Emotion-based control systems

Abstract

The relevance of the model presented to the control and the supervision of systems lies in the fact that, in this context, it is very important to respond quickly and efficiently to unexpected situations, by learning associations between current situations and control strategies. The inputs and the state variables of a system can be considered as stimuli to feed a double processing system. The cognitive image can be considered as the set of values collected in a time frame. On the other hand, the perceptual image can result from the determination of certain characteristics such as overshoot, rate of variation of state variables, and so on. The next step is to establish a basic set of associations in order to allow the system to respond to urgent situations (solely based on the perceptual image). As the supervisor starts marking cognitive images with perceptual ones (a basic mechanism of learning), it becomes able to anticipate those situations (this is what humans apparently do when using the somatic marker). On the other hand, the matching of a certain configuration with one previously stored in memory can be assessed in terms of the positiveness or negativeness of the present situation by consulting the cognitive/perceptual mark. The control and supervision of large scale, non-linear, and non time-invariant systems ought to incorporate planning and decision making mechanisms together with low-level controllers, integrated in such a way that performance (both in terms of learning, quality of response, and efficiency) is ensured.