Applying Model-Driven Development to Environment Monitoring
System
Tran Cao De
Can Tho University, 1 Ly Tu Trong, Ninh Kieu, Can Tho, Vietnam
Keywords: Environment Management, Environment Monitoring System, Formal Method, Model-Driven Development,
Sensor System.
Abstract: Environmental monitoring is critical in understanding whether the quality of our environment is getting
better or worse. Information gathered by using an environmental monitoring system is important to make
decisions. Vietnam is a vulnerable country of climate change. Specially, in the South of Vietnam, the
Mekong delta is known as the region getting the most impact of sea level rise in Vietnam. That leads to a lot
of problems making the worst effects to residents in the area, who are mainly still very poor. On the other
hand, Vietnam is going on industrialization process that makes a strong effect on the environment. To deal
with these challenges, different projects of environment management have been proposed and implemented
and many monitoring systems have been built in those projects. Those systems are basically sensor
networks with high cost in developing and maintaining. They are related to modern technology such as
cloud, communication mobile and wireless. They provide the data for large community for different
purposes. Therefore, building such a system is normally a long term project that requires an incremental and
modular development for a complex system. This paper, on one hand, represents some common
characteristics of an environment monitoring system that requires more study to develop a formal model and
a methodology for their specifications, implementations and verification. On the other hand, we would like
to adapt the formal model approach proposed for Intelligent Transport Systems (ITS) to an environmental
monitoring system. The framework of Baobab is also introduced as an example for transformation from
model to code.
1 INTRODUCTION
Environmental monitoring is critical in
understanding the quality of our environment. The
natural environment encompasses all living and non-
living things occurring naturally on Earth or some
regions. Many environment monitoring systems are
concerned to large fields of interesting in
environment, including air, water and other things
related to social-economic development of a region.
Recently, Vietnam has been known as a vulnerable
country of climate change and sea level rise.
Mekong delta is the region getting the most impact
of sea level rise. That leads to a lot of problems such
as flooding, saltwater intrusion, and soil changes.
These problems make extremely bad effects to
millions of peoples, 80% of whom are farmers living
poorly in rural areas. Vietnam is going on
industrialization process that makes strong effects on
the environment. To deal with these challenges,
Vietnamese government, with support of different
international program funds, is investing in many
projects of environment management. Information
technology has played a central role in these projects
by the present of environment monitoring systems.
One key step in developing an environmental
management program is to assess the current state of
the environment, normally, with supporting of an
environment monitoring system. Such a system is
based on sensor networks with high cost in
development and maintenance. They are related to
modern technologies such as cloud, communication
mobile and wireless, and provide the data for large
community including central government, local
government, science and civil social-economic. The
fact is that a monitoring system is complex and high
cost. Therefore, it must be invested in a long term,
different projects, and spent on different phases for a
long term service. In one hand, it requires
developing in an incremental cycle with the
soundness of modular. It also requires engineering
and re-engineering to adapt to new technologies of
information technology, and assessing of
577
Cao De T..
Applying Model-Driven Development to Environment Monitoring System.
DOI: 10.5220/0005338205770584
In Proceedings of the 3rd International Conference on Model-Driven Engineering and Software Development (CMDD-2015), pages 577-584
ISBN: 978-989-758-083-3
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
environment quality. In the other hand, there are a
lot of entities involved, and various scripts and
decisions which need to be implemented and
verified to make sure the correctness in
functionalities. In that context, Model Driven
Development (France and Rumpe, 2007)0, (Hailpern
and Tarr, 2006) appears to be a good practice on
such systems.
Model-Driven Engineering (MDE) (Schmidt,
2006) is an approach to create software systems that
involves creating models and applying automated
transformations to them. The models are expressed
in modeling languages (e.g., UML) which describe
the structure and behavior of a system. MDE is often
indistinctively associated to OMG's Model-Driven
Architecture and Model-Driven Development. The
ability to create a software design and apply
automated transformations to generate an
implementation helps to avoid the complexity of
platforms, component technologies and frameworks.
