholders. Atop of such

services, Smart Green Buildings are fostering a plethora of processes within their sustainability life-cycle. This

introduces a number of challenges, as how to support multiple perspectives of domain-speciﬁc variability and

how to deal with large collections of related process variants. To tackle this, there is a need to handle multi-

perspective variability for processes. This paper introduces an approach to manage multi-perspective process

variability by means of a meta-model and a modeling methodology, representing separately people and things

variability perspectives in smart environments. Initial experimental results are also described, which indicate

encouraging results for managing highly complex variability models.

1 INTRODUCTION

The world is getting more smarter in this age, i.e.,

more instrumented, interconnected and lead to better

decision making devices (e.g. the Nest

thermostat

or its Europe’s answer Tado

) currently are arising

and pointing out a number of challenges for smart en-

vironments (such as Smart Green Buildings (SGBs),

smart cities or smart healthcare). Central to this vi-

sion, agile service-based applications require to effec-

tively be deployed into a variety of devices and diffe-

rent implementation platforms (Gurgen et al., 2013).

Such services are globally dispersed under the con-

trol of different service providers and ready to be used

whenever needed. Atop such services, a number of

process models may exist which orchestrate domain-

speciﬁc activities related to aforementioned services

and deal with a huge data streams for analytics. In

such process-intensive environments, such as SGBs,

the management of domain-speciﬁc variability and in-

herent process models variability is becoming really

challenging (Castro et al., 2012).

In an SGB, several processes are deﬁned within

its sustainability governance life-cycle (i.e. installa-

tion and commissioning processes, conﬁguration pro-

cesses, operation processes and surveying processes)

http://nest.com

http://tado.com

to look at two major concerns, such as, improving

user experience and maintaining building sustainabi-

lity (Wang et al., 2012). For attainment of those pur-

poses, these all processes may be related to different

multiple dimensions - people and things respectively.

On the one hand, stakeholders (representing re-

lated roles and users) and associated operation con-

tracts, including a number of agreement policies

among stakeholders, (e.g. premium maintenance con-

tract) will need different process models, or simply

processes, to “coordinated (parallel and/or serial) set

of activities that are connected in order to achieve a

common goal”

in each life-cycle phase (e.g. energy

consumption of an SGB). This gives rise to a large

collection of process variants which could have co-

mmon and individual parts. Since living with this

complexity is not an agile practice, process variabi-

lity have to be managed from the people perspective.

On the other hand, a similar case is found in terms

of things. Beyond people perspective process variabi-

lity, a plethora of process variants could also be cre-

ated (e.g. energy efﬁciency of an SGB) for a diversity

of monitored assets (e.g. building facilities - ﬂoor,

room, etc., and equipment - chiller, boiler, etc.) and

building types (e.g. residential, commercial). There-

fore, the variability management becomes here also

the quid of things perspective complexity.

Workﬂow Management Coalition: http://wfmc.org

549

Murguzur A., De Carlos X., Trujillo S. and Sagardui G..

On the Support of Multi-perspective Process Models Variability for Smart Environments.

DOI: 10.5220/0004758205490554

In Proceedings of the 2nd International Conference on Model-Driven Engineering and Software Development (MODELSWARD-2014), pages 549-554

ISBN: 978-989-758-007-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

Regarding both perspectives, we have noticed

that existing initiatives concerning process variability

have not been adequate for supporting large process

variants collections in process-intensive smart envi-

ronments. This is due to they do not allow for multi-

perspective support, but also because there is a lack of

understanding and isolating both perspectives (people

and things) in process models variability.

In this paper, our aim is to support process varia-

bility research by dissecting multi-perspective and va-

riability modeling aspects. To this end, we contribute

(i) a meta-model of multi-perspective process variabi-

lity by a comprehensive analysis of multi-perspective

properties, and (ii) a modeling methodology for repre-

senting separately people and things variability pers-

pectives in smart environments.

The remainder of this paper is organized as fo-

llows: in Section 2, we summarize related research.

