Adaptive Discrete Stochastic Controller: Application for fed-batch fermentation

For a class of industrial processes where functional discrete states can be defined, a novel control strategy is proposed. A process can be then modelled by a controlled Markov chain and relation of the discrete states to measurements by probability density functions. The described control strategy comprises a current state recognizer, a Markov chain parameter estimator, a probability density estimator, and a receding horizon controller based on the dynamic programming principle. Results of application of the controller to a fed batch fermentation are presented.