Ninth International IBPSA Conference

Montral, Canada
August 15-18, 2005

INTERFACING BUILDING PERFORMANCE SIMULATION WITH CONTROL

MODELING USING INTERNET SOCKETS

Azzedine Yahiaoui*1, Jan Hensen1, Luc Soethout2, Dolf van Paassen3

1
Center for Building & Systems TNO-TU/e
P.O. Box 513, 5600 MB Eindhoven, The Netherlands
2
TNO Built Environment and Geosciences
Postbus 49, 2600 AA Delft, The Netherlands
3
TU Delft, Department of Mechanical Engineering
Mekelweg 2, 2628 CD Delft, The Netherlands

as for heating, ventilation and air-conditioning

ABSTRACT (HVAC) and lighting. They also perform tasks such
This paper reports on progress of an ongoing as access control, energy management, and fault
research project, which aims to achieve better control detection and diagnoses. The principal setup of a
modeling in building performance simulation by BEMS as depicted in Figure 1 is to provide a healthy,
integrating distributed computer programs. Recent comfortable and productive indoor environment
developments show that there is a need to enhance while reducing fossil fuel consumption and green
building performance assessments by integrating new house gas emissions, and to achieve this in an
simulation features in order to predict the overall efficient and rational way (see e.g. IEA 2002).
effect of innovative control strategies for integrated Central computer
building systems. However, both domain
independent control modeling environments and
domain specific building performance simulation,
have their own restrictions. For example, certain Netwerk

control features are represented in one simulation

Substation
environment while others are only available in other (computer) Substation

simulation software. To alleviate these practical

problems, this paper describes a mechanism that can +
be used to allow a building simulation environment
to exchange data with an external control simulation -

environment.
In particular, this paper focuses on the problem of Figure1 General architecture of a building energy
developing run-time coupling of control and building management system
performance environments over TCP/IP using
Internet sockets. The socket implementation is As mentioned by e.g. Van Paassen (1986), a
analyzed in terms of minimizing overhead, computer based control system is much more flexible
communication efficiency, and the integration into than traditional control systems. The software allows
existing software tools. Perspectives for a run-time adapting the control strategy to changing conditions.
coupling specification are given to enable The central computer defines which control loops are
connection-oriented sockets to easily exchange data suitable to be activated and what would be the
as well as coupling software. Data requirements in appropriate set-points for the current situation. This
view of integration in real building control protocols computer is connected to substations via a network.
(BACnet and LonWorks) are discussed. An early Substations constitute the automation level that
implementation of run-time coupling is demonstrated collects message (data) and regulates actuators. For
with a case-study, and the paper finishes with some communication, a protocol is used to exchange those
conclusions and directions for future work. data between the central computer and the system
components. Usually, a message includes the address
INTRODUCTION of the component and the series that refer to the real
physical quantity (e.g. the desired position of valve).
With the rapid advances in microelectronics,
powerful computer-based building monitoring and It is very important that the advanced control
control systems have emerged. These Building strategies that are possible with BEMS are taken into
Automation Systems (BAS) or Building (Energy) account already during the design phase of the
Management Systems (BEMS) control systems such building and its installation. This requires that such

*
Corresponding e-mail: [email protected]