Section 3 details our proposed meta-model and me-

thodology for managing multi-perspective process va-

riability. The proof-of-concept prototype that we have

built is described in Section 4, and concluding re-

marks and future work towards real world implemen-

tation can be found in Section 5.

2 RELATED WORK

To best of our knowledge the work presented in

this article is the ﬁrst tackling the multi-perspective

and the impact on process variability management in

smart environments. On a more general level, this

work is related to other ﬁelds of research that can be

divided into two main blocks: (i) multi-perspective in

process variability and (ii) multi-perspective in Soft-

ware Product Line Engineering (SPLE).

2.1 Multi-perspective in Process

Variability

Process variability has been studied for almost a

decade including a number of approaches which uti-

lize different process variability realization tech-

niques to conﬁgure and adapt process variants, such

as single element-based process variant construction

technique (Schnieders and Puhlmann, 2006; Rosa

et al., 2011), which gathers all variability aspects in

a single conﬁgurable process model, representing the

whole process family (i.e. the collection of process

variants), and fragment-based re-use process vari-

ant construction technique (Hallerbach et al., 2010;

Baresi et al., 2012), which employs different mo-

dels for variability representation, modeling variabi-

lity separately from a base model. Although process

variability itself has been investigated, just a few stud-

ies have addressed the issue of multi-perspective.

For instance, the C-iEPC was presented as an

extension of the Conﬁgurable Event-driven Process

Chain (C-EPC) notation with the notions of roles (i.e.

for resource perspective) and objects (i.e. for data

perspective), enabling multi-perspective of business

processes (Rosa et al., 2011). Orthogonally, the multi-

perspective variants approach (Meerkamm, 2012) in-

volves multiple types of domain stakeholders in the

process conﬁguration, offering diverse levels of abs-

traction and reducing the number of decisions each

domain user has to make during the process mode-

ling and/or execution. Personalized views may also

be created in large process models by hiding and

abstracting irrelevant information for separated user

groups as well (Kolb and Reichert, 2013).

Consequently, we may recognize two different

meanings for multi-perspective in process variabi-

lity, as indicated in Table 1: the one related to pro-

cess elements, i.e., the partition of elements found

in process models themselves and the other to pro-

cess stakeholders, i.e., the individual requirements of

various users and groups for which particular pro-

cess variants exist. Given deﬁnitions, with respect

to the mentioned work, the main novelties of our

approach are two-fold: our approach clearly separates

domain-speciﬁc processes and domain-speciﬁc varia-

bility perspectives representation through a fragment-

based re-use approach. Secondly, we are not just fo-

cusing on stakeholders, otherwise we consider both

Table 1: Multi-perspectives in process variability.

Variability technique/Multi-perspective Process elements Process stakeholders

Single element-based construction (Rosa et al., 2011) (Meerkamm, 2012)(Kolb and Reichert, 2013)

Fragment-based re-use construction (Our approach)

Table 2: Multi-perspective in SPLE.

Variability modeling/Multi-perspective Product-line stakeholders

Single variability model (Czarnecki et al., 2004; Classen et al., 2009; Mendonc¸a et al., 2008)

(Garc

ıa-Gal

an et al., 2013) (Our approach)

Multiple variability models (Rosenm

uller et al., 2011; Schroeter et al., 2012) (Our approach)

MODELSWARD2014-InternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

550

people and things perspectives enabling full-support

for multi-perspective variability.

2.2 Multi-perspective in SPLE

In the SPLE, several approaches have also been pro-

posed as a means to facilitate conﬁguration of the

software product line by applying disparate variabi-

lity modeling techniques to deal with multiple dimen-

sions and conﬁgurations of a product-line.

Essentially, the variability modeling represents

the commonality and variability of a family of sys-

tems in the domain and relationships between them

by employing a particular variability modeling lan-

guage (e.g. feature models, decision models, or

the OMG initiative Common Variability Language

(CVL)

(Berger et al., 2013). Such variability is of-

ten described with a feature model using, for ins-

tance, (i) a single variability model to handle all do-

main variability and leverage multi-dimensional se-

paration of concerns in feature-based conﬁgurations

(Hubaux et al., 2011), or (ii) multiple variability mo-

dels to separate multiple dimensions such as, domain

variability and implementation variability. Although

the former is about visibility, the latter is easier to han-

dle than a large single model which could avoid high

complex conﬁgurations and thus enhance manageabi-

lity (Rosenm

uller et al., 2011; Schroeter et al., 2012).