Many MDE solutions focus on generating code to
fulfill the functional requirements as well as the
quality attribute requirements such as performance
or reliability.
In this paper, we represent some common
characteristics of an environment monitoring system
that can be considered as specifications in a high
abstraction of these systems. Then, the proposed
approach for Intelligence Transport systems is
adapted to give a formal model for an environmental
monitoring system. The framework of Baobab also
is introduced as an example of generating code from
the model in UML notation.
2 OVERVIEW ON
ENVIRONMENTAL
MONITORING SYSTEM
Figure 1: Six stations of the water environment monitoring
system of Can Tho city.
Common Characteristics
Environmental monitoring system is a complex
system that composes the processes and activities
needed to characterize and monitor the quality of the
environment. It is often used in the preparation of
environmental impact assessments. It can be used as
an information system that is designed to capture the
current status of an environment or to establish
trends in environmental parameters. The system
automatically measures or captures the factors of
environment characteristics, normally, in real time
and continuity. There are innumerous concerns
about the environment but a lot of monitoring
systems is related to water, air, and soil. There are
many parameters needed to measure in order to
understand the quality of the environment. For
example, some parameters of an air monitoring
system of Vietnam (Circular 2/2014/TT-BNNPTNT)
should be:
- Weather and climate: Temperature, wind
direction, humidity, atmosphere, solar
radiation, etc.
- Dust: SPM, PM10, PM2.5 and PM1
- Chemical: SO
2
, NO and NO
2,
etc.
Figure 2: A fixed station for measuring water level of Can
Tho city.
All environment monitoring systems (EVMS) must
capture the interesting data from the environment.
Therefore, they share some common characteristics:
Sensor based system, Mobile system, and
Distributed system.
From the architecture model, an EVMS is often
designed for monitoring a geographical area that can
be very different in scale from worldwide to a local
area like an industrial zone. Monitoring system
needs capture data from different sites in the
monitoring area. Therefore, it is naturally a
distributed system in which data is collected from
different stations. Figure 1 gives an example about
the water monitoring system in Can Tho city, which
is composed by six stations.
MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment
578
Figure 3: Mobile station for monitoring the air.
A station may be a fixed one as in Figure 2 or
mobile one as in figure 3. A station can be seen as a
set of devices being set up at a place, including
measurement instruments or sensors to capture
environmental data. It provides data in the real time
and in continuity. A station needs to be autonomy,
i.e. so that it should properly work without a keeper.
Therefore, it needs power to function and sends the
captured data to a data center. A station can be a
mobile station (see 0) to capture a status of
environment in an area temporally. 0 represents a
fixed station to measure water level of Can Tho city.
It works in autonomy mode with solar power source,
radio transmission for data communication, and a
security camera. A station may use various types of
sensors to get interesting data. The environmental
status is interpreted from the captured data and it can
be used as an indication of environmental quality at
the monitored site. They are also transmitted to a
monitoring center to be analyzed in associating to
data coming from other stations. It is evident that
this architecture can be repeated at a higher level.
Many monitoring centers can share their own data to
others or there is a bigger center that collects data
from different monitoring centers in a wide
geographical area.
Figure 4 gives an example of an EVMS in a
general context. The system is composed of different
monitoring stations. Each station can communicate
with its center by a communication channel (radio
transmission, for example). Then, the monitoring
center collects data from its stations. In higher level,
a monitoring center may be in a larger system that
connects difference centers national-widely or
world-widely. In a general view, collecting data
from monitoring centers may not require any
restriction on the type of center. These centers need
not be homogeneity to provide the same kind of
data, but a rich variety of data for different purposes.
An EVMS is naturally sensor-based or instrument-
based. EVMS uses sensors or measurement
instruments to capture data from the environment. It
will share common characteristics that synthesize in
Figure 4: Distributed model of an ENMS.
