Page 17 - Handbook of Biomechatronics

10 Ahmed R. Arshi

functions require mastery, fluency, and command over complex interacting
biochemical, biomagnetic, bioelectrical, heat and mass transfer, biofluid
dynamics, and movement biomechanics. Tissue biomechanics in conjunc-
tionwithneuromusculoskeletaldescriptionsarerequiredattimestoallowfull
investigations of the manipulation and locomotion while a large set of data is
being processed to implement any control strategies by the central nervous
system.
There are two basic approaches to modeling in biomedical engineering.
The first utilizes classical disciplinary mathematical modeling where a
description of a combination of function and structure are produced to sim-
ulate the system. The second approach is in favor of looking at the physio-
logical systems as a black box and various algorithms such as neural networks
are adopted to learn the dynamics of the system. These two, often conflicting
modes of thought, should in biomechatronics be considered as two sides of
the same coin. The importance of constructional modeling cannot be over
emphasized as the current applications of such intelligent algorithms or soft
computing in design of biomechatronic systems is in need of further devel-
opment. The black box approach, however, can be used effectively in design
of the control strategies. The fundamental problem with the current knowl-
edge of human physiology is that although a vast array of knowledge is con-
stantly being produced by biological, physiological, or electrophysiological
laboratories, there still is a wealth of knowledge to be gained so that the exis-
ting gaps are covered. Furthermore, the current mathematical tools used in
modeling also require further developments. The continuous advancements
of microprocessors are reaching the state where principles of predictive con-
troller could be revisited so that real-time simulation results could predict
immediate necessary responses of the biomechatronic system in daily inter-
action of human subject with his/her environment. Here, the mentality of a
generalized mathematical model could shift toward tailored solutions. Tai-
lored biomechatronic systems require individualized and personalized
models of the system which could in turn play an important role in control
strategy.
Furthermore, problems such as intent are increasingly recognized as
high-level cost functions against which standard neurophysiologically
obtained parameters do not necessarily lead to suitable models. Intent rec-
ognition could require real-time integration and processing of a multitude of
sensory inputs. Modeling of such complex systems require an alternative but
reliable technology. Bond graph technology could provide a measurable
solution to modeling and design problems.