Most of the variability conﬁgurations techniques

are concerned with multi-user support for a particu-

lar variability modeling language. Feature models

staged-conﬁguration was proposed with the aim of

providing step-by-step conﬁgurations in which each

user reﬁnes previous user decisions (Czarnecki et al.,

2004; Classen et al., 2009). Multi-user collabo-

ration product-line conﬁguration improves staged-

conﬁguration by allowing users to make decisions in

a concrete space (i.e. limiting the space for each user)

at the same time without conﬂicts (Mendonc¸a et al.,

2008). General approaches, without space limitation

and enabling user satisfaction, have also been studied

in the literature (Garc

ıa-Gal

an et al., 2013). There-

fore, multi-perspective in SPLE is related to multiple

stakeholders’ conﬁgurations support, providing them

more accurate views.

Bearing in mind the aforementioned work (see

Table 2), our approach differs from traditional

approaches adopting a single or multiple models for

tackling multi-perspective conﬁgurations in process

variability. We provide an approach by means of a

meta-model and three modeling strategies for sup-

porting multi-perspective in process variability.

http://variabilitymodeling.org

3 MULTI-PERSPECTIVE IN

PROCESS VARIABILITY

3.1 Formalizing Multi-perspective

Process Variability

As it was remarked, the succeeding conceptualiza-

tion pursues a fragment-based re-use variant cons-

truction approach which employs different models

for multi-perspective process variability representa-

tion in smart environments (e.g. SGBs), modeling va-

riability separately from a base model. Especially for

that purpose, we make use of a slightly altered Base-

Variation-Resolution (BVR) method (including addi-

tional concepts as described below) from the Soft-

ware Product Line Engineering (SPLE) (Bayer et al.,

2006) to deﬁne a total of three models for supporting

multi-perspective process models variability: a base

model (BM) - representing commonalities, a varia-

tion model (VM) - representing individualities, and a

resolution model (RM) - representing process variant

conﬁgurations, as illustrated in Figure 1.

3.1.1 Base Model

The base model describes, employing a business pro-

cess modeling language (e.g. Business Process Model

and Notation (BPMN), Business Process Execution

Language (BPEL)), the commonalities shared by all

process variants and placeholder activities that are

subjected to vary. Placeholder activities (i.e. varia-

tion points) identify concrete parts in a BM in which

variant binding occurs. A base model element stands

for any kind of model asset in a BM (i.e. a common

or placeholder activity). This relationship has a zero-

to-one cardinality, since not all model elements (e.g.

common activities among related process models) are

affected with variability (i.e. some activities are not

placed as variation points).

A variation point represents a conﬁgurable acti-

vity element, in which a conﬁguration may be se-

lected, after which it is no longer possible to change

the selection. This selection is called binding, and

the stage in the life-cycle at which binding occurs is

called binding time. We distinguish three kind of vari-

ation points - static, partial and dynamic:

• Static is resolved at build-time, resolving variation

points by selecting suitable alternatives from a set

of available ones. Therefore, process variants are

fully speciﬁed at conﬁguration-time and deployed

into a process engine for later enactment.

• Partial extracts a pattern with a small number of

fragments, simplifying the spectrum of fragment

OntheSupportofMulti-perspectiveProcessModelsVariabilityforSmartEnvironments

551

Figure 1: The Ecore meta-model for multi-perspective process variability.

choices for runtime, helpful when a certain thre-

shold of variation points or fragments are reached.

• Dynamic is bound at runtime since we cannot

pre-speciﬁed which of the applicable fragments

will be suitable considering monitoring dynamic

context data (e.g. ﬂoor internal temperature),

as in Dynamic Software Product Lines (DSPLs)

(Baresi et al., 2012).

