Optimal control of iterative processes in problems of trajectory information processing

A condition for constructing optimal controls is derived, according to which the choice of a specific optimal control variant during the course of an iterative process is made automatically. If the initial approximation is only roughly specified and the partial derivatives are approximate, the motion can be in an off-gradient direction. As the solution is approached, the process generally enters into the region of steady convergence of gradient methods, and for physically determinate problems of trajectory information processing the process enters the region of stable convergence of Newton's method.