- 1377 -
strategies are properly modeled and incorporated in Language (XML; see e.g. Chervet 2004). Nowadays
building performance simulation. However, existing XML might even be used as a gateway between
domain specific software for building performance different communication protocols such as LonMark,
simulation (BPS) is usually relatively basic in terms BACnet and Modbus, all of which are used in
of control modeling and simulation capabilities (e.g. BEMS. Communication protocols that are used by
ESP-r, TRNSYS). On the other hand, we now have BEMS have come up with a structure to describe the
domain independent control modeling environments information that is sent across the network. The same
(CME), which are very advanced in control modeling kind of information needs to be exchanged in our
and simulation features, but rather limited in building distributed simulation environments where building
performance simulation concepts (e.g. Matlab/ modeling and control system are separated and work
Simulink). Marrying the two approaches by run-time together through run-time coupling. This motivates
coupling would potentially enable integrated our choice to use internet techniques based on socket
performance assessment by predicting the overall communication for the run-time coupling between
effect of innovative control strategies for integrated simulation environments
building systems.
The remainder of this paper is organized as follows:
Although, our current work starts from specific BPS the next section describes the background of run-
and CME simulation environments, the time coupling. Then it follows the reasoning behind
communication mechanism and data-exchange using Internet sockets in the implementation. This is
protocol that is being developed and implemented, followed by case study demonstration, and finally
have much wider and more general applicability. conclusions and directions for future work.
ESP-r and Matlab / Simulink were selected as BPS
and CME software tools respectively. ESP-r and BACKGROUND
Matlab/Simuling can run on different operating
Coupling different applications (simulation
system platforms. ESP-r is used for building
executables) at run-time provides the facility to
modeling only and Matlab/Simulink is considered as
exchange the information in a co-operative way.
a controller that can be configured remotely. It is felt
Usually, one application controls the overall
that this type of technology development will enable
simulation procedure and requests the other
design and evaluation of advanced building control
application (s) when necessary. For example, Janak
applications that are currently infeasible.
(1997) has enabled run-time coupling between ESP-r
and the ray-tracing lighting simulation and
visualization software Radiance.
In principle, the coupling of tools requires an
underlying communication mechanism to allow the
transfer of both data and control between codes. The
data transfer corresponds to the sending of the results
calculated by one of the processes and to the
receiving by another process whereas the transfer of
control is the execution of one particular function to
be performed remotely in another process. Most of
the attempts to run-time couple codes rely on the use
of shared (or intermediate) files, which is not a very
efficient way to exchange data at an adequate level of
abstraction (Hughes and Hughes 2003, Ranganathan
et al. 1996). In (Yahiaoui et al. 2003), we described
and compared also various other options for inter-
Figure 2 Networked building simulation and application data transfer facilities.
automatic control systems
Inter-process communication (IPC) can be viewed at
For the coupling between the different simulation different levels of abstraction. At a higher level, it
environments, we want to mimic as much as possible appears as an exchange of information between
the situation of a real BEMS. As schematically processes. These processes are either contained in a
shown in Figure 2, overall system integration is in single parallel application or in distributed
our approach achieved by connecting the various applications, or may otherwise belong to different
hardware and software subsystems to a network, applications that need to communicate. At a lower
such as the Internet or a local area network (LAN). level, communication appears as the mere traffic of
In order that the subsystems can communicate bits from one address space to another. Distributed
information, it is necessary to use a common data simulation requires support for expressing
exchange format such as the eXtensible Markup communication at high-level of abstraction. A

- 1378 -
powerful abstraction in distributed applications is the made up of an IP-address concatenated with a port
IPC mechanism that deals with distributed object number. In general, sockets use a client-server
systems, such as CORBA (Common Object Request architecture. The server waits for incoming client
Broker Architecture), in which computer request by listening to a specified port. Once a
applications, written in any oriented-object request is received, the server accepts a connection
programming language, can work together over the from the client socket to complete the connection. In
network. However, as reported earlier (Yahiaoui et our case, ESP-r is the client/master and
al. 2004), adapting CORBA to explicitly run-time Matlab/simulink is the server/slave. Hence,
couple ESP-r with Matlab (neither of which are Matlab/simulink can be launched at every ESP-r
written in an object-oriented language) raises major time-step as a separate process.
difficulties in dealing with the CORBA interface
definition language (IDL). IPC USING INTERNET SOCKETS

