A Domain Speciﬁc Platform for Engineering Well Founded Measurement

Applications

Florent Bourgeois

1,2

, Philippe Studer

, Bernard Thirion

and Jean-Marc Perronne

Universit

e de Haute Alsace, MIPS EA 2332, 12 rue des Fr

eres Lumi

ere, F-68093 Mulhouse, France

Actimage GmbH, Hafenstraße 3, 77694 Kehl, Germany

Keywords:

Metrology, Measuring Process, Mobile Measurement Applications, Development Platform.

Abstract:

Mobile platforms, such as smartphones, are now embedding more processing and communication capabilities

than ever. They offer generally a set of standard built-in sensors to measure their surroundings and potentially

increase their knowledge about the environment. Moreover their communication capabilities allow easy access

to external devices and remotely accessible sensing nodes or more general services. Nevertheless, despite their

obvious ability to provide rich data visualization, only a few applications propose using mobile platforms as

a ﬂexible and user-friendly measuring process assistant. This paper proposes the description of a system able

to model and design mobile and well-founded domain speciﬁc measuring processes, supporting physical as

well as non-physical quantities. The soundness of the application and its conformance to metrology rules is

ensured through the use of quantities semantic, dimensional analysis and adherence to the representational

theory of measurement. The conformance veriﬁcation gives to non-metrology specialists the ability to design

and conﬁgure rigorous mobile applications dedicated to assist an end-user in its usual and speciﬁc measuring

needs and habits, while limiting erroneous results due to manipulation errors.

1 INTRODUCTION

In recent years, we have seen the global widespread

use of powerful sensors embedded into smartphones

(Daponte et al., 2013). Sensors increase our capabil-

ity to measure things, which is an unbelievable oppor-

tunity for craftsmen who need to know their work sur-

roundings. Moreover, sensors are useful tools in ev-

eryday activities such as cooking, health care, sport or

interior design. This leads to the recurring use of mea-

suring operations and vocabulary. They are so useful

and widespread in popular culture that anyone is able

to deﬁne a way to measure a table length, the weight

of a person or the temperature of a room. Further-

more, people naturally use units in order to commu-

nicate obtained values.

Nevertheless, accidents such as the loss of the

Mars Climate Orbiter (Stephenson et al., 1999) re-

minds us that errors in measurements can lead to

disastrous situations and are quite difﬁcult to detect.

Even though this example is critical, it shows that

even experts are subject to make mistakes, which

means that non-specialists are, a fortiori, likely to ob-

tain wrong results. Such wrong results can waste time

and affect production for craftsmen which may have

important economic repercussions.

The newly built-in mobile sensors, such as ac-

celerometers, gyroscope, barometer, compass, GPS,

etc. are not always straightly compatible with users’

needs. Indeed, they present values that were not ac-

cessible to users a few years ago. For example, if a

user wants to know about the tilt of a surface, he might

try to use his mobile gyroscope. But this requires pre-

cise knowledge of what the sensor returned values are

and how to evaluate a tilt indication from these values.

This kind of problem can be alleviated by using a do-

mains speciﬁc measuring assistant. As an illustration,

a typical application is described.

The application proposes to the user to estimate

the production of a solar panels installation using lo-

calization data to deﬁne the positions of the panels.

The measuring process requests the user to place the

mobile device on the roof. Using the embedded sen-

sors, the application gathers data on the geolocation,

the orientation and the inclination of the device. With

such data the application knows the solar panels the-

oretical position. Then the application is able to esti-

mate quantities such as the weather condition of the

location and mean daylight time. If the user then in-

puts the surface and the kind of solar panels, it is pos-

309

Bourgeois F., Studer P., Thirion B. and Perronne J..

A Domain Speciﬁc Platform for Engineering Well Founded Measurement Applications.

DOI: 10.5220/0005512103090318

In Proceedings of the 10th International Conference on Software Engineering and Applications (ICSOFT-EA-2015), pages 309-318

ISBN: 978-989-758-114-4

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

sible estimate their production of electricity.

It might also be interesting to extend this scenario

by adding the ability to consider the roof nature, the

weight of the solar panels, the kind of roof covering

products. Such data might be interesting to estimate if

the roof can bear the installation. Also, most of solar

panels owners are looking for electricity cost savings.

It is then possible to estimate the money saved thanks

to the installation. Furthermore since the installation

comes with a price, it would be possible to estimate

the time required to cover the installation price with

savings.

This application makes use of physical and non

physical quantities altogether in order to produce its

result. These quantities belong to different measure-

ment scales (c.f. Section 2). The complexity of the

calculation is not unbearable, but it is subject to mis-

takes and wrong calculations when using such differ-

ent quantity kinds.

Metrology, the science of measurement, deﬁnes

rules to follow during measuring. Thanks to these

rules it is possible to avoid erroneous results due to

manipulation errors. Basics of this science are known

by everyone but advanced rules are only accessible

to specialists. Such a situation is not satisfactory con-

sidering the growing number of measuring tools users

can access and the risk of increased measurement er-

rors. The mentioned errors encompass the absence

of references leading to conversions errors, the lack

of knowledge in the sensors characteristics, process

steps omission, mishandled operations on quantities

and false decisions based on poor quality results.