ApplyingModel-DrivenDevelopmenttoEnvironmentMonitoringSystem
579
(Meritxell et al., 2011), including:
- Physical distribution in deployment of system:
the operation involves computation and
information resources that are physically
distributed.
- Resource boundedness: sensors must be able to
operate though they are low-powered; their
computation capabilities are low; and
communication is slow and intermittent.
- Information uncertainty: incomplete information
about the states of sensors, the network, and the
environment
- Large scale: a numerous number of sensors,
making unfeasible computational approaches that
do not scale with the number of sensors.
- Decentralized control: sensors may need to
coordinate to achieve tasks that cannot be
achieved by the operation of a single sensor.
Such coordination must occur without a
centralized control.
- Adaptiveness: The operation of a sensor network
is expected to be adaptive, namely the network
must autonomously and dynamically adapt to
either external and internal changes
Functionality
From the functionality view, a monitoring system is
a long term service system. However, it is normally
outdoor and vulnerability that needs maintenance to
keep well-functioning. In this context, hardware may
be changed that lead to re-implement the system to
adapt it to new interface or services provided by the
hardware. This is an important characteristic of an
EVMS that may need to apply Modem-driven
development to deal with.
An EVMS might be a large system, rich in
hardware and functionalities but it might follow
some patterns:
Sensor Level: at the input of any monitoring
system, there are a lot of sensors for capturing
different aspects of the environment. However, each
sensor provides a kind of raw data, for example,
water temperature, PH of the water, etc. In some
case, more than one sensor is used to measure one
feature of the environment. Normally, sensors can
provide data continuously, but they later is collected
and stored periodically. The first simple function of
any monitoring system is to capture the simple value
of each environmental quality factor and to make
some warnings as soon as possible if the values
obtained are out of acceptable range. Table 1
represents a sample data of a water monitoring
station in Can Tho city, including water level,
temperature, total dissolved solid, oxygen, electrical
conductivity, turbidity, and salinity.
Figure 5: An air monitoring station in Ho Chi Minh City.
Table 1:A sample data of a monitoring station in Can Tho city.
Time
Water
level
Temperature
(
o
C) PH
Total Dissolved
Solids (TDS)
Oxygen
(DO)
Electrical
conductivity Turbidity Salinity
2013/12/31 00:01 2.54 23.69 8.21 0.1 0.09 0.16 2083.0 0.08
2013/12/31 00:06 2.57 23.7 8.21 0.1 0.09 0.16 2083.0 0.08
2013/12/31 00:11 2.6 23.69 8.17 0.1 0.09 0.16 2084.0 0.08
2013/12/31 00:16 2.63 23.68 8.19 0.1 0.09 0.16 2083.0 0.08
2013/12/31 00:21 2.65 23.69 8.19 0.1 0.09 0.15 2084.0 0.08
2013/12/31 00:26 2.67 23.68 8.16 0.1 0.09 0.15 2083.0 0.08
2013/12/31 00:31 2.7 23.67 8.16 0.1 0.09 0.15 2084.0 0.08
2013/12/31 00:36 2.72 23.68 8.15 0.1 0.09 0.15 2083.0 0.08
2013/12/31 00:41 2.74 23.67 8.15 0.1 0.09 0.15 2083.0 0.08
2013/12/31 00:46 2.76 23.66 8.12 0.1 0.1 0.15 2083.0 0.08
2013/12/31 00:52 2.78 23.65 8.18 0.1 0.1 0.15 2083.0 0.08
2013/12/31 00:57 2.8 23.66 8.14 0.1 0.1 0.15 2084.0 0.07
MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment
580
Station Level: a station may be composed by many
sensors of several types to give a set of features
about the environment. Its sensor system can be
designed to give a set of values for a special
purpose. Figure 5 gives an example of air
monitoring station in Ho Chi Minh City. The
functionalities of a monitoring station may be quite
simple and programmed to manage a set of sensors,
collecting data from its sensors periodically and
sending them to a data center to store and use. This
collection task and transferring data to data center,
normally, are the charge of a programmable
controller. In this simple model, a monitoring station
may be designed to work as an autonomy system
including: power system and the security system (for
the station itself). It also is designed for providing
more complex function, for example, displaying
collected data and the status of the environment or
alerting automatically if any environment
characteristic violated.
