MODEL-PREDICTIVE CONTROL - PAST, PRESENT AND FUTURE

Model predictive control was emerging into the process industries during the 1980's. Today it is commonplace and, in some sense, mature. Commercial packages are available on a variety of platforms and ''do it yourself'' implementations are also readily found. This paper takes a look at where this technology has been and where it is, and tries to predict important future developments. Some trends are clear but the range of possibilities is wide. There are many buzzwords that could describe aspects of future model predictive controllers: adaptive, non-linear, embedded, object oriented, open systems and many more. How these might fit into future model predictive controllers is explored in this paper.