Based on an interdisciplinary approach combining

quantities semantic, dimensional analysis and repre-

sentative theory of measurement, as well as software

modeling, software engineering and software archi-

tectures, this paper proposes a model-driven approach

for building well-founded mobile measurement ap-

plications. First an architect has to model a domain

speciﬁc measuring process. The obtained models are

then validated for coherence with real world accord-

ing to the process inputs semantic speciﬁed by the em-

ployed sensor models. Following the use of model-

driven principles and tools allows for creating an ap-

plication that conforms to the validated models. The

resulting application supports the end-user when he

tries to apply the measuring process while ensuring

more conﬁdence in the produced results.

This paper presents related works by presenting

the metrology rules deduced from the literature and

the various development existing in the domain. Af-

terward, it shows the requirement analysis which lead

then to our design and prototype. Last, the paper con-

cludes by presenting future developments.

2 RELATED WORK

2.1 Metrology: Foundations, Standards

Metrology is an active research domain covering all

of the sciences. Several foundational frameworks ex-

ist to model the domain. Some based on mathemat-

ics such as (Khan and Finkelstein, 2013), others on

algebra like (Domotor and Batitsky, 2010) or some

on the Representational Theory of Measurement in-

cluding (Mari, 2013). Within this section we present

the standard of metrology used to represent observa-

tions. Then we point out that quantities are not only

numbers and present the importance of constraints on

their manipulations. And last we explain the terms of

measuring process and uncertainty.

2.1.1 Measurement Standards

Metrology is the science used to communicate about

the surrounding world in terms of measurable objects

properties. Being multidisciplinary, it is used in any

physical science (i.e. mechanics, physics) but also in

psychological sciences (i.e. IQ, pain). International

standards have been created lately to ease this pro-

cess. The Joint Committee for Guides in Metrology

(JCGM) published a uniﬁed vocabulary in (OIML

et al., 2012), the International Vocabulary of Metrol-

ogy (VIM), which deﬁnes itself as a common ref-

erence to anyone looking to perform measurements.

The VIM deﬁnes a quantity as the property of a phe-

nomenon, body, or substance, where the property has

a magnitude and a reference. The most usual kind of

reference is a measurement unit. Quantities are clas-

siﬁed by quantity kinds.

Physical properties, represented by quantity kinds

(e.g. length, time, speed), are associated one to

another with physical laws. Hence the quantity kinds

are organized using systems of quantities based on a

set of base quantities and derived quantities, which

result from algebraically combining quantities. The

most recognized system is the International System

of Quantities (ISQ) described by (ISO, 2009). This

system presents the advantage that any quantity kind

can be represented using a combination of seven base

quantities through multiplication and division oper-

ations. The SI has been created upon this standard.

(BIPM, 2006) describes the International System of

units (SI) which presents one speciﬁc unit for each

quantity kind. Other quantities are represented as a

combination of these base units. Also some of the

derived units are called by a special name, whose pur-

pose is to ease the notation of complex derived units.

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

310

The reader is referred to the NIST diagrams

for an overview of the ISQ and SI. Ultimately only

SI units should be used to communicate but non-SI

units are still used widely. Therefore, even though it

is possible to convert from one unit to another, this

can lead to misunderstanding when quantities are

communicated. (Astarita, 1997) shows that using a

system of quantities allows us to use dimensional

analysis. This method is able to check quantity

equation homogeneity, to ﬁnd equation simpliﬁcation

and also to determine the appropriate scaling of a

quantity.

2.1.2 Representational Theory of Measurement

Nowadays it is natural to use quantity values associ-

ated with SI units to describe objects. This quanti-

tative representation is so widespread that the non-

expert tends to think that he simply is manipulat-

ing numbers. However, on a more foundational per-

spective, the representational theory of measurement

(RTM), initiated by Stevens (Stevens, 1946), explains

the relations between the empirical observation of

events and their formal representation and manipu-

lation as quantities. (Krantz et al., 1971) explains

how to extract from the different empirical relations

on object properties a system of operations that can

be safely applied on the formal representation and are

still relevant in the real world.

Since several quantity kinds have the same mathe-

matical structure, the notion of scales is introduced. A

scale represents a given set of permissible operations

for a quantity as illustrated in table 1. (Hand, 1996)

denotes that the scale attributed to a given observa-

tion result depends on the knowledge of the property

analyzed and the value assignment method. Such a

knowledge enables us to detect when common opera-

tions cannot be applied (e.g. mean of values).

It is important to note that RTM is not only re-

served for the measurement of physical quantities,

but also for more subjective measurements of quan-

tities, such as those performed in psychosocial sci-

ences. This implies quantities that are not only repre-

sented by numbers but also by symbols because of the

possibility to use nominal scales in order to categorize

various properties (i.e. color, gender).

2.1.3 Measuring Process and Uncertainty

Although several processes can deﬁne a given object

property, some are more convenient than others and

can have an impact on uncertainties.

http://physics.nist.gov/cuu/Units/SIdiagram2.html

The deﬁnition of a domain speciﬁc measuring pro-

cess is a difﬁcult task. Indeed, the process has to de-

ﬁne any interacting parts to obtain the result. (Ellison

et al., 2000) explains that it is possible to ease this task

using a cause and effect diagram with input branches

for the method, the operator, the instruments, the en-

vironment and the subject under consideration.