A station can be seen as an outdoor system, some
important characteristics are required:
- Fault tolerance: a station still works properly
even if some sensors are corrupted.
- Easy to maintain: all components of the station
can be replaced easily for example, replacing
the power system (or battery), replacing the
sensor, replacing the main controller board.
The replacement may not need to keep the
same kind of hardware.
A station is programmed to function. All stations in
a monitoring system are not necessary homogeneity,
i.e. the same kinds of hardware, drivers and
architecture. Developing software for a station may
need a flexible model to deal with these
characteristics, specially, to deal with change
(hardware and drivers) and regression test (or retest)
the system. Such a model is also needed to deal with
the instantiation of a system for the same
functionality but different on hardware and
architecture. Figure 6 represents the detailed model
of water monitoring system of Can Tho city. In the
system there 6 stations with the same functionality
that is measuring a set of characteristics of water,
but they are not homogeneity, for example in data
transmission way. How Model-driven development
can contribute to this issue?
Monitoring Center: main functions of a monitoring
center are to receive the collected data from different
stations, store them to analyze the environment
status, and then predict the environment trends. It
also exchanges its own data with the other
monitoring systems world-widely or national-
widely. Therefore, it can be seen as a data center for
multi-purposes for a long time. For example (see
figure 6) the water monitoring system of Can Tho
city collects data from its six stations, store them in a
private cloud. It also transfers its own data to NISCI,
a national monitoring center via the Internet.
Actually, this system is in function to capture
environment data. The use of these data now is quite
simple: showing the variations daily, weekly,
monthly or yearly (see figure 7); alerting the
violation of standard safety criteria.
Figure 6: The detailed model of water monitoring system of Can Tho city.
ApplyingModel-DrivenDevelopmenttoEnvironmentMonitoringSystem
581
The functionality of the water monitoring system
of Can Tho illustrates a simple function of a
monitoring center: data collection, data store, and
simple analysis to alert the standard violation,
providing information for administration and
management purposes, and transferring data to
another monitoring system at a higher level.
Figure 7: Variation of DO indices at the station E of the
water monitoring center of Can Tho on September 1st,
2014.
Exploiting collect data for complex analysis to
establish environment status or predict the trends of
the environment is very special issues that need
different advance math models or machine learning
techniques. We do not intent to go in depth to these
subjects. Here we limit ourselves at simple
functionalities of a monitoring center.
In data collection and storing, the system needs
to be changed to add new indices of new
characteristics of the environment. At the processing
tasks to alert the standard violation criteria, the
system need to be changed for updating new
calculation model or math formula or adapting to
new criteria (i.e. new thresholds).
Figure 8: Average of PH of water of Tam Giang lagoon in
the dry season (no rain fall).
Providing information to administrative or
management is a complex issue. The collected data
need to be analyzed, synthesized, and represented in
an appropriate format. For example, in figure 7, the
raw data about DO index on 1/9/2014 is synthesized
and drawn continuously. A similar synthesis may be
done weekly or monthly. The collected data from
different stations can be synthesized and represented
as a map to give a general sense about the
environment as in figure 8.
The web page of Vietnam Environment
Administration
(http://www.quantracmoitruong.gov.vn) provides
draw data and synthesized information in a few
formats from a lot of monitoring centers in Vietnam.
In a higher hierarchy level, a monitoring center
can be a branch of another monitoring center, or
simply transferring its own data to a lager data
center to store and use.
3 APPLICATION OF FORMAL
MODEL
In the previous section, the common functionalities
and characteristics of an EVMS have been
represented. Such a system shares many
characteristics of an ITS (Intelligent Transport
Systems) introduced in (Bonnefoi et al.,2007). In
fact, an EVMS also has 4 problems identified by
(Bonnefoi et al, 2007):
Problem 1, Modeling Diagrams: it needs to ensure
that the specifications will be adapted to perform a
formal verification for a large set of main concerns,
including sensors, batteries (or power sources),
controller boards, communication devices, ports, etc.
