You are on page 1of 8

Copyright © IFAC Algorithms and Architectures for

Real-Time Control, Vilamoura, Portugal, 1997

REALTIME CONTROL IN FACTORY AUTOMATION

P. Kopacek

Institute for Handling Devices and Robotics


Vienna University o/Technology
Floragasse 7a. A-J040 Vienna. AUSTRIA
Tel: +43-1-5041835. FAX: +43-1-5041835-9
e-mail: kopacek@ihrt1.ihrt.tuwien.ac.at

Abstract: Automation today consists of two main fields: the automation of continuos
processes - process automation- and the automation of discontinuous processes -
production automation. Both fields get closer together under the influence of
microelectronics or modem computing. In both fields AI methods are introduced. This
leads to new headlines in factory automation, like intelligent manufacturing systems (ims),
service robots, with new tasks for real time control.

Keywords: Real time control, manufacturing automation, Robots

1. INTRODUCTION control in chemical and process engineering, electric


power generation, position control in navigation and
Some "tools" of manufacturing automation and in biomedicine.
robotics necessary for realtime control will be shortly
presented.
1.2 Man - Machine Communication

1.1 Trends in realtime process control The user interfaces in process automation systems
get more and more user friendly because of
Process automation or the automation of continuos developments in "Man Machine Communication
processes is and will be dominated by "classical" (MMC)" an absolutely necessary tool for real time
control engineers. Therefore topics like adaptive control. Fully graphic screens operated by mouse or
control, nonlinear or timevarying systems, joystick, logical cross checks (thinking systems) and
identification and parameter estimation and the possibility to generate a distinct layout by users
advanced control algorithms are on top. are available. Such user friendly interfaces require
an improved computer hardware - storage capacity,
Some years ago the first intelligent sensors and clock frequency (computing speed )
actuators, equipped with one chip microprocessors,
were commercially available. Furthermore modem
computing technologies make it possible to realize 1.3 ArtifiCial Intelligence
modem control concepts like adaptive or "advanced"
control algorithms . The main problem is which Artificial Intelligence - AI - is since some decades an
control method should be applied for distinct ex1ensive research field. The first knowledge based
realtime tasks. According to a study some years ago, or expert systems were available in the late seventies
"advanced" algorithms are used for aircraft control, preferably for special fields in medicine. One of the

15
reasons was the computer hardware because AI 1.4 External sensors
methods require very high computer power. In the
last years the computer capacity increases Autonomously or (partly autonomously) guided
dramatically - especially from PC's - and therefore systems need to move in an unstructured
some methods of "Artificial Intelligence - AI" can environment without having a-priori knowledge.
be applied in an efficient way for real industrial Regarding to economical reasons, one is interested to
processes. For such purposes programming increase the autonomy of these systems so that they
languages and software tools are available for "low can take over some tasks of the operator, who again
cost" industrial applications. can spend his time for other, more special tasks. To
provide this autonomy, there is a need for sensor
One of the "magic terms" in automation in the last based guidance - thinking in terms of "low~st"
years was "fuzzy". Created by the mathematician solutions - operating on the base of ultrasonic
Lotfi Zadeh approximately 30 years ago and sensors.
introduced in automation in the early eighties by the
Japanese. In Europe - with the usual dead time - ten The "smart sensors" concept is based on the idea to
years later fuzzy logic and fuzzy control become use animal behaviour as a paradigm for the control
headlines. A lot of industrial applications appears in of man-made machines. Analyzing the complex
a short time - most of them only for ,,PR" purposes - behaviour of animals or human beings, one can
and nearly every producer of control devices has at isolate releasing mechanisms that cause a specific
least one with the label "fuzzy". In process response to certain situations. The design of these
automation fuzzy control can be used for the control behaviours can be considered as a fast human copy
of highly nonlinear processes, time varying systems, of the evolution process resulting in reliable and
multivariable control systems and for modeling and efficient control systems (Fig. 1).
identification of some complex processes.