Deﬁning the right measuring process also implies

an uncertainty estimation process. This uncertainty

results from instruments precision limitations and the

combination of the different errors induced by the

measurement process (Farrance and Frenkel, 2012).

The Uncertainty of Measurement Guide (OIML et al.,

2008), or GUM, is the international reference con-

cerning uncertainty estimation. It presents the vocab-

ulary and the methods for handling uncertainty.

2.2 Metrology in Computer Science

In this section we present an overview of the develop-

ments that integrate metrology and computer science.

2.2.1 Representing Observations

The encoding of metrological concepts in software

occurs at different conceptual levels : ontological,

modeling, units or dimensions encoding, etc. (Foster,

2013) discusses the lack of a standard formalization

of the concepts, which brought an environment with

no cohesion, every system using measurements being

based on its own implementation or encoding.

(Fowler, 1997) speciﬁes a model for observation

and measurement. This model proposes considering

categories which are deﬁned as a range of values for

a given quantity. An observation holds a protocol

to perform the measurement. Hence with the cate-

gories and the protocol it is possible to model any

measurement of physical and non physical quantities.

Derived from this model, the Open Geospatial Con-

sortium proposed (ISO, 2011), which describes Ob-

servations and Measurements (O&M). This is a stan-

dard for modeling quantities in geographic informa-

tion systems. It is able to represent measurements of

any domain as long as the process description is pro-

vided. Physical quantities are represented by a num-

ber and a unit of measurement encoded as identiﬁers

recognized by the system. An XML schema of O&M

is provided by the Open Geospatial Consortium.

On a more ontological level, quantities represent

knowledge about a physical object. This is made

through the use of a standardized vocabulary, which

ontologies are used for. So it is natural to ﬁnd on-

tologies that represent quantities and units. (Hodgson

et al., 2013) presents QUDT ontologies which rep-

resent any system of quantities associated with its di-

ADomainSpecificPlatformforEngineeringWellFoundedMeasurementApplications

311

Table 1: Representation of the four basic measurement scales (inspired from (Stevens, 1946)).

Scale Elements relation Acceptable Operations Example

Nominal Categorized = (is a) Gender, Colors

Ordinal Ordered =, < IQ, school grades

Interval Equidistant magnitude =, <, + Temperatures (

◦

Ratio Equidistant magnitude and Absolute zero =, <, +, ∗ Weight, Length

mensions and system of units. While QUDT proposes

a base reusable in any software to handle units and

quantities, it does not resolve issues such as the units

presenting the same dimension. Such issues appear

typically with mechanical torque and energy having

the same dimension but different units and also with

dimensionless units as angles and counts.

(Rijgersberg et al., 2011) and (Rijgersberg, 2013)

present OM, which is “an Ontology of units of Mea-

sure and related concepts”. They also compare OM to

related existing ontologies. It appears that OM tends

to propose more capabilities than other ontologies

thanks to its ability to operate dimensional analysis.

Indeed, to propose a dimensional analysis that is more

accurate and avoids conﬂicts, OM proposes deﬁning

applications domains for quantities and units. Know-

ing the domain of an end-user, it is possible to reduce

unit conﬂicts. However and according to (Rijgers-

berg, 2013) their interpretation of the term scale is

different from that of RTM. Thus it would not be pos-

sible to represent non physical measurements. OM

proposes web services whose purposes are to share

measurements data in the semantic web.

The Uniﬁed Code For Units of Measure (UCUM),

presented by (Schadow and McDonald, 2009), is a

coding system intended to include all units of mea-

sures being contemporarily used. The document

presents a language capable of expressing any unit.

Preﬁxed units can be created using units and preﬁxes.

Also, UCUM has been able to deﬁne a language with

no conﬂict as long as the non-SI units do not use SI

preﬁxes. Consequently it is a simple and efﬁcient way

to represent quantities as a structure composed of a

number and a UCUM coded unit.

(Cmelik and Gehani, 1988) describe a way to use

preprocessing in order to write source code embeding

quantities and units. The direct usage of such entities

in the source code makes comprehension easier when

modeling physical properties and recall that the data

manipulated are more than simple numbers. One unit

is deﬁned as the base unit for each quantity. In order

to accept different units, it is possible to deﬁne a new

one by selecting the unit it is based on and a conver-

sion factor to this unit. Thanks to this operation the

user can code by using the usual units and the pre-

processor will do the conversion. Libraries based on

this idea such as (Schnabel and Watanebe, 2013) are

still used. While useful, such method has no impact

beyond the code level.

2.2.2 Quantities Processing

Most implementations essentially propose encoding

the observations. We now focus on systems able to

manipulate the quantities.

(Rijgersberg, 2013) shows that OM is able to han-

dle dimensional analysis. When an equation based on

quantities is proposed, a web service is able to de-

termine if the equation is dimensionally correct and

if the units used are coherent. Their framework pro-

poses an abstraction of mathematical operations and a

workﬂow that has to be processed to achieve the cal-

culation. Indeed, in calculations performed by mathe-

matical tools such as MATLAB, the semantic data has

to be removed and treated separately from the numer-

ical data in order to generate separate numerical and

semantic results that are then combined. The units of

the results are assigned using dimensional analysis.

