FIELD: automatic optimization of multiparameter objects of control based on certain criterion of optimality. SUBSTANCE: step in random direction from initial condition in space of optimized parameters is made in correspondence with uniform law of distribution. If value of quality function under new condition is larger or equal to value of quality function in starting point then system returns to initial state after which new random step counted from old state is formed anew. Adaptation of distribution of direction of random steps consists in change of their mathematical expectation based on sign of increment of quality function and adaptation of intensity of self-learning depends on absolute value of difference of derived quality function of optimized object under successful runs. EFFECT: increased speed of self-tuning. 1 dwg, 1 tbl
