Advanced control techniques based in artificial intelligence for robotics manipulators

Abstract

The performance quality in nonlinear model based control of mechanical manipulators is conditioned to the reliability of the mathematical model and precision in the knowledge of all the involved parameters. Control methods based on artificial intelligence techniques (learning algorithms, system identification and neural networks) can be applied to improve its performance. A neural control scheme is proposed, consisting basically of a neural network for learning the robot inverse dynamics and online generating the control signal. Also an online supervision based on optimisation techniques is designed and implemented for such neural control. Simulation results are provided to evaluate the alternative variations to the proposed central scheme.