IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 1, NO.

1, MARCH 1996 5

What is Mechatronics?
David M. Auslander

Abstract-With the explosively increasing costlsize-effectiveness This definition is a generalization of the original definition
of computers, mechatronic systems are becoming common in any of mechatronics which encompassed mechanical components
engineering discipline dealing with the modulation of physical and electronically based decision making (digital and analog
power. In this article, I attempt a definition of mechatronics
that will allow for its differentiation as an identifiable field of circuitry). The term “mechatronics” has been attributed to the
engineering and then follow this direction to discuss some of Yaskawa Electric Co. in the early 1970’s, although I have not
the associated design philosophy. The definition draws on the seen any authority for that attribution. It makes sense in the
centrality of computing in mechatronic systems. context, for example, of the brushless dc motor which uses
electronics for its commutation decisions rather than relying
I. WHAT’S UNIQUE? on mechanically based decision making and, in the process,
results in a motor with dramatically different properties than
I F mechatronics is to be recognized as a discipline we
must articulate the aspects of it that set it apart from
other engineering activities. A definition that is too broad, by
the original.
With the advent of computing as the preferred decision-
encompassing everything, encompasses nothing. Likewise, to making medium, the definition needed to be broadened. It
has also become evident that a wider variety of physical
be too narrow does not do justice to the richness of the field.
systems can be encompassed than just mechanical motion.
I believe that the uniqueness associated with mechatronics
Examples of these include traditional mechanical systems such
comes from its exploitation of computation to create systems
that are qualitatively unlike any that came before. Although as machine tools, manufacturing automation, as well as vehicle
the name implies that to be “mechatronic” a system must only control (automotivie, airplanes), process systems, thermal and
environmental control, building vibration control, etc.
contain mechanical and electronic elements, that definition
Finally, it seems likely that while software (algorithmic
appears much too broad to be useful. In other respects it
languages) is the current decision-making medium, other tech-
is too narrow. I see mechatronics including many physical
nologies could supplement or supplant it. Current candidates
systems, not just traditional mechanical systems. It also begs
the question of what software is and ‘how it fits in, which include artificial neural nets and fuzzy logic. Thus, the defi-
nition points to the centrality of decision making rather than
is understandable since control software was not a significant
factor in the early 1970’s when the term was coined. the means of realizing it.
Software is currently implemented using electronic deviczs,
but is really a medium in itself, independent of implementa- A. Examples
tion. It is the source of unparalleled complexity-for better
or worse! Engineering systems with software based control Some examples can flesh out this definition. In the auto-
can work miracles, but the technology also brings risks. New motive area, for example, mechatronic engine controls have
technologies call for new methodologies but people don t enabled the modern automobile to simultaneously meet emis-
change their ways of thinking easily. sion, fuel economy, and drivability specifications that could
I think that this at least lays the groundwork for a definition not be duplicated using the traditional engine controller, the
of mechatronics that fits current technology and has some hope carburetor. This is an example in which a mechatronic solution
of covering future developments as well. produces essential pegormance improvements.
Another example is the modern machine tool. The CNC
controller defines the operation and use of such a machine.
11. DEFINITION OF MECHATRONICS
The entire manufacturing process has been affected, from
The application of complex decision making to the oper- designing the parts to be manufactured to the potential for
ation of physical systems. fully automatic, untended operation of the machines. This is a
“Complex decision making” covers today’s software and case in which the introduction of mechatronics to an existing
distinguishes modern mechatronics from earlier attempts at technology resulted in a qualitative improvement.
complexity. It identifies as “mechatronic” systems that depend The above examples show the application of mechatronics
for their unique functionality on software. It also identifies as to existing technologies. An example of a technology that
practitioners of mechatronics those who have the skills and is impossible without mechatronics is the automated teller
understanding to create and manage the complexity. machine (ATM) used to deliver cash to bank customers. Such
a machine is not even conceivable without mechatronics.
Manuscript received October 11, 1995; revised December 12, 1995. Finally, as noted above, new technologies carry new risks
The author is with the Department of Mechanical Engineering, University
of California, Berkeley, CA 94720-1740 USA. with them. A recent spectacular example of the risk at work is
Publisher Item Identifier S 1083-4435(96)02287-9. the baggage handling system for the new Denver International
10834435/96$05.00 0 1996 IEEE
6 IEEEIASME TRANSACTIONS ON MECHATRONICS, VOL. 1, NO. 1, MARCH 1996