In contrast, socket programming is complicated, but While run-time coupling involves many aspects, the
we can build more sophisticated communication current paper focuses on programming network
abstractions on top of sockets (e.g. Stevens 1999). communication. Internet domain communication
Socket communication has been used earlier for which uses the TCP/IP (Transmission Control
evaluation of HVAC control performance (e.g. Protocol/Internet Protocol) suite is the most prevalent
ASHREA 1997). In addition, a socket domain is an set of communications protocols in use today.
abstraction that provides addressing structure and a TCP/IP provides communication services that are
set of protocols such as: Internet Protocol, BacNet, interoperable with virtually any network
LonTalk, and so on. For such reasons, the use of configuration, allowing the internet sockets to act as
internet sockets seems a good alternative for the basic element of the network.
developing and implementing a coherent paradigm
UDP vs. TCP protocol
for communication between BPS and CME software.
Socket types define the communication proprieties
DEVELOPING RUN-TIME COUPLING visible to the application. Indeed, the term TCP/IP
FOR SIMULATIONS refers to more than just the TCP protocol itself. It
specifies an entire that bridges communications
The initial objective of the current research is to between the transport layer and the upper (session/
marry the ESP-r and Matlab environments in order to presentation/ application) layers through a network
be able to perform distributed building control medium. Within the TCP/IP suite, two protocols are
simulations. This involved identifying the specific mainly used: TCP (Transmission Control Protocol)
components that should be connected in order to and UDP (User Datagram Protocol). Both of them
communicate data between ESP-r and Matlab, which, have their own advantages and disadvantages. Each
coincidentally, may run on different machines using of them can be better for a particular problem related
different operating systems. That is to say that in the to run-time integration of BPS and CME software.
run-time coupling approach, applications can be What follows is a short description:
distributed over heterogeneous environments. In this
regard, socket based IPC has been developed with a 1. TCP is the most prevalent of internet domain
set of techniques that permit communication over the communication. It supports connection-oriented
network as shown in Figure 3. communication between client (e.g. ESP-r) and
server (e.g. Matlab/ Simulink). Before any data is
transmitted, TCP establishes a mutually
acknowledged connection between the sending
components and the receiving components (or
nodes). During communication, this connection is
maintained through the transmission of data by
one component followed by an acknowledgement
(ACK) of receiving data at the other node. If the
sender of data does not receive this ACK, the
data in question will be retransmitted. Finally
when all data has been transmitted, the
connection is closed. Although TCP offers the
Figure 3 Networked communication between ESP-r advantage of reliability of transmissions (sent
and Matlab using sockets data is guaranteed to be delivered), this feature
costs time in connection set-up and error
Sockets are abstract communication endpoints. They checking, especially during large transmitting
are basic components of inter-process and inter- volumes. In essence, TCP may experience low
system communicating over a network. A socket is

