Hot rolling is a competitive business in which quality and efficiency are of paramount importance. Methods for improving the accuracy and precision of the product can yield significant savings for the steel producer. Automatic control of the process is now widespread, and the inclusion of adaptive techniques to control systems has shown notable improvements in meeting tightening tolerances and reducing scrap. The paper describes some of the techniques used for on-line identification and control during the hot rolling of steel. A general adaptive control scheme is described which is applicable to the majority of rolling mills, and practical guidelines for success are outlined. Two case studies are included : the first is the application of extended Kalman filtering techniques to pass-by-pass adaptation within a section mill; the second is the application of similar techniques to feedforward adaptive control of a hot strip mill finishing train.