Airport [3].Failures attributed to the software controlling the

highly automated baggage-handling system caused nearly a
year-and-a-half delay in opening the airport at a cost of the
order of a billion dollars.

111. CONTEMPORARY
REALIZATION Steam K
-
Engine Main Sha?
Computers of all sizes and capabilities are used as the
decision making elements in products affecting virtually every
Fig. 1. The Watt speed governor
resident in the developed world. From a numerical perspective,
the most common type of computer in use today is the micro-
controller-very small, highly integrated microprocessors used
as decision-making elements in engineering systems. Several
hundred million of these are sold every year!
The following item from a daily newspaper highlights
the extent to which embedded computing in a mechatronics I software. eledronics 1
context is entering daily life and consciousness:
SIDEBAR: USA Today, May 1, 1995, page 1B
USA SNAPSHOTS-A look at statistics that shape
your finances
Micromanaging our lives
Microcontrollers-tiny computer chips-are now run-
ning things from microwave ovens (one) to cars (up to
IO) to jets (up to 1000). Fig. 2. Mechatronic system elements.
How many microcontrollers we encounter daily:
1985 less than 3 spin outward, moving the linkage so as to close the steam
1990 10 valve and thus slow the engine. This oft-quoted example of an
1995 50. early feedback control system is typical of the systems on this
list-its decision making (computation) is done in the same
IV. A DESIGNPHILOSOPHY medium as the major power flows in the system. This makes
these systems premechatronic for reasons discussed below.
The unique factor in mechatronics is its dependence on
sophisticated real-time computation to define the nature of
the engineering system in which the computation is em- B. Mechatronic System Organization
bedded. Successful mechatronic system designs require an The developments in electronics have made it possible to
organized approach to the design of such software. As part of energetically isolate the four central components making up a
the introduction of mechatronic principles, a focused design controlled mechanical system, Fig. 2:
methodology will be followed so that conceptualization and Computation: At the center;
documentation of mechatronic software can be done indepen- Actuation: Controlling power delivery;
dent of the medium used for its implementation. Instrumentation: What's happening;
Target System: What all the fuss is about!
A. A History of Increasing Complexity In addition, the peripheral elements connect the mechatronic
Technological history shows that there has always been a system to the world of its operators and managers:
striving for complexity in mechanical systems. The follow- Operator Integace: Normal and emergency;
ing list shows some examples of machines that show the Other: Communication with other systems, etc.
marvelous ingenuity of their inventors: Once isolated from the instruments on one side and the
(not strictly chronological!); actuators on the other, computation could be implemented
Watt governor; using the most effective computing medium, independent of
Jacquard loom; any needs to pass power through the computational element.
typewriter; That medium has been the digital computer, and the medium
mechanical adding machine; of expression for digital computers is software.
pneumatic process controller; This ability to isolate is recent. Watt's speed governor, for
carburetor; example, combined the instrument, computation and actuation
sewing machine; into the flyball element. Its computational capability was
tracer lathe; severely limited by the necessity that it pass power from the
cam grinder. power shaft back to the steam valve. Other such examples,
A diagram of the Watt steam engine speed control (gover- where computation is tied to measurement and/or actuation,
nor) is shown in Fig. 1. As the engine speeds up, the flyballs include automotive carburetors, mastered cam grinders, tracer
AUSLANDER: WHAT IS MECHATRONICS?

Fig. 3. Operational amplifier; computing configuration.

