Application of a general learning algorithm to the control of robotic manipulators
International Journal of Robotics Research
Hi-index | 0.98 |
In this paper, some successful case studies are presented on the intelligent control of time-varying linear/nonlinear dynamical systems using the Cerebellar Model Arithmetic Computer (CMAC) artificial neural network, and a recently developed unified linear system theory and a novel control technique called Extended-Mean Assignment Control (EMAC). Thanks to CMAC's simple and effective training algorithm and fast learning convergence, satisfactory and encouraging simulation results of angle-of-attack control of an aerobreak re-entry spacecraft and attitude control of an earth orbiting space vehicle subject to gravity gradient torque are obtained and presented in this paper. Suggestions for further investigations on the control of time-varying dynamical systems using CMAC and EMAC are proposed.