On limits of performance analysis and what can we do about it?

A cursory look at the programs of recent CDC or IFAC meetings reveals the fact that the control system community is searching for a new paradigm. It is healthy and invigorating as we search for new domains to conquer. However, we are reminded of the dictum “In time of stress, people revert to the tools they are comfortable with”, i.e., it is easy and natural for us to attempt to use the same real variable and differential equation based tools to attack these new problems. The purpose of this note is (i) to bring up some unpleasant and no doubt controversial (to some) facts concerning the limitations on how far we can push these tools with respect to the nature of the new problems we wish to solve, and (ii) to speculate that a softened approach involving computational intelligence such as fuzzy logic, neural networks, genetic algorithms, and ordinal optimization has a place in our arsenal of tools. It is the author´s belief after years subscribing to the same standard that at least an open minded discussion on these issues in our community is desirable for the benefit of the general readers. After all, one of the professed goals of the society is in solving real world problems. The fundamental limitations described in the paper with respect to the nature of these problems force us to rethink our paradigm