A connectionist approach to PID autotuning

A connectionist approach to the problem of PID autotuning is proposed. Multilayer perceptrons, based on integral measures of the step response, are used to supply the PID parameters. Tuning can be achieved both in open loop and closed loop. A different formulation of the learning problem for a class of multilayer perceptrons is presented. This approach, besides reducing the size of the problem, gives a major reduction in the number of iterations needed to achieve a local minimum.<
>