4
lathes, DC motor commutation, timing belts used in a va- Fig. 4. Operational amplifier; follower configuration.
riety of machines to coordinate motions, rapid transit car
couplings (which are only present to maintain distance; cars Grindstone
are individually powered), and myriads of machines that
use linkages, gears, cams, etc., to produce desired motions.
Many such systems are being redesigned with software based
controllers to maintain these relationships, with associated
/"'\
improvements in performance, productivity (due to much
shorter times needed to change products), and reliability.

C. Amplijiers and Isolation

The development of electronic amplification techniques
opened the door to mechatronics. Operational amplifier (op- Fig. 5. Masterless cam grinder.
amp) based circuits, such as that shown in Fig. 3, could provide
both isolation and generation of a wide variety of linear and This is an excellent example of how mechatronics can
nonlinear, static and dynamic computing functions. change the way thlat problems are approached. In a conven-
Fig. 4 shows an alternate operational amplifier configu- tional, mastered, cam grinder the motion is controlled by
ration, which provides even stronger isolation, but cannot a mechanical system which traces the contour of a master
perform as flexible computation. cam. The mechanical tracing mechanism must fill the role
Developments in power electronics provided the same kind of computation (transmission of the desired contour) but also
of flexibility and isolation on the actuation side that op-amps carry enough power to operate the follower mechanism. This
provide on the signal side, and gave a means for translating is a limit to performance, and causes considerable loss of
commands computed in electronic form into action. productivity since the time to change the machine from one
Digital electronics added enormously to the range of com- cam profile to another is considerable.
putational functions that could be realized. Digital electronics In the mechatronic-based cam grinder, the profile is stored as
also formed the basic technology for implementing computers, an array of numbers in software, so the machine's performance
but with the use of computers it is software rather than is limited only by its inherent mechanical properties and the
electronics that represents the medium of expression. Although ability of the control computer to store and process the data.
mechatronics based on electronic computing elements opened Changeover is also much faster since it only requires loading
a new world of mechanical system control, software has of new data.
extended that world in unimaginable ways. Because this The real-time software used in most mechanical system
software must operate within timing and response constraints control is also safety-critical. Software malfunction can re-
imposed by the system under control, it is known as real sult in serious injury and/or significant property damage. In
time software and, by the present definition, is at the heart discussing software-related accidents which resulted in deaths
of mechatronic systems. and serious injuries from clinical use of a radiation therapy
machine (Therac-25), [4] established a set of software design
principles, ". . . that apparently were violated with the Therac-
V. REAL-TIMESOFTWARE 25 . . ." Those are the following.
Real-time software differs from conventional software in Documentation should not be an afterthought.
that its results must not only be numerically and logically Software quality assurance practices and standards should
correct, they must also be delivered at the correct time. A be established.
design corollary following from this definition, is that real time Designs should be kept simple.
software must embody the concept of duration, which, again, Ways to get information about errors-for example, soft-
is not part of conventional software. An example showing the ware audit trails-should be designed into the software
need of on-time delivery of results is given in Fig. 5. In the from the beginning.
masterless cam grinder the rotation of cam must be precisely The software should be subjected to extensive testing and
coordinated with transverse motion to cut the noncircular cam formal analysis at the module and software level; system
profile. testing alone is not adequate.
8 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 1, NO. 1, MARCH 1996