(Botts and Robin, 2007), who use O&M as a basis

to express quantities, is a framework dedicated to the

representation of processes implying the transforma-

tion of input stimuli to output measurement quanti-

ties. Initially aimed at processing sensors data, O&M

is also able to represent operations on quantities. In-

deed it is possible to consider such operations as pro-

cesses which stimuli and output are quantities. Yet,

the framework is not able to deduce the resulting

quantity kind from the input quantities or to determine

whether an operation is possible. Indeed the O&M

framework does not provide implementation for the

different features of interest or for the units. So its ca-

pabilities of mathematical operations are bound to the

user knowledge and the implementations used.

The same limitations are found in most of the

tools. Indeed, since the frameworks to represent ob-

servations are limited to physical quantities most of

the time, they rarely consider the Representational

Theory of Measurement. Consequently the opera-

tions they accept to perform are limited to additions

of quantities of the same quantity kind, or multiplica-

tion and division of quantities, such as (Schnabel and

Watanebe, 2013) proposes.

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

312

2.2.3 Uncertainty Calculation

Uncertainty calculation is of particular importance for

measurement. Some frameworks propose implemen-

tations of such calculations.

(Hall, 2006) proposes computing the measuring

process output values and uncertainties from the dif-

ferent inputs. The measuring process is built as a

set of interconnected components which represent the

different operations performed. The components im-

plement the uncertainty propagation law and recur-

sively request the uncertainty value of their inputs

from the previous components.

The OpenTURNS framework (EDF et al., 2013)

implements an uncertainty propagation methodology.

It requires a description of the measuring process as

an equation. Knowing this equation, and the values of

the inputs, OpenTURNS is able to compute the out-

puts uncertainty values of any measuring process.

2.2.4 Workﬂow

Being both the source of errors and the list of activi-

ties, leading to the measurement result, the measuring

process is of ﬁrst interest. Frameworks able to model

a measuring process are presented here.

Simple activities such as using sensors and mathe-

matical operations on quantities are described as pro-

cesses by (Botts and Robin, 2007). Moreover, their

model proposes process chains which are composite

processes able to sequence several simples processes,

leading to complete measuring processes.

Measuring processes, being simply the applica-

tion of observations and calculations activities con-

strained by a speciﬁc order, present the same charac-

teristics as any kind of workﬂow. The essential dif-

ference is the format of the inputs and outputs which

represent quantities. (WFMC, 1999) describes work-

ﬂows as the automation of a business process pre-

senting several activities, processing any kind of data

and connected through transitions. Such a standard

description may be used to model a measuring pro-

cess, considering that an activity is the production of a

quantity through the use of a sensor or a mathematical

operation. A workﬂow transition condition might be

enabled by the presence of a quantity resulting from

an upstream activity. But it might also be enabled by

a user action, such as turning on a sensor.

3 REQUIREMENTS ANALYSIS

The previous section presented several frameworks

that aim to implement metrology concepts and rules.

Although a lot of works exist, we did not ﬁnd any

solution able to model and design a mobile applica-

tion which assists an end-user during his own mea-

suring process and asserts that the application model

is coherent with the rules of metrology. This section

adopts a stakeholders point of view to present the re-

quirements analysis for a development platform that

a software architect can use in order to create such a

domain speciﬁc measurement application.

3.1 Developing Organisation

As a developing organisation, the industrial partner of

this project aims to propose a family of applications

dedicated to assist craftsmen during their daily work.

The application described in the introduction is one

of them, other examples are applications dedicated to

the estimation of room’s covering costs and assistance

of interior design. The production of such a family

of applications sharing similar properties will be sup-

ported by a model-driven development approach. In

this approach, a platform architect has to engineer do-

main speciﬁc measurement platforms and measure-

ment processes based on the description of end-user

domains.

3.2 End-users

End-users think with their speciﬁc business domain

vocabulary in mind while the architect conﬁguring an

application manipulates platform descriptions as well

as metrology concepts. A main requirement is to pro-

pose tools to reduce this semantic gap.

A speciﬁc measuring process model must be built

according to the description of a process an end-user

can formulate. This description yields knowledge

about the speciﬁc hardware and the measuring and

visualization instruments the end-user intends to use.

In addition to the prescription of quantities and units

for the process inputs and outputs, the end-user might

deﬁne uncertainties with maximum ranges. Thus, the

platform has to be able to compute the uncertainties

implied by the measuring process and propose to de-

ﬁne the inputs estimated worst cases uncertainties de-

rived from the end-user instruments and outputs un-

certainties.

To resume, in order to be valid the application

must :

• Produce the quantities the end-user requires.

• Be based on methods and speciﬁc instruments

handled by the end-user.

• Produce a result with an uncertainty range con-

form to the end-users requirements.

ADomainSpecificPlatformforEngineeringWellFoundedMeasurementApplications

313

Such a measuring process assistant becomes re-

ally pertinent when it is supported by the computing

capabilities of mobile devices which would be able to

accompany the end-user during his daily work. It is a

further requirement to be able to generate applications

targeting speciﬁc mobile devices.