- 1379 -
 performance due to significant overhead per in sockets application programming interface (Sock
transmitted data segment. APIs). The underlying concept upon which those
functions are constructed is that of a socket, which
2. UDP provides communication in connectionless simply refers to the communication between ESP-r
mode (no notion of connection). UDP and Matlab through the network.
communication consists only of transmitting one
data packet with attached routing information. The data exchanged with UDP represent the physical
UDP is not responsible for notification prior to quantities as they are measured in a real application.
the transmission of data nor for confirmation of However the data that should be exchanged from
data delivery. As a result, UDP offers no delivery ESP-r to Matlab and required passing back from
guarantees. This simplicity, however, offers Matlab to ESP-r are first converted by flattening of
significant benefits in saving time for connection structured data items to external (network)
set-up and error checking. Furthermore, since no representation on one side, and then reconverted by
connection must be established, the components rebuilding data structures to internal representation
of sending and receiving data can be made to run on other side. This exchange format has for the
promptly, regardless of whether they are purpose to transfer data in safe mode. The main
successfully completed. UDP is usually used to advantage of this approach is to enable run-time
efficiently exchange small datagrams (<64MB). coupling of the simulation environment with a real
building (e.g. for control purposes) or with building
UDP allows applications to exchange data across
components (e.g. HVAC systems).
the network with a minimum of overhead.
In the development of run-time coupling between DATA COMMUNICATION IN A REAL
ESP-r and Matlab over the network, more BUILDING
particularly between controller and building model, In real operations, it is desirable for control loops to
the time that it takes to send and receive small data communicate with building components or HVAC
volumes might be the limiting factor on allowable systems. The communication between controllers and
sampling interval. This can happen when a type of building systems requires an interface or a gateway,
controller, which uses or depends on dynamic due to their different communication speeds and data
parameters, requires to instantaneously changing the formatting. The proper operation of the gateway is
position of a specific building component. For speed dependent on the continued use at the corporate level
of network communications, the UDP protocol to implement between controllers and building
would be a better choice than TCP. We envisage systems. In consequence, two standard protocols
using a building domain specific control protocol, i.e. BACnet and LonWorks (see e.g. SmartHome
BACnet/IP (a data communication protocol for 2000) may be distinguished:
Building Automation and Control Networks). The
Both protocols cover most or all layers of the OSI
BACnet protocol itself provides the guaranteed (Open System Interconnection) model for network
delivery of data. This does not require the use of communication. For applying some of the ideas
TCP. Moreover, the added reliability of TCP may developed for these protocols in our case of run-time
seem to give an advantage over UDP, but this may coupling, only application layer is considered. The
not necessarily be the case. Improved reliability in other layers deal with hardware specifications and
TCP is achieved by the re-transmitting data feature the low-level details of the network communication.
that guarantees delivery but also guarantees that the In our case, these details are already covered by the
re-transmitted data will arrive later than intended. socket implementation. The application layer
The delivery of this late data is not worth the however deals with the conceptual part, i.e. with the
additional network bandwidth it requires, as closed way of information is modeled.
loop would be better served by a new transmission of
recent data than by re-transmission old data. Indeed, BACnet is a specification for a standard protocol
the timing of a run-time coupling mechanism published by ASHRAE organization. The run-time
requires the ability to execute send and receive coupling mechanism is implemented in order to
functions with great precision. Summing up, UDP create a communication protocol that complies with
this specification. The standard defines protocol
seems to be the better choice for our type of
implementation conformance statements (PICS) that
applications. Therefore an appropriate signal for
identify different levels of compliance. This meets a
necessary error checking and possible repeated
field level of Building Energy Management System
function calls is implemented (i.e., select() function
(BEMS) that run-time coupling between ESP-r and
is used).
Matlab uses to communicate over BACnet protocol.
Exchange data with UDP To support communications on a variety of LANs or
At the top layer, application level communications to allow different components to co-exist in the same
within the control loop are performed using functions network, LON (or LonWorks) that communicates via

- 1380 -
Standard Network Variable Types (SNVT) facilitates Matlab as regular Matlab. In case we need to deal
interoperability by providing a well-defined with Simulink, the link should also be performed
interface. Basically a proprietary communication between each other. Practically the same approach is
protocol is called LonTalk, created by ECHELON used for Simulink, although MEX S-functions are
Corporation. A chip is required for any device that used and are dynamically linked programs within
uses LON. Standard network variable formats can be Simulink. Thus, when an application requires
established by run-time coupling to allow the transfer computational control dynamics, it should call
of data between Matlab and ESP-r using the Matlab code. For the use of Matlab / Simulink
LonMark subset of LON capabilities to interoperate interoperability; technically permutations flexible
with each other. usage model between Matlab, Simulink and C/C++
are described in table 1. More detail about can be
IPC TO ESP-R AND TO MATLAB/SIMULINK found in (MathWorks, 2005).
BINDING
To use sockets IPC, C and C++ programming Table 1 Interoperability between MATLAB,
languages are used of which socket libraries are Simulink, and C/C++
exploited. Neither building performance simulation
FROM CALL HOW?
software (ESP-r) nor control modeling environments
C/C++ Matlab MATLAB Engine
(Matlab/Simulink) have simple interface with sockets
ActiveX/COM & DDE
communication mechanism. Therefore we have MATLAB Compiler & runtime
describes an approach, which is developed in usual Matlab Simulink Using sim function:
way of interfacing sockets communication >> sim(mymodel)
mechanism to ESP-r and Matlab/Simulink as well. Simulink C/C++ S-function Builder
C-code S-function
Interfacing to ESP-r
C/C++ Simulink Engine or COM
Currently ESP-r is the legacy of Fortran codes. [x,y,t] = sim(mymodel,u0)
Mixed language programming is used to interface Simulink Matlab M-code S-function
between Fortran and C/C++ programs (e.g. Einarsson Caution: interpreted, but not
1995). This is often a very appropriate way of compiled
combining the strengths of various programming Matlab C/C++ MEX
languages, in which the language subset is common ActiveX/COM & DDE
between C and C++, and between Fortran 77 and 90.
The developed approach consists of using a method Moreover, those external interface functions (MEX-
that can be functional for both Unix and Linux files and MEX S-functions) as described previously
Operating Systems in which ESP-r can be installed. can send and receive data to/from C program through
This method is based on external structure function a gateway (or a bridge) that was made. This can be
(extern struct) in order to add new variables that need done once we compile them with MEX utility. After
to be exchanged without modifying programming the compilation, it generates a DLL (Dynamic Link
codes. With this function, a ESP-r subroutine can Library) file for any operating system. With this
exchange data with the C client code of sockets procedure, Matlab control files can be invoked when
communication mechanism through the function was the server receives data from a client process. A
made. Afterward, the C client codes, which server which is generated by compiling MEX-file
developed and support sockets communication are and sockets communication codes in a group is like a
compiled within the concerned ESP-r packages. This DLL file which is used and recognized by Matlab.
creates one executable, a client process which is
RUN-TIME COUPLING OF BUILDING
formed by combining ESP-r packages and sockets
CONTROL STRATEGY
communication codes together.
In the current implemented approach of run-time
Interfacing to Matlab/Simulink coupling between ESP-r and Matlab, it is ESP-r
Matlab has a built-in utility called MEX (Matlab which starts simulation. As Matlab is a server for
Executable), which is used to convert C/C++ to ESP-r (a client). If the connection is not established
Matlab-Executable format. The original sense of the between client and server the data to be exchanged
Matlab/Simulink word represents two different are not transferred.
environments, which are a high-level programming The run-time coupling of ESP-r and Matlab strategy
environment and a graphical user interface developed in this work takes the form of a closed-
respectively. The link to external programs depends loop control system, as shown in figure 4. This can
on which environment we would interface. For performed either for the feedback control structure or
Matlab, MEX-files are used and are dynamically for the feedforward control structure. Both control
linked programs that can be called from within loops can be used for a simulation of building control