In particular, it was determined that a number of these prob- intertwined with the computation aspects. This situation relates
lems were associated with asynchronous operations, which directly to the first design principle listed above, “documenta-
while uncommon in conventional software, are the heart and tion should not be an afterthought.” If the system engineering
soul of real-time software. Asynchronous operations arise from is separated from the computational technology, then the
preemptive, prioritized execution of software modules, and documentation will have to exist independent of the program;
from the interaction of the software with the physical system otherwise, documentation can be viewed as an additional
under control. burden.
Because of preemptive execution, it becomes impossible The following definitions will be used to separate these roles
to predict when groups of program statements will execute as they appear in the proposed methodology:
relative to other groups, as it is in synchronous software. Errors System engineering: Detailed specification of the relation-
that depend on particular execution order will only show up ship between the control software and the mechanical sys-
on a statistical basis, not predictably. Thus, the technique of tem.
repeating execution until the source of an error is found, which Computational technology: A combination of computa-
is so effective for synchronous (conventional) software, will tional hardware and system software that enables application
not work with for this class of real time error. software based on the engineering specification to meet its
In a like manner, there are errors that depend on the operational specifications.
coincidence of certain sections of program code and events Using these definitions, the engineering specification de-
taking place in the physical system. Again, because these scribes how the system works; the computational specification
systems are not strictly synchronized, they can only have determines its peformance. As a result of this separation, if
statistical characterization. a suitable paradigm is adopted to describe the engineering
specification, a much broader discussion and examination can
A. Nasty Software Properties be brought to bear because, the details of the engineering can
Software complexity tends to increase exponentially with be discussed by project participants familiar with the problem,
size. What starts out as a small, manageable project, can ex- not just those familiar with computing languages and real
pand into something which, quite literally, never gets finished. time programming conventions. This provides for meaningful
To compound this problem, the economics of software based design review of projects that are software intensive.
products is that over the lifetime of the product, maintenance
costs dominate the total expense.
Both the complexity and the asynchronous nature of real- C. Software Portability
time software lead to situations in which bugs can remain
In mechanical system control, portability has consequences
dormant for long time periods, waiting to do their dirty work
both for product lifetime and for the desigddevelopment cycle.
at the most inopportune time.
The mechanical part of the system can have a commercial
The antidotes to these problems flow from adopting an
lifetime of anywhere from 5-20 years. On the other hand,
engineering perspective so that code writing (programming)
the computational technology used for its control only has a
should not be viewed as a creative activity!
lifetime of 3-5 years. To remain competitive, new versions of
On the engineering side, a method must be found to describe
the product need to use new computers to take advantage of
in detail how the control software should behave. This allows
the ever-increasing computational capability. Doing this cost-
both for the needed engineering review function and enables
effectively requires software that will “port” easily from the
the act of code production to become much more routine since
the creative engineering part has been done at the descriptive original target processor to new ones.
Software portability seriously affects the de-
level.
The second way of avoiding at least some of the complexity sigdimplementation cycle as well. Early stages of the
software tend to be simulations, done to test hypotheses and
problems is to modularize the software and its production
to substitute for hardware not yet built. Later stages use
process. In this way, specific pieces of software are directly
laboratory prototypes, then pilot prototypes, then, finally,
connected to certain machine functionality. System malfunc-
tions can thus be traced to the relevant software more quickly, the actual production system. If software can’t migrate from
step-to-step in this process, the whole process can become
system changes can be made with reasonable assurances that
greatly elongated as new software must be created for each
the changes will not adversely affect some unexpected part of
step, and there are significant possibilities for introducing
the machine, and individual software elements can be tested
new bugs at each stage.
more easily.
Portability is complicated by the real-time constraints. If
real-time software environments are used as aids in meet-
B. Engineering DesignKornputational Perj4ormance ing those constraints (kernels, schedulers, real-time operating
Too often, the only detailed documentation of real-time systems), software written for one environment can require
software is the program itself. Furthermore, the program is substantial rewriting to run in another. Crossing the full spec-
usually designed and written for a specific real-time target trum from simulation to production traverses environments in
environment. The unfortunate consequence of this is that the which program structure itself is incompatible. The proposed
engineering aspects of the design problem become inextricably methodology provides a layer of abstraction one higher than
AUSLANDER: WHAT IS MECHATRONICS? 9

the real-time environments, so offers a means of bypassing The organization into processes and threads is purely for
these incompatibility problems. performance purposes. As is noted below, there is no theoret-
Further portability challenges arise from the operator inter- ical necessity for such organization at all. The organization
face section of the program, and the interprogram communica- serves only to meet performance specifications when the
tion for those solutions implemented with multiple processors. chosen processor is not sufficiently fast to meet the specs
with a single-thread structure. It does this by shifting processor
D. Control System Organization resources from low-priority to high-priority tasks. As a further
A two-level organization is used on both the engineering note, the portability requirements enumerated above imply that
and computational sides of the control software design. On several different computational organizations will be used in
the engineering side, a job is organized into the course of a development project.

