Abstract: In a process and a system for automatic, computer-system-supported optimization and the search for a state x dependent on at least one parameter P, for which f(x) as a measure of the quality of x assumes an extremum, an initial solution, i.e. a state x, is found as an initial state and a further state y different from x is found by means of an elementary change in an individual parameter P from x. If the result of a comparison of the target function values f(x) and f(y) is that f(y) is poorer than f(x) by more than a threshold value T, y is rejected, x is retained and a fresh adjacent state of x is found. If f(y) is at least as great as f(x)-T, there is a transition to state y. The threshold T is successively reduced to zero.