3.3 Metrology Specialist

The different required metrology concepts are : repre-

sentations of observable attributes of objects, dimen-

sional analysis, uncertainty evaluation and measure-

ment scales brought by the Representational Theory

of Measurement. Indeed, the ability to handle non-

physical quantities is needed by the typical intended

end-user domains (e.g. type of roof tile). Since nei-

ther the end-users nor the architect are metrology spe-

cialists the purpose is to propose a platform which

can be used by these non specialists. So, the plat-

form must be able to deduce from a formal repre-

sentation of quantities the performable operations and

their impact in order to propose to the end-users a

veriﬁed result. Moreover, the model has to embed

enough knowledge about the empirical relations be-

tween properties manifested in the real world and

quantity semantics in order to deduce the correctness

of the operations performed on the values resulting

from the measurements. The term quantity semantics

encompasses any metadata (i.e. numbers, symbols,

scales, etc.) which allows to represent an information

based on the concepts and rules of metrology.

Thanks to the quantity semantics, a model check-

ing tool dedicated to metrology will be able to pro-

cess the whole model and assert its coherence. To

achieve this assertion, the model checker must be able

to interpret the quantities manipulation functions and

proceed dimensional analysis as well as restrictions

based on the measurement scales. To enhance usabil-

ity by non-specialists, each quantity has to be asso-

ciated to a default measurement scale. Indeed, the

different quantities brought in the process are consid-

ered captured by a real or virtual sensor able to di-

rectly measure the quantity. This naturally associate

a measured quantity to a given scale. Moreover, each

speciﬁc sensor description may contain an associated

measurement scale.

Finally, any observable property might ultimately

become a quantity. It is therefore necessary for the

platform to be open and give the possibility to a spe-

cialist to deﬁne new importable quantities.

3.4 Architect

The role of the architect is to design a domain spe-

ciﬁc measurement platform (c.f. ﬁgure 1). This role

implies two complementary activities. First, during a

platform engineering activity he conﬁgures the plat-

form, based on an end-user domain description. Sen-

sor speciﬁc models and formal representations are

used to reduce the gap between end-user and architect

language levels. Then, using a conﬁgured platform,

the architect realizes a process engineering activity.

During this activity he models and veriﬁes the domain

speciﬁc measuring processes based on an end-user de-

scription. The description provided by the end-user

contains knowledge about the end-user usual measur-

ing activities and their sequencing. From this descrip-

tion the architect builds a workﬂow model of the mea-

suring process. To this end, the platform has to pro-

vide tools to compose workﬂows. This encompasses

activities requiring the user participation to measure a

quantity, activities applying automatic measurement

operations and sequencing of activities.

Process

engineering

<Role>

Domain Speciﬁc

Measurement

Platform

Designer

Measuring

process model

realize

in charge of

out

Architect

played by

Platform

engineering

Domain Speciﬁc

system conﬁguration

realize

in charge of

out

Process

description

Sensors

models

Target model

inin

Domain model

Figure 1: Architect Activities.

The measuring process model can contain any

kind of function the end-user might need. These func-

tions are classiﬁed as :

• Measuring, this implies the use of measuring in-

struments to generate quantity values. These

functions imply the association of quantity se-

mantics to the values.

• Processing, this implies the use of mathematical

operations on the quantity values. These functions

imply the manipulation of the quantities seman-

tics according to the process performed.

• Presenting, this implies the use of speciﬁc inter-

faces to display the quantity values. These func-

tions suppress the quantities semantics and con-

vert the results to adequate formats.

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

314

4 DESIGN AND PROTOTYPE

The generation of a mobile application relies on a

suite of interpreted tools following a model driven ap-

proach (c.f. ﬁgure 2). Each tool produces a model

which is consumed by the next tool. The three main

tools are a Process Modeler to compose a measuring

process model, sequence activities, edit the associated

displays and deﬁne the targeted devices, a Metrol-

ogy Core which veriﬁes the conformance of the pro-

cess model to the metrology rules and a Model Trans-

former which transforms the process model into an

application model that governs the behavior of the

Mobile Embedded Application. The Model Trans-

former also transforms the process model into a for-

mat used to estimate uncertainties.

Error in model

Architect

Model

Transformer

Metrology

Core

Process model

Process

Modeler

Use

Veriﬁed model

Uncertainty

Valid model

MEA

Executable

model

Figure 2: Domain Speciﬁc Measurement Platform.

4.1 The Process Modeler

The Process Modeler is the tool the architect uses to

generate a process model (called MEASURE model).

This model is the result of the transposition of the

end-user measuring practices in the software domain,

taking into account the measuring devices he owns

and handles. The MEASURE model is composed of

a workﬂow layer and a measure layer.

The workﬂow layer, models the sequencing of the

measuring activities. The activities model measur-

ing operations and transitions model quantities eval-

uations. The Process Modeler allows the architect to

produce the workﬂow by sequencing activities imply-

ing either automatic (i.e. combination of quantities)

or user involved (i.e. use of a sensor to grab a quan-

tity value) metrological operations. To assist the user,

any user involved activity proposes an associated user

interface. Depending on the activity, this interface

presents guidelines, interactions to input values, or

displays of captured data needing a conﬁrmation.