3.1.2 Variation Model

The variation model gathers all the individualities in-

troduced by each process model variant characteri-

zing individualities and constraints, i.e., relationships

among particular domain features and variables.

The variability perspective describes a pre-

conﬁguration of a reference BM, i.e., it takes the

aggregation of a particular variability perspective (e.g.

one stakeholder view) under consideration. Beyond

people’s perspective, process variability could also

be resolved interpreting context data from embedded

and smart devices (e.g. Heating, Ventilation and Air

Conditioning (HVAC), etc.). Hence, two variability

perspectives are interwoven - people perspective and

things perspective respectively, see Figure 1:

• People Perspective: Following the terminology

of (Schroeter et al., 2012) two main items are out-

lined regarding people perspective. On one hand,

a view shows only process variants that could be

conﬁgured for a speciﬁc stakeholder or role. Ac-

cording to the ISO/IEC/IEEE 42010:2011, Sys-

tems and software engineering standard

, “a view

is a representation of the whole system from the

perspective of a related set of concerns”. On the

other hand, a viewpoint represents a set of views,

each restricting a reference BM to a subset of con-

ﬁguration, e.g. a particular operation contract re-

levant for a group of stakeholders.

http://www.iso-architecture.org/ieee-1471/

• Things Perspective: In terms of things, process

variants could also be created for the same pur-

pose (e.g. energy consumption or chiller opti-

mization) but considering resource types and asset

speciﬁcations. A resource type describes each

different source of information from which con-

text data is gathered (e.g. residential, commercial

buildings). Similarly, an asset represents some-

thing that is built, installed, or established to serve

a particular purpose which might include a num-

ber of facilities and monitored equipment to per-

form different actions.

The variability speciﬁcation represents the actual

variability of a variation point which is directly re-

lated to a variability perspective element (i.e. view,

viewpoint, resource type or asset speciﬁcation). Just

like in domain feature modeling, it stands for a family

of systems in the domain and relationships between

them (i.e. mandatory, optional, alternative, and or re-

lations), their corresponding domain and constraints.

Here, we identify three different types of elements

which are fundamental in our approach:

• Variability Constraints represent constrains on

valid resolutions (i.e. requires, excludes and

cross-tree constraints). Constraint sub-model

could include constrains on valid process vari-

ant conﬁgurations that may be deﬁned at three

levels: complex feature-feature, feature-variable

and variable-variable cross-tree relationships.

• Fragment describes a single realization option for

each variation point placed in the BM speciﬁca-

tion. Depending on variability speciﬁcation in-

cluding people and things perspectives, and con-

sidering context data, applicable fragments are se-

lected and executed to resolve multi-perspective

process models variability.

• Transformers have concrete transformations asso-

ciated to a variability speciﬁcation. When con-

text values are bound to a concrete transformer,

MODELSWARD2014-InternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

552

it executes the corresponding resolution. In other

words, it is the responsible for deciding which re-

solutions should be adopted by the current data

values (value, fragment or perspective).

3.1.3 Resolution Model

A resolution model represents a binding of variabi-

lity speciﬁcations, which may be used to derive a new

process variant considering context data. It speciﬁes

which alternatives described in the VM are valid for

settled placeholder activities in the BM considering

current conditions. To that end, a set of resolution

elements deﬁne how a VM is bound. A resolution

element represents a binding of variability. It could

either do a complete binding in which all variability

from a variability speciﬁcation is resolved, or a partial

resolution in which some variability is still present. It

has three types of subtypes:

• Value resolution determines a value for the varia-

bility where context values are mapped to vari-

ables’ or features’ resolutions in the variability

speciﬁcation.

• Fragment resolution represents resolutions for

variation points in the variability speciﬁcation as

well as the corresponding mappings in the BM.

• Perspective resolution considers context data re-

lated to people and/or things perspective resolu-

tion. It is used to minimize the process variant

conﬁguration options reducing alternatives. Once

the multi-perspective variability speciﬁcation is

fully described, our system starts to collect con-

text data in order to start a perspective variabi-

lity resolution. This separation of concerns allows

to perform different strategies for and automated

staged-conﬁguration within our toolkit - people-

ﬁrst, things-ﬁrst and in-row.