The standard UML notion and the adapted
methodology proposed in (Bonnefoi et al.,2007)
might be applied.
Problem 2, Structuring the Specification: EVMS has
the same context as ITS. Beside of a large set of
heterogeneous elements to specify, there are also
numerous actors that are communicating and
interacting in parallel, such as time event, sensor
status, threshold came over… An appropriate
template as in the case study of (Bonnefoi et
al.,2007), that enables variation of architectures and
components parameters, might be a good idea.
Problem 3, Specifying Behaviors: UML State
diagram is good for expressing internal behaviors of
components. However, as stated in (Bonnefoi et
al.,2007)., under some conditions, it is possible to
MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment
582
transform a state chart into a formal notation like
Petri Nets for analysis purpose.
Problem 4, Analyzing the System: Analysis of a
system, formal methods as theorem proving and
model checking are normally used.
By sharing the common characteristics and problems
between EVMS and ITS, we adapt the solutions
proposed in for ITS to EVMS. This approach, by
using UML and Petri net, can resolve some issues,
including:
- Modular architecture of the system
- Definition of components interfaces
- Definition of components behavior
- Assembling of the formal model
- Analysis.
The proposed approach is compatible with software
design; the methodology relies on UML and sticks
to the V-model for software life cycle. It introduces
a guideline to designers to achieve EVMS
specifications that is suitable for formal analysis and
verification. It also shows a way to enhance the use
of the achieved results from one modeling formalism
to the others. The interesting point is the full
automation of translations from UML models into
Symmetric Nets models in the framework of a
development tool as Eclipse Modeling Framework
(Budinsky et al., 2003), a basis for such an
implementation.
The case study represented in (Bonnefoi et
al.,2007). illustrates that we can use UML notation
and Petri net to specify an ITS that can be
implemented by a development tool as Eclipse.
However, there is no detail of the implementation,
neither transformation to code. For this issue,
Baobab (Bahar et al., 2009) might give us a good
example.
At a higher abstraction, EVMS naturally is a
kind of wireless sensor system (WSN). Baobab is
known as a model driven development (MDD)
framework for WSN applications. It provides a
generic meta-model (GMM) which is versatile
across different application domains including
EVMS. Baobab allows meta-model users to extend
GMM for defining their own domain-specific meta-
models and platform-specific meta-models.
Baobab’s model-to-code transformer type-checks
and validates a given application model and
generates application code in nesC (Gay et al., 2003)
for TinyOS (ww.tinyos.net/). Here, we do not intend
to go further in Baobab. We just introduce its meta-
models and models for WSN applications. In fact,
Baobab defined Generic Meta-model Elements
which represent sensor devices used in WSNs. All
sensor classes, representing a specific type of
sensors, extend from the base class Sensor. The most
common sensor types that can be used in a variety of
applications are defined in the generic meta-model.
For functional requirement specification, Baobab
also defines several elements to express the most
common functional aspects of WSNs. The functional
tasks can be modeled by using
ConditionalFunctional element of the GMM and
they can further be specialized into RepetitiveTask.
There are other tasks used to define the receipt of
data as DataReceipt and to request a waiting period
before another task can be called, as WaitingTask.
Sensing phenomena can be modeled with
SensingTask. This element retrieves the new-created
SensorData from the Sensor.
The non-functional requirements can be modeled
explicitly by means of NonFunctionalTask class in
the GMM. Some examples of non-functional tasks
defined in the GMM are: ClusterFormation for
dividing the network into clusters; ChangeSleepTime
for adjusting the sleep time to minimize energy.
Main classes defined in Baobab are represented in
Figure 9.
Baobab provides a domain specific modeling
language to specify WSNs and generate code of
such systems. Moreover, it also illustrated that we
can build the model with UML notation and
implements it by using a development tool as
Eclipse.
