Modeling and optical control of a multi-link robot manipulator using model-reality algorithm

Abstract

In this chapter we consider a multilink robot manipulator problem with two-link manipulator arm. The perturbation model is expressed as nonlinear ?rst order “real” model which exhibit a high degree of nonlinearity. When a linearized model is used to solve the optimal control it creates large modelreality di?erences, which cannot be ignored. An algorithm that takes account of model-reality di?erences is developed and utilized to ?nd the optimal control of the robot manipulator. The algorithm uses linear quadratic (LQR) problem as the model. The high degree of nonlinearity makes the linearized model very unstable, which inevitably causes divergence. The large di?erences between model and reality on the other hand cause slow rate of convergence to the algorithm. In this chapter we present preliminary ?ndings on one method to overcome this slow convergence which is the selection of proper weighting matrix to the LQR.