USING FORWARD AND INVERSE NEURAL MODELS FOR SOLVING OPTIMAL TRACKING PROBLEM OF NON-LINEAR SYSTEM

Anna Jadlovská

   This paper presents the use of the theory of optimal control and linear- quadratic approach to the deterministic controller design for non-linear dynamical system - robot. In the first stage the optimal target trajectories are off-line determined that serve as the required inputs for the feedback tracking control of the robot, whereby the feedback is determined from the linearized robot model. The results of this classical approach are used to design an optimal tracking neuro-controller (OTNC) with quadratic cost function. The proposed feedforward control scheme with neural models will be demonstrated on a typical non-linear system, a two-link robot. Computer simulations of the control algorithm are made by the simulation language Matlab and the presented results are analyzed.

Keywords: programmed optimal control, two-bounded problem, sensitive function, linear-quadratic control, tracking problem, Riccati equations, forward neural model, inverse neural model

[full-paper]