Today fuzzy methods are mainly


used for control tasks in household
and utility goods, adaption of
controller parameters, as modules
Sensor Design
for PLC' s and microcomputer
controllers. Future applications
could be classification and pattern Physical Sensor Configuratio
recognition, fuzzy techniques for Smart Sensor
expert systems, neuro-fuzzy Sensor Management
techniques, fuzzy techniques for
Petri-Nets. The efficiency of fuzzy Behaviour Control
methods depends mainly from the
quality of the rule base. Furthermore
no stability criteria and only few Fig. 1. Smart Sensor Concept (probst, 1996).
optimization methods for fuzzy
controllers are published. At behaviour control each control system is
implemented as an independent module that
Artificial neural networks (ANN) for control generates a special response as a function f of the
purposes were introduced a little bit later than fuzzy input of sensors and other modules (f = f(xI, xl),
methods. The original idea was to copy some parts of with xl is the input of the sensors and xl is the
the human brain; e.g. a neural net is able to learn status of the other modules). It can be processed
from the environment. Applications are in robotics,
parallel to other Behaviour Modules - these modules
in steel production, in paper industry and in communicate via messages to each other. To manage
banking. The field of applications for neural nets are conflicts, the input information can be inhibited by
not so clear defined than for fuzzy concepts. another behaviour or an output response can be
suppressed.
Both concepts are merged to neuro-fuzzy in the last
time. Neuro-fuzzy is a combination of the ability to Classical approaches of control systems for mobile
learn from the neural nets and the transparency of robots are typically of a cascaded structure. Any
fuzzy methods. Today only few pilot applications are changes of the lower levels do have an influence to
known e.g. control of a washing machine, adaptive
the higher ones. The structure of a "Behaviour
control of steam generator and control of a grinding Control" is similar to the Object-Oriented approach:
process. These developments yield to new subjects in A behaviour is an abstraction of a set of rules that
automation like genetic algorithms. neuro- generate the specific response to a specific stimulus.
informatics and soft computing.

16
A ..Behaviour Module" receives messages from and 2.1 "Low cost" Computer Integrated
sends messages to other modules. The internal Manufacturing
infonnation processing is hidden - this encapsulation
allows an easy maintenance of the modules. The Especially SME's had a lot of difficulties to
modules can be developed in parallel and the introduce ClM or parts of CIM. Today only in few
adaptation to application-specific requirements can SME's a total ClM concept is realized. In most of
simply be done by adding new behaviour modules. the SME's only some ClM components (CAD,
CAP, CAM, CAQ, PPS) or parts of these are
An ultrasonic sensor system used for the autonomous realized. European and especially Austrian SME's
guidance is required to reliably detect a vast variety need more time to get familiar with the ClM
of objects. The degree of reflection of the ultrasonic philosophy and to introduce more components or
energy differs from almost 100% for a huge plane parts of them.
wall perpendicular to the acoustical axis of the
sensor down to 20010 for the legs of a hwnan being or Software packages for the ClM components are
even 1% for a tiny object such as the leg of a chair. available commercially from various sellers. Usually
"Smart Sensors", the combination between sensor these packages are only partially suitable for SME's.
element and behaviour control, are designed for the They offer a lot of features, often not necessary for
reliable and robust execution of sensor based the demands of SME's, on the other hand features
functions in the real world. necessary are missing. Adding some of these takes a
long time and it is usually very expensive.
Furthermore most of these packages requires a cost
2. COMPUTER INTEGRATED intensive computer hardware.
MANUFACTURING
Especially for the demands of SME's flexible,
The field of production or manufacturing automation modular "low cost" ClM concepts are necessary.
was dominated by ClM - Computer Integrated From the side of hardware, the basic philosophy is to
Manufacturing - in the last decade. The original idea use PC's (e.g. 486 compatibles or similar) connected
of a total computer aided or controlled production by a local area network (LAN) with a host computer
from the order or first product draft until the delivery for database tasks. The operating system for the PC's
was only realized in few cases worldwide mainly in is MS-DOS, for the host computer UNIX. As a LAN
large companies. serves ETHERNET.

According to common definitions, ClM consists of The modular "low cost" ClM concept is shown in
the following components: Fig.2. It uses two types of computers: A UNIX
machine serves as a database and network server,
CAD: includes all data processing activities, related various MS-DOS computers (AT or 386) work as
to development and construction, together network stations with different tasks. For OCA
with calculation and simulation (CAE) special terminals are used.

