Self-tuning neurofuzzy control for nonlinear systems with offset
C.W. Chan, Xueshuo Liu, Wai-lan Yeung · 2002
A self-tuning neurofuzzy controller with an ability to remove offsets is derived, based on a self-tuning integrating controller derived for a local linear model. The training target for the proposed controllers is derived, and they can be trained by the simplified recursive least squares (RLS) method with a computing time that is linear instead of geometric in the number of weights in the network. Further, the simplified RLS method not only has the same convergence property as the RLS method, it also has a better ability in tracking varying parameters. The performance of the self-tuning neurofuzzy controller is illustrated by examples involving both linear and nonlinear systems.