– People-first: an initial pre-conﬁguration on

people perspective variability is performed, re-

ducing process variants alternatives for a latter

things perspective resolution.

– Things-first: In contrast, this strategy en-

ables to be things related data ﬁrst recognized

in order to perform a staged-conﬁguration of

process variability.

– In-row: The perspective resolution is just

done by interpreting context data as it is col-

lected to resolve multi-perspective process va-

riability.

To perform the aforementioned types of resoluti-

ons, we classify two types of context data: static con-

text data variable-value pairs (e.g. temperature sensor

type or stakeholder) representing static preferences of

a particular domain that are known just before the pro-

cess instantiation and which rarely change over time,

and dynamic context data pairs (e.g. speciﬁc temper-

ature of a room) which change frequently over time

and thus only available when the selection of a place-

holder activity occurs. Once a set of context pairs is

fulﬁlled (i.e. assigning a value to context variables

from context data), value, fragment and perspective

resolutions are executed.

3.2 Modeling Multi-perspective Process

Variability

One of the advantage of separating commonality and

variability speciﬁcation is that there may be more than

one VM for each BM, as illustrated in Figure 1. Pro-

cess modelers may associate as many variation mo-

dels as they require to an unique BM even represen-

ting multiple perspectives. Ideally, each smart en-

vironment solution (e.g. a concrete SGB solution)

would require a separate VM, so we will build a diffe-

rent VM for each domain solution (specifying all par-

ticularities by multiple perspectives support) where a

BM is applied. Therefore, this set of variation models

will represent different variability scenarios at diffe-

rent levels regarding a common shared BM.

The cornerstone of our approach is multi-

perspective support, which may be modeled in a sin-

gle VM or also taking advantage of separated models

for each perspective description. However, from our

previous work (Murguzur et al., 2013), we have no-

ticed that to avoid highly complex models, we should

try to separate different variability dimensions such

as people variability and things variability. Thus, our

approach now enables the process variability mode-

ling of people perspective (PP) and things perspective

(TP) in disjointed models as follows:

• In-PP: people variability perspective may only

contain variation points.

• In-TP: things variability perspective may only

contain variation points.

• SharedBy: both variability perspectives may

contain elements directly related to variation

points.

4 PROTOTYPE AND TOOLKIT

To implement our multi-perspective process variabi-

lity we utilize Activiti Designer

for base model and

fragments speciﬁcation and Clafer

(CLAss, FEature,

http://activiti.org

http://clafer.org

OntheSupportofMulti-perspectiveProcessModelsVariabilityforSmartEnvironments

553

Reference) for modeling people and things perspec-

tive variability. The Activiti Designer is an open-

source and light-weight Business Process Manage-

ment (BPM) platform which offers among other fea-

tures a powerful designer and a solid BPMN2 pro-

cess engine. Clafer is also a lightweight modeling

language which enables to express variability among

products and systems.

Following the presented approach, we model the

variability of an elevators remote predictive main-

tenance and monitoring process in an SGB domain

using two separated variation models and techniques

(In-PP, In-TP and SharedBy): one for people pers-

pective (e.g. maintenance contracts) and the other for

things perspective (e.g. elevator’s modules - door and

pulley) variability modeling.

Overall, our prototype shows encouraging results

to cope with complex large models, by adopting base

models, variation models (representing things and

people perspective separately) and process fragments

for representing multi-perspective process variability.

However, it is restricted in the sense that it purely fo-

cuses on the control-ﬂow.

5 CONCLUSIONS

Likewise in SPLE, process variability needs to con-

sider the multiple dimensions coming from smart en-

vironments, i.e., it needs to support the multiplicity of

stakeholders, devices and data derived from process-

intensive applications. This premise is the underly-

ing idea of our paper that presents a novel approach

for supporting multi-perspective process variability

by using a fragment-based re-use approach. After dis-

cussing the impact of multiple dimensions in SGBs

with respect to multi-perspective process variability,

we pointed out the related work and also described

a methodology by means of a meta-model for sup-

porting multi-perspective process variability adopting

a fragment-based re-use approach.