tasks VI. CONCLUSION

states I have presented an approach to the definition of what
and, on the computational side, into constitutes the field of mechatronics and an approach to the
design of systems based on that definition. The purpose for
processes doing this is to establish the central skills and experience that
threads. are expected of those who are or wish to be mechatronics
specialists. If that is not done, mechatronics will remain a
The breakdown of the engineering specification into tasks
loosely defined “mterdisciplinary” area which will greatly
and then states is a subjective process, requiring considerable hamper attempts to move the field forward in a reasonably
engineering judgment. This stage is the primary creative design orderly way and will make it difficult to define educational
step.
paths that will lead to this specialty. I expect that not all
Tasks represent units of work; they are the functional sub-
readers share this vision of mechatronics but I hope that
units of the job. Tasks can often be designed by breaking a
this article might lead to enough further discussion so that
system down along both time scales and physical components. a consensus definition will emerge-perhaps in the pages of
For example, there might be separate tasks to handle the high- new journal.
level part of controls for several motor-driven axes, and other
tasks to handle the higher-speed operations associated with REFERENCES
measurement and actuation of those same motors.
Tasks, in general, are active simultaneously. They are used D. M. Auslander and P. Sagues, Microprocessors for Measurement and
Control. Berkeley, CA: OsborneiMcGraw-Hill, 1981.
to describe the parallelism inherent in the physical system. D. A. Dornfeld, I).M. Auslander, and P. Sagues, “Programming and op-
Internally tasks are organized in a strictly sequential manner timization of mull i-microprocessor controlled manufacturing processes,”
into states [l], [2], [ 5 ] , and [ 6 ] . States describe specific Mech. Eng., vol. 102, no. 13, pp. 34-41, 1980.
W. W. Gibb, “Software’s chronic crisis,” Sei. Amer., pp. 8 6 9 5 , Sept.
activities within the task; only one state can be active at a 1994.
time. The active state controls the current activity of the task. N. G. Leveson and C. S. Turner, “An investigation of the Therac-25
accidents,” IEEE Computer, July 1993.
The primary reason for this distinction between the nature M. J. Simms, “Using state machines as a design and coding tool,”
of tasks and states is that sequential activity is a part of many Hewlett-Packard J., vol. 45, no. 6, Dec. 1994.
mechanical systems. Even for systems that do not seem to be P. T. Ward and S . J. Mellor, Structured Development for Real-Time
Systems. Englewood Cliffs, NJ: Prentice-Hall, 1985.
fundamentally sequential, such as process or web systems, the
sequential breakdown within tasks seems to work quite well.
During normal operation, tasks tend to stay in only one or two
states. However, during start-up and shutdown, the taskhtate
structure describes the operation very well.
The sequential state structure also serves as one test for David M. Auslander completed undergraduate
appropriate task definition. If tasks are aggregated to much, studies at the Cooper Union and graduate studies at
they will end up with parallel operations which cannot be MIT, both in mechanical engineenng.
He is currently a Professor of Mechanical
described effectively by states. Breaking the task into several Engineenng at the University of California at
smaller tasks solves the problem. Berkeley, and has interests in dynamic systems and
control. His research and teaching interests include
On the computational side, processes describe computa- bloengineenng, mechanical control, and real-time
tional entities that do not share an address space (this includes software. Current projects in these areas are cardiac
independent processors). Threads are computational entities signal analysis, nanometer mechanical positioning,
design methodology for real-time control software,
that share an address space but can execute asynchronously. and control of bioregenerative life-support systems He conwlts in industrial
A thread can contain one or more tasks; a process can contain servo control systems and other control and computer applications.
one or more threads. Dr. Auslander is a member of ASME.