Neural network learning and control of uncertain robot systems with guaranteed stability

Abstract

This thesis considers neural network learning and control of uncertain robot systems. In robotics areas, nonlinear control schemes such as adaptive control and robust control have been proposed so far in order to overcome the structured and/or unstructured uncertainty of robot systems. However, such control schemes have a critical drawback that they can be applied only in the case of the known dynamic model. Therefore, the neural network control scheme, which is leading toward intelligent control, has been exploited in robot systems control. The generalization and learning capability of the neural network have been widely applied in the areas of control systems and have accomplished outstanding research results. However, researches so far have been directed mainly to the empirical study rather than to the theoretical one. For example, there have been few results that guarantee the closed-loop system stability using the neural network controller with learning. Even though exist, they are for the restricted cases only. Thus, this thesis presents control schemes and learning rules using neural networks that can guarantee the closed-loop stability. The closed-loop stability is proved using the Lyapunov analysis and learning rules are obtained by minimizing the cost function which are chosen for the neural network to learn the system uncertainty. This thesis is organized with two main parts: Part I is on the trajectory tracking control of the robot manipulators and Part II is on the position and force control. Based on the introduction on the neural network control schemes in Chapter 1, we present in Chapter 2 a neural network control scheme by learning the inverse dynamics of the robot systems. The neural network control structure and learning rule derivation are the main part of this chapter. Chapter 3 deals how to guarantee the closed-loop system stability when the neural network is incorporated in control. Two kinds of stability- guaranteeing schemes are propose...