In our future work we plan to consider an au-

tomated resolution of multi-perspective process va-

riability by processing over context data streams, as

well as evaluating the presented perspective resolu-

tion strategies against different modeling methods.

REFERENCES

Baresi, L., Guinea, S., and Pasquale, L. (2012). Service-

Oriented Dynamic Software Product Lines. IEEE

Detailed information at:

http://aitormurguzur.com/projects/lateva/multi-perspective

Computer, 45(10):42–48.

Bayer, J., G

erard, S., Haugen, O., Mansell, J., Moller-

Pedersen, B., Oldevik, J., Tessier, P., Thibault, J., and

Widen, T. (2006). Consolidated product line variabi-

lity modeling. In SPLC.

Berger, T., Rublack, R., Nair, D., Atlee, J. M., Becker, M.,

Czarnecki, K., and Wasowski, A. (2013). A Survey of

Variability Modeling in Industrial Practice. In VaMoS.

Castro, M., Jara, A. J., and Skarmeta, A. F. (2012). An anal-

ysis of m2m platforms: Challenges and opportunities

for the internet of things. In IMIS.

Classen, A., Hubaux, A., and Heymans, P. (2009). A for-

mal semantics for multi-level staged conﬁguration. In

VaMoS.

Czarnecki, K., Helsen, S., and Eisenecker, U. (2004).

Staged conﬁguration using feature models. In SPLC.

Garc

ıa-Gal

an, J., Trinidad, P., and Ruiz-Cort

es, A. (2013).

Multi-user variability conﬁguration: A game theoretic

approach. In ASE.

Gurgen, L., Gunalp, O., Benazzouz, Y., and Gallissot, M.

(2013). Self-aware cyber-physical systems and appli-

cations in smart buildings and cities. In DATE.

Hallerbach, A., Bauer, T., and Reichert, M. (2010). Cap-

turing Variability in Business Process Models : The

Provop Approach. Software Maintenance and Evolu-

tion, 22(6-7):519–546.

Hubaux, A., Heymans, P., Schobbens, P.-Y., Deridder, D.,

and Abbasi, E. (2011). Supporting multiple perspecti-

ves in feature-based conﬁguration. Software and Sys-

tems Modeling, pages 1–23.

Kolb, J. and Reichert, M. (2013). A ﬂexible approach for

abstracting and personalizing large business process

models. Applied Computing Review, 13(1):6–17.

Meerkamm, S. (2012). Staged conﬁguration of multi-

perspectives variants based on a generic data model.

In BPM Workshops.

Mendonc¸a, M., Cowan, D., Malyk, W., and Oliveira, T.

(2008). Collaborative Product Conﬁguration: For-

malization and Efﬁcient Algorithms for Dependency

Analysis. Journal of Software, 3(2):69–82.

Murguzur, A., Sagardui, G., Intxausti, K., and Trujillo, S.

(2013). Process variability through automated late se-

lection of fragments. In CAiSE Workshops.

Rosa, M. L., Dumas, M., ter Hofstede, A. H., and

Mendling, J. (2011). Conﬁgurable multi-perspective

business process models. Information Systems :

Databases : Their Creation, Management and Utiliza-

tion, 36(2):313–340.

Rosenm

uller, M., Siegmund, N., Th

um, T., and Saake, G.

(2011). Multi-dimensional variability modeling. In

VaMoS.

Schnieders, A. and Puhlmann, F. (2006). Variability mech-

anisms in e-business process families. In BIS.

Schroeter, J., Lochau, M., and Winkelmann, T. (2012).

Multi-perspectives on feature models. In MODELS.

Wang, K.-Y., Lin, G., Chou, P., and Chou, A. (2012). Lever-

age smart system services technology for smart green

building management. In EGG.

MODELSWARD2014-InternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

554