Process control via artificial neural networks and learningautomata

摘要

The application of learning automata for control is shown to givesignificantly improved performance by two modifications (a) using a setof proportionality constants of a proportional feed-back controller, asthe action set of learning automata and (b) a scheme which usesartificial neural networks as memory. This work examines a classproblems where the objective function is reduced to inequalityconstraints and also where qualitative information from artificialneural networks can be used with learning automata as an effectivedecision maker. Two nonlinear process control examples are tackled. Itis shown that despite deterioration of net's prediction error thelearning automata-ANN control strategy works while the performanceimproves resulting a better accuracy