- 1381 -
systems. In case we would use open-loop control for socket communication, the use of some signals, like
building application, it would just necessitate to for example SIGIO, allows the transmission of data
setting, in Matlab side the variables requires passing in asynchronous manner (e.g. Leffrer et al. 1993).
back to ESP-r to zero, and more exactly as defined in Although the asynchronous simulation of coupled
Matlab language are: the rhs (right hand side) programs is able to deal with any updated
arguments. Finally, both unidirectional or information, the execution time can be significantly
bidirectional communication are carried out within reduced (e.g Fumagalli et al. 1999). Therefore, the
this run-time coupling strategy. controller is updated each time the events arrive from
building components, and verse versa. This step is
important in control systems when building
components are in the presence of delays or when
objects located in building change their actions
within the time. Consequently, asynchronous
simulation is efficient for an assessment of some
control analysis for integrated building systems. But
it is harder to parallelize, as the receiver does not
Figure 4 Structure of run-time coupling between know when the transmitter is going to send data.
control and building performance simulation Synchronous Mode
A user-interface in ESP-r is modified by adding one This event occurs when the two processes (ESP-r and
row of External Control at the end of the control Matlab) are synchronizing with the same defined
period window. This allows us to setup an external time in execution. Each process (either ESP-r or
controller for a specific period of day. Once the Matlab) waits for incoming data from another
simulation has been started, ESP-r alters data transfer process. In this case, the time step is defined by the
over a network to the involved controller in Matlab client process (ESP-r). On the other hand, the server
by run-time coupling when necessary. Then if data process (Matlab/Simulink) is in charge of the same
transfer requires passing back to ESP-r, the controller time step defined for ESP-r. This is a usual and
invokes ESP-r when its processing in Matlab is simple way of run-time coupling when ESP-r and
terminated. In this case, the implemented run-time Matlab exchange data at end of each time step, as
coupling of ESP-r and Matlab is supported with basis shown in figure 6.
data transfer protocols. The manner of data exchange
with Internet sockets depends on whether the
communication is synchronous or asynchronous. If
some building control applications requires running
in real-time, it is necessary that run-time coupling of
ESP-r with Matlab is in asynchronous mode.
Therefore the different communication events are
briefly described below.

