Knowledge based approach for online self-tuning of PID-control

We present results pertaining to the online adaptation mechanism of a class of linear controllers using tools of soft computing. These are implemented through the readily available linguistic knowledge acquired a priori about the system and its behavior along with the learning achieved in an online and off-line stages of the control design. This is applied to a benchmark experimental model taken here as a DC motor subject to wide range of varying load parameters and external disturbances. The results obtained using an integrated scheme of the optimized Takagi Sugeno scheme are compared with those of a dynamical neural network based scheme. It is shown that significant improvements could be made over the conventional static PID-controller, in particular for load disturbance recovery