CAP: planning of work. programming of NC - The database-server and all workstations are
machines (results are basic data of the PPS - connected by a local area network (ETHERNET). A
system). second network (party-line) connects the OCA-
terminals with the OCA-server, which works as an
CAM: includes the support of computers for the intelligent gateway between the two networks.
control of the working means during the
production (machine tools, handling devices, The various control systems recognize on their own,
transport and stock systems). if the connection to the host server is possible and/or
succeeded. They build up the connections self-acting
CAQ: describes computer supported planning and and adjust the data between the different databases
performance of quality control. automatically.

PPS: includes the use of computer aided systems With this solution the personnel in the production is
for the organisatorial planning, controlling not stressed with additional data-processing
and monitoring of all processes. activities, in order to keep the control system
working. Each workstation has its own server
process running on the network server. These server
processes handle the communication between
workstations, processes and database. The software
is completely modular. "Modules" could be
commercially available or individually written

17
PPS Server
CAQ-

OCA Terminals

OCAGateway

CADINC Fileserver
CIM PC. ',..~...

NC Machines

Fig. 2. "Low cost" CIM concept (Kopacek, 1991).

according to specific demands of the user. These 2.2 Intelligent Manufacturing Systems - "ims"
modules can combined with a minimum of interface
problems - as usual in CIM systems today. The Generally, AI methods are introduced more and
programming language is C or C++. more in production automation or CIM systems.
This results in intelligent CIM components - ICAD,
This concept has following advantages: ICAP, ICAM, ICAQ - and in intelligent CIM
• low cost systems (lCIM) or in intelligent manufacturing
• Possibility of stepwise realization systems (ims). This new philosophy requires a lot of
• easy combination of software modules for a prerequisites and research.
specific solution
• possibility to include AI methods in some "ICIM" or "ims" is partially introduced in the
modules. industry but mainly for large companies. AI in fonn
of knowledge based and e},."J)ert systems is ready to be
From a control engineering viewpoint the operation introduced in an efficient way in CIM components.
of today's "Flexible Manufacturing System (FMS)" The implementation of AI methods in "low cost"
as a important CIM component is planned and open- CIM systems depends from the availability of AI
loop. This might be one of the reasons that FMS software packages for this hardware configuration.
operations are not very reliable. Therefore there is a
trend for introducing feedback and closed loop The results of the 6 test cases of the worldwide ims-
control techniques into FMS subsystems and initiative
operations. The incorporation of automatic tool TC2 Clean manufacturing
monitoring, automatic tool change as well as closed- TC3 Global concurrent engineering
loop quality control are some steps in this direction. TC4 Globeman 21
A more detailed analysis of the various FMS areas TC5 Holonic manufacturing systems
shows that because of the great variety of control TC6 Rapid product development
tasks involved , the application of standard control TC7 Knowledge systemization
techniques may not be sufficient for this complex
industrial environment. "Advanced" control are "tools" for the development and a new dimension
algorithms probably with methods of artificial in real time control.
intelligence have to be introduced.

3. ROBOTS

As a typical example for the present state and future


developing trends in production automation could
serve the industrial robot. Robots were 10 to 15 years

18
~
er;
"0 "0
~ ~

.-'--s::
] t;
Cl:
1.000.000
.s I
t /}

'-
c ./ 0
800.000 o 0
z :z 1.500
600.000
1.000
400.000
500
Year
1991 1992 1993 1994 1995 1996

Fig. 3. Comparison: Installed IR versus service robots (Schraft, 1993)