Asynchronous Mode Figure 6 Synchronous mode

This event occurs when the processing of ESP-r and
Matlab are independent from each other. Figure 5 Since a building control model is executed
shows the event when the time step of one of them sequentially, the transmission by synchronous mode
(either ESP-r or Matlab) is different from another. blocks the whole simulation until the
simulation/execution environment receives data. In
real-time simulation, some of scheduled transitions
may be delayed because of this blockings. As a
result, if the implemented run-time of ESP-r and
Matlab uses synchronous mode, it should be ensured
that the time constraints of scheduled transitions are
satisfied (adjusting a timing of a control loop).

BUILDING CONTROL APPLICATION

Figure 5 Asynchronous mode To demonstrate the application of the developed and
implemented run-time coupling between ESP-r and
However, asynchronous (parallel) simulation of run- Matlab, an application case is presented here. The
time coupling between ESP-r and Matlab is application comprises a working office space unit
characterized in that they contain no-blocking send (4.8*4.2*2 m3) with two radiant-ceilings used for
and no-blocking receive functions. Indeed within both heating and cooling mode. The on/off controller

- 1382 -
is used to regulate the appropriate temperature inside implemented in Matlab. The obtained results for
the room by opening or closing the valves on external control are within the use of implemented
pipelines (either the pipeline of warm water or the run-time coupling between ESP-r and Matlab using
one of cold water). The constructions used in this Internet sockets. This external controller is presented
office space are internally insulated cavity walls and when ESP-r and Matlab are executed in the same
internally single glazed walls. The office is located in host and when they are run in different hosts. In the
a six floor of the building sited around the atrium, as present application, the building model is carried- out
shown in figure 8. in ESP-r and the controller is set-up in Matlab. By
combining them by run-time coupling, the simulation
is performed in synchronous mode. Therefore,
Matlab is launched at every ESP-r time step as a
separate process. However, the obtained results for
internal control and external control are basically
similar during the working period (from 7:00 to
18:00 oclock). An accurate comparison shows that
the simulation results are identical. The results also
show the oscillation of different responses of
controllers around the set-point in working period,
this is due to controller capability in which control
Figure 8 The Building model action takes place every time a deviation occurs from
set point. This action responds quickly but is
The walls facing south and the atrium are in a single sensitive to input noise, which causes chattering at
glazed structure. It has a thermostat in which the user short intervals.
is allowed to set a temperature at five degrees higher The importance of run-time coupling between ESP-r
or lower than the common set-point, which is and Matlab is to incorporate advanced control
21degree-Celsius. This illustrated case-study analysis methods in building performance simulation
investigates an application that has two objectives. for improving all quality aspects of an indoor
The first consists of comparing the results of environment, and for enabling an integrated
(internal and external) controllers to the set-point approach to independent variables such as
within the same time step of 10mn/hour. The second temperature, light, solar penetration, etc. Using
is to qualify the importance of the run-time coupling more than one host at a problem might enhance the
approach. Figure 9 compares the simulated results by performance and the reliability of run-time coupling
regulating a temperature in the office. The simulated when the time plays a critical part for the simulation
temperature results are obtained using internal of advanced building control systems. The principal
control, which refers to the On/Off controller result of such an approach will certainly enhance
implemented and available in ESP-r, and external building efficiency while maintaining optimum
control which refers to the On/Off controller energy consumption.
25
24
23
22
21
20
19
18
17
16
15
0:00
1:00
2:00
3:00
4:00
5:00
6:00
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
0:00

Set-point Internal Control External Control External Control (diff. hosts)

Figure 8 Simulation results

It should be noted that it might be dangerous to use application, as the output signal from the On/Off
the asynchronous communication for the current

- 1383 -
controller is either fully on or completely OFF Current Directions in Numerical Software and
depending on the error deviated from a set point. High Performance Computing, Kyoto, Japan

