Powers [ 7 ] proposes an inherent principle of all living systems that explains and unifies perception and behaviour. In contrast to the more conventional approach of open-loop processing where some form of processing is performed on an input to produce a specific output, Powers' scheme is a closed-loop, continuous, feedback control system where the function of behaviour (non-specific output) is to control a perception (specific input). An outline of the basic functional architecture of Perceptual Control Theory (PCT) is shown in figure 1 a. A reference signal ( r ) represents the goal perception of the system, the perceptual signal ( p ) represents the current input and the error signal ( e ) the difference between the two. The output is a function of the error signal. The perceptual signal is a function of the environment and the effects of the output of the control system.
A graphic example of PCT is demonstrated by the everyday activity of driving a car. The most important aspect of driving is to keep the car in the correct lane in order to avoid accidents. In PCT terms this is achieved simply by adjusting the steering wheel (output) according to a perceived position (input) of the car. The amount the wheel is turned does not need to be specifically calculated, the wheel is simply turned until the car is perceived in the appropriate position. This method will effortlessly overcome any disturbances to the system whether they be due to internal changes or external influences, such as a wind acting upon the car. Contrast this with a non-control system, which could be called a measure and model approach. As with PCT it is necessary to represent in some way how far away the car is from a desired position, but it is now also necessary to relate that error to a specific amount that the steering wheel needs to be turned in order to bring the car back on course. Obviously, this requires precise calibration of the relationship between not only the steering and driving wheels, but also between the driver and the steering wheel. A wind force would influence the car's position so this would need to be measured and its effects modelled and compensated. Many other factors can also affect how the car behaves such as tyre pressure, road surface and engine condition, all of which would need to be measured and modelled in an open-loop approach that requires the computation of a specific output. PCT greatly simplifies the situation by controlling the input and dispensing with the need for measuring and modelling the environment.
Rupert J Young