The measure layer models the real operationaliza-

tion of the process activities as found in the presented

application. To easily create new measuring mod-

els or reconﬁgure existing ones, this layer leverages

on a connectable component model paradigm. Such

paradigm offers combinatory capacities allowing the

architect to re-use models or part of them in order to

propose new behaviors. Figure 3 presents the compo-

nent model. Three main component classes describe

the system functions presented in section 3 and guide

the modeling activity. Indeed, the architect builds

the MEASURE model according to the user descrip-

tion. This description is based on sensors, quantities

manipulations and presentations which are modeled

by classes inheriting respectively from the Measurer,

Processor and Presenter components.

Component

Measurer Presenter Processor

Connector

out in ins

ExternalInterface

Unit

Link

source

sink

outs

Communication

Interface

Figure 3: Components model.

The ExternalInterface components represent in-

puts or outputs. They associate a communication in-

terface with the element meant to capture the mea-

sured value and a unit which speciﬁes the associated

semantic. The Processor components represent alge-

braic manipulation of the quantities.

Sensors and presenters are highly dependent on

the application domain. Indeed, a domain speciﬁc

application description may include specialized sen-

sors and presenters. Therefore, the Process Modeler

proposes to extend a basic conﬁguration of the plat-

form with speciﬁc sensors and presenters libraries ac-

cording to the needs. Such extensions usually impact

the platform’s three tools. It is likely that the archi-

tect would not be able to perform these extensions.

Then, a metrology specialist might be required to han-

dle their implementation. These extension activities

may typically occur during the early stage of the plat-

form usage. When several developments have been

performed using the platform, the different domain

speciﬁc extensions architect might need will likely be

available as imports.

This capacity to extend the platform according to

the speciﬁc application domain reduces the semantic

gap between the architect and an end-user. Indeed,

these extensions give to the architect the ability to

build a MEASURE model using the elements and vo-

cabulary of the description furnished by the end-user.

ADomainSpecificPlatformforEngineeringWellFoundedMeasurementApplications

315

4.2 The Metrology Core

Section 2 highlighted that quantities and units add

knowledge to refer the measured values to the ob-

served world. Also, dimensions and scales are con-

cepts able to deﬁne if an operation on quantities leads

to a meaningfull result. The Metrology Core is a

model checking tool whose purposes is to use such

concepts to verify the MEASURE model in order to

assert that the quantities generated through the mea-

suring process can be used in order to draw valid con-

clusions on the real world. The Metrology Core en-

codes knowledge in ﬁrst-order predicate logic based

and uses the Prolog language to reason about that

knowledge.

Metrology concepts are modeled with a set of

clauses and predicates that can be considered as being

at a meta level. These clauses and predicates are then

instantiated to represent and verify a speciﬁc MEA-

SURE model. A fundamental principle of the metrol-

ogy core is to convert each quantity kind into a canon-

ical representation. Hence a set of rules has been de-

signed that associate one concept to another :

• a base quantity kind to a base unit ; for example :

aliasedU(mile,length).

• a base quantity kind to a canonical representation

that also allows the treatment of derived units : ?-

simplifyURep(length,Res). uniﬁes with :

Res = [power(length,1)].

• a unit to a measurement scale:

defaultScale(celsius,interval).

The representation of units is a prolog implementation

of (Rijgersberg et al., 2011) class diagram of unit of

measure. The scale of a derived quantity is deduced

from the set of operations leading to it. The measure-

ment scale resulting from a combination of quantities

is the most restrictive scale of the participating quan-

tities scales or the scale implied by the operation it-

self. A MEASURE model is then straightforwardly

transformed into a Prolog model. In fact, each MEA-

SURE model component has a predicate counterpart.

For example, the Measurer and Presenter components

become input and output predicates.

The role of an input predicate is to extract a canon-

ical representation and a measurement scale from the

unit of a Measurer component.

Similarly, the role of an output predicate is to ex-

tract a canonical representation from the unit of a Pre-

senter component. This canonical representation is

then compared by uniﬁcation with the canonical rep-

resentation of the variable resulting from the predicate

representing the component it is connected to.

The Processor components are associated to alge-

braic operation clauses. These clauses apply dimen-

sional analysis and scale coherence analysis, through

uniﬁcation, according to the algebraic operation the

components model. If the analysis present no issues,

the result is a variable holding the canonical repre-

sentation and the scale resulting from the operation

application.

Listing 1 shows an excerpt of prolog generated

by the Metrology core on a simple solar panel price

calculation process illustrated by ﬁgure 4. The uni-

ﬁcation returns true meaning that the process model

does not present any issue considering the dimen-

sional analysis and the measuring scales coherence.

This example uses on purpose different units such as

mile, meter, are, euro, dollar to show that the uniﬁca-

tion is done as long as the units are convertibles from

one to another.

Area

product

textIn (mile)

I_L1

textIn (metre)

I_L2

textDisp

(squareMetre)

Out_Area

Price

product

textIn

(euro/are)

I_PP

textDisp

(dollar)

Out_Price

Figure 4: Solar panel price MEASURE model.

Listing 1: Solar panel price process veriﬁcation.