CONCLUSION AND PERSPECTIVES Fumagalli, A. and Grasso, R., 1999. An Efficient

Asynchronous Simulation Technique for High
The paper has described a promising approach to Speed Slotted Networks, Proc. 32nd Annual
run-time coupling between ESP-r and Matlab/ Simulation Symposium, San Diego, USA.
simulink as implemented using internet sockets. This
approach is based on the UDP protocol to exchange IEA, 2002. Control Strategies for Hybrid Ventilation
data between building model and its controller. The in New and Retrofitted Office Buildings
development of this new advent would potentially (HYBVENT), Annex-35 Report, University of
enable integrated performance assessment of new Aalborg, Denmark.
building control strategies that are not yet possible. Hughes, C. and Hughes, T., 2003. Parallel and
This type of approach must be validated through Distributed Programming using C++, Addison-
application in a realistic system, especially since the Wesley, Boston, USA.
developed run-time coupling approach incorporates Janak, M. 1997. Coupling Building Energy and
advanced control strategies for integrated building
Lighting Simulation, Proc. Building Simulation,
systems with time-varying delays. Issues concerning
Vol. 2, pp. 313-319, Prague, Czech Republic
the networks reliability and utilization have been
discussed. The purpose of this research is to evolve Leffler J., Fabry R.S., Joy W.N., Lapsley P., Miller,
the prospects for integration in real building control S. and Torek. C. 1993. An advanced 4.4.BSD
protocols. interprocess communication tutorial, Technical
report, Computer Science Research Group,
Concerning the perspectives of this research, there is
a necessity to formalize the run-time coupling University of California, Berkeley, USA.
mechanism that uses a connection-oriented socket. MathWorks, 2005. MathWorkss website. <
TCP is a reliable mechanism that provides a better http://www.mathworks.com/>
reliability in communication and guarantees transfer
of data. An approach is envisaged to run-time couple SmartHome, 2000. SmartHomeFuroms website <
ESP-r and Matlab with TCP. This approach will be http://www.smarthomeforum.com/>
based on an architecture that consists of reducing the Schmitt, M., Acharya, A., Ibel, M. and Iancu, C.
overhead by minimizing the time for communication, 2001. Service Sockets: A Uniform User-Level
including latency and bandwidth. Furthermore, this Interface for Networking Applications. Technical
overhead can be reduced by minimizing function Report, Computer Science Department,
calls and optimizing the size of communication University of California at Santa Barbara, USA
buffers to data types need to be exchange.
Stevens, W.R., 1999. UNIX Network Programming.
Future research will involve applying this approach Vol. 2: Interprocess Communications, 2nd Ed.,
to realistic application case study settings to prove
Prentice-Hall, Upper Saddle River, NJ, USA
the success of implementing the current run-time
coupling. However, and more importantly, this Ranganathan, M., Acharya, A., Edjlali, G., Sussman,
research will generate associated knowledge for A. and Saltz, J., 1996. Flexible and efficient
general and wider applicability. coupling of data parallel programs, Proc. Parallel
Object-Oriented Methods and Applications, Santa
ACKNOWLEDGMENT Fe, Mexico.
Drs. A. Keizer of Centrum voor Taal en Techniek is Van Paassen, A.H.C., 1986. Control of Indoor
gratefully acknowledged for her helpful assistance in Climate Technology, Journal A, Vol-22, N. 03
editing this paper.
Yahiaoui. A., Hensen J.L.M. and Soethout L.L.
REFERENCES 2003. Integration of control and building
performance simulation software by run-time
ASHRAE 1997. A Standard Simulation Testbed for coupling, Proc. IBPSA Conference and
the Evaluation of Control Algorithms and Exhibition, Vol. 3, pp. 1435-1441, Eindhoven,
Strategies, ASHRAE 825-RP, GA, USA. Netherlands.
Chervet, A., 2004. XML and Building Automation, Yahiaoui, A., Hensen, J., and Soethout, L., 2004.
Journal ASHREA, N. 12, pp 24-32 Developing CORBA-based distributed control
Einarsson, B., 1995. Mixed Language Programming, and building performance environments by run-
Part 4, Mixing ANSI-C with Fortran 77 or time coupling, Proc. 10th ICCCBE, Weimar,
Fortran 90, Proc. International Workshop on Germany.

- 1384 -