On the structure and the evolutionary computation of H∞/L1 sub-optimal robust controllers for nonlinear systems

This paper addresses the problem of the automatic computation of robust controllers for nonlinear systems inspired at mixed H_∞/L¹ optimal control criteria. As a matter of fact, seeking the optimal solution of the standard H_∞ control problem, for some general classes of systems, yields "very degenerative" controllers. In order to overcome this limitation, an L¹ optimum criterion is added to the problem. It is introduced a class of neural networks all consisting of positive definite and proper functions, together with the description of an evolutionary numerical strategy specifically designed for the actual computation of suboptimal solutions of the problem outlined.