?- input(mile,I_L1),

input(metre,I_L2),

input(unitQuotient(euro,are),I_PP),

operation(multiplication,I_L1,I_L2,Area),

output(Area, squareMetre),

operation(multiplication,Area,I_PP,Price),

output(Price,dollar).

Our Metrology Core is not limited to physical

quantities, as listing 1 suggests it by using money

quantities. As long as a base quantity is associated to

the units and a scale is associated to the base quantity,

it is possible to use all types of quantities.

As mentioned in the Process Modeler description,

it is possible to import domain speciﬁc extensions to

the platform. Such extensions are handled through the

prolog consult mechanism.

4.3 The Model Transformer

To propose a process uncertainties estimation during

the process validation phase, the Model Transformer

translates the measuring process to the OpenTURNS

framework (EDF et al., 2013) and let this framework

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

316

evaluate the uncertainties. This solution requires con-

sidering the on-line adaptation of the measuring pro-

cess. Indeed, applications are able to consider their

surrounding using sensor values, it is then possible to

generate a process which considers the surrounding

state such as the luminosity, the temperature or the

end-user experience with the application to adapt the

activities sequencing. Such on-line adaptation implies

that the uncertainty estimation process has to consider

all the possible measuring sequences, but highly im-

proves the resulting quality due to ﬂexibility and con-

text aware usage. The Model Transformer deduces

from the MEASURE model the possible activities se-

quencing and submits them to the uncertainty estima-

tion.

The Model Transformer also generates the Mo-

bile Embedded Application’s executable components

conﬁguration model (c.f. Section 4.4) if the model’s

uncertainty estimation matches the user needs. Also,

any model created by an architect can be reused in

an other application as is. Indeed, the communica-

tion between the different components being seman-

tically checked quantities, it is possible to exchange

any component by a compliant one. This feature im-

plies modeling modularity and reusability. The con-

struction of applications is then facilitated by pre-

senting complex structures as simple components im-

ported from previous constructions. All the complex-

ity of building the application is hidden in the Model

Transformer.

4.4 The Mobile Embedded Application

After building and veriﬁcation of MEASURE mod-

els, the next step concerns the generation of speciﬁc

mobile embedded applications.

A Mobile Embedded Application instantiates and

executes a MEASURE model. Since the MEASURE

model is built on a component model, the application

runtime is also built on the same paradigm. Such

choice presents two advantages. First keeping the

same paradigm will ease the transformation between

the MEASURE model and the application descrip-

tion. Second, in later evolutions of the application

it might be possible to embed the Process Modeler in

the ﬁnal application, enabling an experimented user

to modify the process it is performing considering not

forecasted situations. The modiﬁcations on the model

could then be added dynamically to an application.

In order to propose such capabilities, we deﬁned

the MEASURE model executable counterparts as

connectable components activated by a synchronous

model of computation. The application proposes

three main elements. A process builder which con-

nects and conﬁgures the components handling the

measuring process the user will follow. A process

executer which is able to sequence the components

of the process enabling the user to perform its mea-

suring process. A components library which con-

tains the MEASURE model executable counterparts

deﬁned for the speciﬁc mobile platform the applica-

tion is installed on.

Internally all operations rely on a canonical repre-

sentation of units. Hence conversions only happens at

the inputs and outputs. Each Measurer or Presenter

component is associated to a speciﬁc component han-

dling the interaction with sensor or display and with a

component handling the conversions between the re-

ceived or wanted quantity unit

We made a prototype implementation using Ob-

jective C to target iOS devices. Currently it is possible

to generate the component conﬁguration from a de-

scription based on the application network model and

to execute this network. We implemented the iPhone

5 embedded gyroscope, a bluetooth communicating

laser rangeﬁnder and textual input as sensors, simple

math operations as processes and text presenter.

In this section we highlighted that designing a Do-

main Speciﬁc Measurement Platform able to generate

an application to assist an end-user during its mea-

surement activities based on a veriﬁed measuring pro-

cess requires combining several technologies. A way

to tackle the induced complexity is to rely on a model

driven approach.

5 CONCLUSION, FUTURE WORK

This paper introduced that non-metrology specialists

may have access to more and more embedded sen-

sors and measuring tools. This can lead to erroneous

deductions in a measurement scenario. These errors

encompass conversions errors, communication issues,

sensors misuse, process steps omission, mishandled

operations on quantities and false decisions based on

poor quality results.

The importance of metrology concepts in measur-

ing activities is highlighted and a general framework

to reduce the mentioned errors occurrences is de-

duced from the literature. It appears that even though

metrology is an active domain of research, metrol-

ogy sensitive software frameworks do not cover all

the concepts mentioned in the design of well founded

measuring assistant applications.

An analysis, using a stakeholders point of view,

showed the requirements such a system has to ful-

ﬁll. To try to integrate the various implied aspects,

a model driven approach was used.

ADomainSpecificPlatformforEngineeringWellFoundedMeasurementApplications

317

Following requirements analysis, a software ar-

chitecture based on three distinct tools and a plat-

form embedded application was presented. Each tool

is a response to the presented requirements. A Pro-

cess Modeler enables an architect to model an end-

user process. Proposing importable sensors and do-

main speciﬁc vocabulary, it indexes the communica-

tion and sensors issues. A Metrology Core, based

on dimensional analysis and measurement scales ver-

iﬁcation asserts that the measuring process presents

no mishandled operations on quantities. A Model

Transformer translates the process model to the Open-

TURNS framework to estimate results uncertainties.

