Optimal design of CMAC neural-network controller for robotmanipulators

摘要

This paper is concerned with the application of quadraticnoptimization for motion control to feedback control of robotic systemsnusing cerebellar model arithmetic computer (CMAC) neural networks.nExplicit solutions to the Hamilton-Jacobi-Bellman (H-J-B) equation fornoptimal control of robotic systems are found by solving an algebraicnRiccati equation. It is shown how the CMAC can cope with nonlinearitiesnthrough optimization with no preliminary off-line learning phasenrequired. The adaptive-learning algorithm is derived from Lyapunovnstability analysis, so that both system-tracking stability and errornconvergence can be guaranteed in the closed-loop system. Thenfiltered-tracking error or critic gain and the Lyapunov function for thennonlinear analysis are derived from the user input in terms of anspecified quadratic-performance index. Simulation results from antwo-link robot manipulator show the satisfactory performance of thenproposed control schemes even in the presence of large modelingnuncertainties and external disturbances