Page 414 - Biomimetics : Biologically Inspired Technologies

Bar-Cohen : Biomimetics: Biologically Inspired Technologies DK3163_c016 Final Proof page 400 21.9.2005 11:49pm

400 Biomimetics: Biologically Inspired Technologies

unify the interpretation of living organisms and machines from the viewpoint of organized systems.
Several theoretical concepts have been evolved in control theory, typified by feedback control,
optimal control, sequence control, and so on. The main roles of feedback control are regulation
and adjustment, whereas optimal control involves planning and supervision with a higher level
of control state than feedback control. Sequence control, on the other hand, has the objective of
rationalizing logical procedures, scheduling, and decision.
Basically, the objective of feedback control is to follow up a desired command input (set point)
by controlled variables subjected to unpredictable external disturbance. For the constant set point,
the control problem is referred to as a constant value control (regulator problem), and if the
command changes over time, it is referred to as a follow-up control (servomechanism problem).
The linear feedback control theory using transfer functions has already reached its maturity in the
1940s. The similarity of biological and mechanical systems in terms of homeostasis was considered
within the framework of feedback control in that time. For example, tremor can be seen in many
neurological disorders, their hand trembles when they attempt to grasp an object. Some tremors
may be related to feedback instabilities. Since the feedback control is used to suppress disturbances
and to maintain a set point in constant value control, it corresponds to homeostasis in living
organisms.
In the 1960s, modern control theory based on the foundation of state space methods has been
established. This leads to the rapid appearance and development of the pole assignment principle,
observer theory, optimal regulator, Kalman filter, the internal model principle, and their extension
to multiple-input–multiple-output systems. While the modern control theory contributed to the
refinement of the linear feedback theory, an even more important development was the clarification
of the principles related to duality of control and observation. Specifically, Kalman (1960)
introduced the concepts of controllability and observability and demonstrated the duality of control
and observation for the linear systems. He showed that it is possible to construct the pole-
assignment of a closed-loop system and the state-estimation of a controlled system in the same
framework. Furthermore, optimal regulator and Kalman filter can also be constructed in the
same way. Therefore, if a system (the controlled system) is controllable and observable, then it is
possible to construct any kind of desired dynamics (poles) for a closed-loop system.
Although the similarity of biological and mechanical systems was studied in terms of feedback
control, often the subject was limited to the range in the vicinity of operating point in which the
linear theory holds. In the 1980s, the control design theory has returned to the frequency response
methods and robust control theories based on H1 norm were developed (Doyle et al., 1989). H1
control theory makes it possible to quantitatively handle the influence from the variations of the
controlled object’s dynamic characteristics, which had been previously treated only qualitatively.
However, the control concept remains unchanged within feedback control.
On the other hand, adaptive control is a more direct strategy than robust control to handle
the influence of the object’s variations. In robust control, the controller itself does not change
even if the controlled object was fluctuated. However, adaptive control has the function of
identification to monitor the fluctuation in the controlled object, and based on the result of
identification, the controller is adjusted. Accordingly, identification and adjustment are the two
basic functions of adaptive control. Recently, robust adaptive control has also been developed
to increase the system’s robustness against the fluctuations of the controlled object. Adaptive
control system can be regarded as a nonlinear system in the sense that the compensation is adjusted.
However, since the controlled system and the compensation element are simply linear, and the
purposes of the control are still limited to suppress disturbances and to track command input as in
feedback control, present adaptive control does not have the ability to self-organize the system’s
internal states with respect to the environmental dynamic variations as seen in biological systems. It
left far from the goal of automatic system control design. The meaning of adaptation in the adaptive
control systems that has been developed so far is still quite different from the one featured in living
systems.