RESUMO

The gradient method of solution of quadric programming problem under parametric uncertainty and with measurement of the minimized function with bounded noise is considered. It is shown, that convergence of gradient minimization procedure, i.e., existence of a bounded minimal invariant set, takes place only when uncertainty in a priori parameter estimation is within some bounds. For preservation of convergence of iterative procedure when parametric uncertainty increases, use of "direct" adaptive control is suggested.