ago one of the headlines in factory automation. Integration of industrial robots in the manufacturing
Meanwhile robots are a tool on the manufacturing environment requires enhanced capabilities of the
level. hardware structure of the robot controller as well as
the robot programming system. Related to the first
aspect, fault tolerant multiprocessor architectures
3.1 Industrial robots represent a perforrnant solution, for nowadays
multitasking robot controllers. Robot languages
The worldwide robot population was growing up integrate sensory-based commands for adaption to
dramatically in the last years. Approximately ten the working environment.
years ago the number of robots in industry were
nearly doubled every three years. Therefore As an example could selVe a robotized assembly cell.
according to the forecasts five years ago the number To handle all the necessary different assembly
of robots working in industry world-wide should be sequences a hierarchical control structure was
increased 30% every year. One of the reasons is that chosen as shown in Fig. 4. At the lowest level there
in classical application fields like spot welding, are only local control functions, which are divided
spray painting, coating, materials and parts handling into subsystems with own controllers such as:
a saturation can be obtained especially in the last -7 Robots
three years. Work places requiring industrial robots -7 Screwing devices
are usually equipped with such robots. Three years -7 Soldering device
ago the industry was waiting for the intelligent robot
equipped with external sensors. As already known and Systems without controller like:
the development of such e:\."ternal sensors like visual, -7 Transportation system
auditive, force torque etc. is going on very well in -7 Part feeders
research institutes and laboratories. But until now
only few of these sensor concepts are available for All these subsystems interface to the sequence
industrial applications at a reasonable price. controller. This module is connected to a database
containing the following informations:
Therefore between 1992 and 1994 the producers of • All single robot programs
industrial robots were confronted with a decreasing • Start conditions for one part
demand of industrial robots. The number of installed • Definition of the sequence of assembly
robots in the industry worldwide was approximately • Information for plausibility check of robot
30 % less than the estimations five years ago. As a program and tooVgripper
logical consequence the prices for industrial robots • Information about the used storing devices or
decreases in these years dramatically. Latest tools.
developments deal with a modularisation of the • Information to avoid collision.
robots as well as the control system. Probably in two
or three years it might be possible to buy an The tasks of the sequence controller are to check the
industrial robot and separately from another availability of all components in each stage of the
company an advanced control computer at low assembly process, to start the corresponding robot
prices. program sequence and to send all the necessary

19
interface
toCIM
(Ethernet • lAN)

Man-Machine-
Interface

Robot- Part- Transportation Sensordatas external


controllers feeder systems controllers

Tools Screwing de.;c,;s


TooI-Changing- Soldering and
Systems Glueing devices
Grippers

Fig. 4. Control Structure of an Assembly Cell (probst, 1996)

commands to the connected subsystems. On the 3.2 Service Robots


other hand the sequence controller is connected to a
man-machine-interface. All error and status- Producers of industrial robots tried to recognize new
conditions are interfaced to this module. application areas. A broad field are so-called
unconventional areas. This unconventional areas can
The module for collision avoidance is mainly be divided in two classes:
responsible for the coordination of all assembling
operations where both robots are involved (e.g. one • Robots to replace human beings at work in dirty,
robot move a wire towards the surface of the printed hazardous and/or tidious operations
circuit, the other solders the two parts.). Furthennore
this module watches all parallel moving sequences of • Robots to operate on human beings to elevate
the robots and stops one, until the other robot incornodity or to increase comfort pleasure.
reaches the same point within the assembling
sequence. The Man-Machine-Interface module To the first category what is necessary for:
displays all error conditions and status infonnations • Operation in hazardous environment (e.g. in
of the cell and generates automatically report-files. high temperature, vacuum, underwater,
radioactive environment)
Therefore future industrial applications of industrial • Fire fighting (fire extinction, life saving)
robots can be divided in two directions. On the one • Police applications
hand side intelligent robots are necessary for • Military applications
increasing the number in classical application fields. • Architecture (inspection/maintenance of
In addition to the application fields mentioned before buildings/structures)
some other classical fields growing up dramatically • Publicity (attracting customers, etc.)
in the last two years, e.g. assembling as well as • Entertainment (appearance in films)
disassembling. For disassembling applications new • Housework (floor sweeping, cooking)
features of the industrial robots are necessary. These
robots require a combined position and force control To the second category robots can applicate in the
as well as relatively high payload. That requires the following fields:
availability of cheap ex"ternal sensors in the nex"t
• Medicine (patient care)
years for industrial applications.

20
• Guiding the blind additional features like combined force and position
• Entertainment (use in gaming houses, in control, external sensors based on microsystems,
amusement parches) flexible and light weight robots. Because of the
• Housework (table services) decreasing number of installed robots new
application fields will be recognized. One of these
Comparing with industrial robots, service robots fields are the service robots. Service robots look
would be characterized by the following facilities, to quite different than conventional and therefore
permit the operating effectively and unobtrusively in research have going on in additional directions such
the human living environment: as external sensors, new grippers and gripping
• Mobility devices, new kinematic structures. Efforts have to be
• Portability undertaken to further develop key components of
• Operating case these robots towards efficiency, performance,
• Sensingllearning/judging functions (artificial miniaturization and cost. Here the collaboration of
intelligence) research institutions, service industry and robot and
• Adaptability to widely varying operations and component manufacturers has the potential to create
environmental conditions valuable synergies.