This tool also transforms the process model into an

executable model consumed by a Mobile Embedded

Application.

Based on this architectures and a model-driven ap-

proach the paper also presented a functional proto-

type using executable connectable components and a

Metrology Core implemented in Prolog.

Further work consist in strengthening the meta-

models of the process model and the model consumed

by the embedded application to further automatize the

transformations. Open issues include the analysis of

the impact and the implementation of the workﬂow

layers in each of the tools presented, process exten-

sions using more elaborated operations (e.g. statis-

tics) and implication of targeted mobile platforms in

the process modeling.

REFERENCES

Astarita, G. (1997). Dimensional analysis, scaling, and or-

ders of magnitude. Chemical Engineering Science,

52(24):4681 – 4698.

BIPM (2006). SI brochure, 8th edition - the international

system of units (SI).

Botts, M. and Robin, A. (2007). OpenGIS sensor model

language (SensorML) implementation speciﬁcation.

OpenGIS Implementation Speciﬁcation OGC 07–000.

Cmelik, R. F. and Gehani, N. H. (1988). Dimensional anal-

ysis with c++. Software, IEEE, 5(3):21–27.

Daponte, P., Vito, L. D., Picariello, F., and Riccio, M.

(2013). State of the art and future developments of

measurement applications on smartphones. Measure-

ment, 46(9):3291 – 3307.

Domotor, Z. and Batitsky, V. (2010). An algebraic–analytic

framework for measurement theory. Measurement,

43(9):1142 – 1164.

EDF, EADS, and PhiMeca (2013). Reference guide Open-

TURNS version 1.1.

Ellison, S. L. R., Rosslein, M., Williams, A., et al. (2000).

Quantifying uncertainty in analytical measurement. In

Eurachem/CITAC Guide. Eurachem.

Farrance, I. and Frenkel, R. (2012). Uncertainty of mea-

surement: A review of the rules for calculating uncer-

tainty components through functional relationships.

The Clinical Biochemist Reviews, 33(2):49–75.

Foster, M. P. (2013). Quantities, units and computing. Com-

puter Standards & Interfaces, 35(5):529 – 535.

Fowler, M. (1997). Analysis patterns: reusable object mod-

els. Addison Wesley Professional.

Hall, B. D. (2006). Component interfaces that support mea-

surement uncertainty. Computer Standards & Inter-

faces, 28(3):306 – 310.

Hand, D. J. (1996). Statistics and the theory of measure-

ment. Journal of the Royal Statistical Society. Series

A (Statistics in Society), 159(3):pp. 445–492.

Hodgson, R., Keller, P. J., Hodges, J., and Spivak, J. (2013).

QUDT - quantities, units, dimensions and types.

ISO (2009). Iso 80000-1:2009 quantities and units – part 1:

General. International Standards Organization.

ISO (2011). ISO 19156:2011 geographic information –

observations and measurements. International Stan-

dards Organization.

Khan, S. and Finkelstein, L. (2013). Mathematical mod-

elling in the analysis and design of hard and soft mea-

surement systems. Measurement, 46(8):2936 – 2941.

Krantz, D., Luce, D., Suppes, P., and Tversky, A. (1971).

Foundations of Measurement, Vol. I: Additive and

Polynomial Representations. NY Academic Press.

Mari, L. (2013). A quest for the deﬁnition of measurement.

Measurement, 46(8):2889 – 2895.

OIML, BIPM, CEI, IFCC, ILAC, ISO, UICPA, and UIPPA

(2008). Evaluation of measurement data - guide to

the expression of uncertainty in measurement - JCGM

100:2008 (GUM) - NF ENV 13005:2009.

OIML, BIPM, CEI, IFCC, ILAC, ISO, UICPA, and UIPPA

(2012). International vocabulary of metrology – ba-

sic and general concepts and associated terms (VIM).

JCGM 200:2012 [ISO/IEC guide 99].

Rijgersberg, H. (2013). Semantic support for quantitative

research. PhD thesis, s.n.], S.l.

Rijgersberg, H., Wigham, M., and Top, J. L. (2011). How

semantics can improve engineering processes: A case

of units of measure and quantities. Advanced Engi-

neering Informatics, 25(2):276 – 287.

Schadow, G. and McDonald, C. J. (2009). The uniﬁed code

for units of measure. Regenstrief Institute and UCUM

Organization: Indianapolis, IN, USA.

Schnabel, M. C. and Watanebe, S. (2013). Boost c++ li-

braries boost.units 1.1.0.

Stephenson, A. G., LaPiana, L. S., Mulville, D. R., Rut-

ledge, P. J., Bauer, F. H., Folta, D., Dukeman, G. A.,

Sackheim, R., and Norvig, P. (1999). Mars climate or-

biter mishap investigation board phase I report novem-

ber 10, 1999.

Stevens, S. S. (1946). On the theory of scales of measure-

ment. Science, 103(2684):677–680.

WFMC (1999). Terminology and glossary. Technical Re-

port WFMC-TC-1011, Issue 3.0, Workﬂow Manage-

ment Coalition.

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

318