An improved cerebellar model articulation controller based on the compound algorithms of credit assignment and optimized smoothness for a three-axis inertially stabilized platform

Abstract

An improved cerebellar model articulation controller (CMAC) based on the compound algorithms of credit assignment (CA) and optimized smoothness (OS) is proposed to improve the control performances of a three-axis inertially stabilized platform (ISP). On the basis of both advantages of CA and OS algorithms, the improved CMAC neural network controller can deal with the disadvantages of conventional CMAC in learning interference and output fluctuation. As a result, the main dynamic performances of the ISP control system are systematically promoted. To verify the method, the simulations and experiments are carried out respectively, in which the LuGre friction model is introduced to represent the main disturbances. The results show that the CA&OS-CMAC controller can efficiently restrain the nonlinear disturbances and obviously improve the ISP's pointing precision and output smoothness. Compared with the conventional CMAC controller, the root-mean-square (RMS) errors of tracking the angular position under the static base swinging and dynamic base leveling conditions are reduced by 17.17% and 30.55%, respectively.