4 CONCLUSION
In this paper, we presented common characteristics
of EVMS that share characteristics of ITS. An
EVMS is a complex system based on sensor
networks, organized in hierarchy structure, including
sensor level, station level and monitoring center. An
EVMS consists of sensing devices distributed in a
wide area, which coordinate to produce meaningful
information. Sensor networks are usually wireless,
dynamic networks composed of a large number of
(possibly heterogeneous) sensors. These sensors
acquire pieces of information that may substantially
differ in content, resolution, and accuracy. It is
related to a large set of devices that are placed
outdoor, in the environment. Therefore, it needs to
do a lot of work on maintenance to ensure it works
properly. As a consequence, it is necessary to have
some appropriate approaches to facilitate the
specification, design, development and maintenance
of the system. In fact, it needs some formalism for
modeling, structuring the specification, and
ApplyingModel-DrivenDevelopmenttoEnvironmentMonitoringSystem
583
Figure 9: Functional Aspects of the Generic Meta-model in Baobab (Bahar et al., 2009).
specifying system behaviors and analyzing.
Specially, it also needs a flexible model to deal with
changes in its life cycle. The formalism proposed for
an ITS based on UML and Petri net might be well-
adapted to an EVMS. It is suitable for adding new
sensors, replacing sensors, changing data collection
schedules, changing data representation, adding new
environmental criteria.
An EMVS is basically a sensor network and
might be a kind of wireless sensor network.
Therefore, the framework provided by Baobab has
illustrated a way to specify an EMVS (specialized of
WSN) and generate its code. It can be seen that the
formalism in Baobab can be expressed in UML
notation and implemented in other language as well.
These features provide a lot of advantages for
incrementally developing and maintaining an EVMS
as a long term project.
REFERENCES
France R, Rumpe B., Model-Driven Development of
Complex Software: A Research Roadmap, Proceeding
Future of Software Engineering, Fose'07, 2007, Pp.
37-54.
Hailpern B. and Tarr P., Model-Driven Development: the
Good, the Bad and the Ugly, Ibm Systems Journal,
Vol. 45, No. 3, Pp. 451-461, 2006.
Schmidt, D.: Model-Driven Engineering. Ieee Computer
Magazine 39 (2), 2006.
Circular 32 of the Ministry of Agriculture and Rural
Development, 32/2014/Tt-Bnnptnt, Dated 10/09/2014,
Vietnam.
Meritxell Vinyals, Juan a. Rodriguez-Aguilar, Jesus
Cerquides , a Survey on Sensor Networks from a
Multi-Agent Perspective, the Computer Journal
Archive, Volume 54 Issue 3, March 2011, Pp. 455-470.
Bahar Akbal-Delibas, Pruet Boonma, Junichi Suzuki,
Extensible and Precise Modeling for Wireless Sensor
Networks, Information Systems: Modeling,
Development, and Integration, Lecture Notes in
Business Information Processing Volume 20, 2009, Pp
551-562.
David Gay, Phil Levis, Rob Von Behren, Matt Welsh, Eric
Brewer, and David Culler, the Nesc Language: a
Holistic Approach to Networked Embedded Systems,
Proceedings of Programming Language Design and
Implementation (Pldi) 2003, June 2003.
Http://Www.Tinyos.Net/
Bonnefoi, F., Hillah, L.M. ; Kordon, F. ; Renault, X.
Design, Modeling and Analysis of Its using Uml and
Petri Nets, Intelligent Transportation Systems
Conference, Itsc 2007. Pp. 314-319.
Http://Www.Quantracmoitruong.Gov.Vn/En-Us/En/Home
.Aspx
F. Budinsky, D. Steinberg, E. Merks, R. Ellersick, and T.
Grose, Eclipse Modeling Framework, Ser. the Eclipse
Series, E. Gamma, L. Nackman, and J. Wiegand, Eds.
Addison-Wesley Professional, August 2003.
MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment
584
