Extremum or hill-climbing regulation: a statistical theory involving lags, disturbances and noise

An example of a perturbation-type extremum controller

“…:i t is the current expected value of x, and o, is an information-seeking perturbation. Some authors have used this technique to derive observations of the gradient of c/J, which can be more legitimately approximated by a linear function; see Box and Chanmugam (1962) and Roberts (1965). Here the observations will be used more directly.…”