The ultimate target to be reached would be a robot


that possesses faculties approaching that of human 4. SUMMARY
beings. Leaving such an ideal robot as a goal for the
future, intennediate robots that only satisfy a limited Microelectronics is the driving force to close the
selection of the most requisite functions should still historical gap between theory and practice in
find good use in human society. Among the faculties automation. We are now in the position to realize
cited above, what is the most indispensable for a modem control concepts like "advanced" control
service robot would be mobility. algoritluns, fuzzy methods, neural networks, expert
or knowledge based systems but we have no
e~:perience when such a modem concept is necessary
Today the number of service robots in operation is
very small. However, the evolution of people and efficient.
working in the service field shows a constant growth
rate and consequently enormous potential. Great The future of automation in the broadest sense will
hopes are attached to the e~:pansion of this sector, be determined by the development of
however the possibility of the automated microcomputers e.g. single chip processors from the
implementation of services through robot systems is hardware side. The software tends towards "user
barely apparent, neither to the supplier and friendly" with all advantages and disadvantages.
manufacturer, nor t9 the user/customer. The use of Advantages are e.g. the possibility in process control
robot systems in the field of services offers, in systems to create an own layout for the process or to
principle, advantages to all those involved. develop an own e~:pert system for a special purpose.
Disadvantages are the higher computer power for
To fulfill these tasks, research would be necessary in this tasks as well as the complexity of the new
the nearest future in the field of microsystems or software tools.
micro-mechanics. Microsystems or micro-mechanics
is an interdisciplinary research field growing up Introduction of methods of AI in automation is in
dramatically in the last five years. As a research field progress and will be increasing in the future. New
between mechanical engineering, electronics, control concepts like neural networks and genetic
computer science, Microsystems should be algoritluns are introduced in industrial applications
applicable in the nearest future for industrial robots. in the nearest future. The next generation is ready
from the theory and waits for practical applications.
Another direction taken in service robot
development is toward limited purpose robots to The future of production automation is dominated by
serve such specific purposes as "new" headlines like intelligent manufacturing
systems, agile manufacturing, holonic systems, rapid
• guiding the blind
prototyping, etc.). Historically production
• patient care
automation was always and will be in the future the
• floor sweeping
field of headlines. In reality we have to introduce -
• building/civil construction work
as pointed out in the paper - AI methods, starting on
• inspection, operation.
a very low level, in this field of automation. This
development requires new and advanced realtime
Future application oriented research in robotics is
control concepts.
dominated by two mean directions. Robots for
classical applications have to be equipped with

21
5. LITERATURE

Kopacek, P., A. Frotschnig. and V. Kacani (1991).


User interface to a CIM-database. In:
Proceedings of the Second International
Workshop on " Computer Aided Systems Theory
EUROCAST'9I". Lecture Notes in Computer
Sciences, Vol. 585, pp. 592-601. Springer
Verlag, Krems.

Kopacek, P. (1994). Robotics research today and in


the future. In: Preprinls of the Symposium
"Robotics in Alpe Adria Region - RAA 1994",
p. 213-217. Bled.

Kopacek, P. (1995). Trends in process and


production automation from the viewpoint of
European small and medium sized companies.
Bulletin of the Polytechnic Institute of Iasi Vol.
45, pp. 7-21.

Probst, R and P. Kopacek (1992). Robotized


assembly cells in low cost CIM concepts. In:
Proceedings of the 23rd "International
Symposium on Industrial Robots - ISIR'92",
p. 191-195. Barcelona.

Probst, R and P. Kopacek (1996). Service robots -


Present situation and future trends. In:
Proceedings of the 2nd ECPD International
Conference on "Advanced Robotics. Intelligent
Automation and Active Systems ". Vienna.

Schmidt, G. (1991). Towards integration of


autonomous subsystems for assembly and
mobility into flexible manufacturing. In:
Proceedings of the Int. Workshop "Information
Processing in Autonomous Mobile Robots",
pp.3-20. Springer.

Schraft, RD. (1994). Service Robots: Opportunities,


Possibilities and Potentials for the Ne:\.1 Decade.
In: Proceedings of the " IEEE Robotics
Conference ". San Diego.

22

You might also like