Neuro-sliding mode controllers for systems with uncertainties

Official URL: http://risc01.sabanciuniv.edu/record=b1123085 (Table of Contents)

A Neuro - Sliding Mode Controller was designed for systems that have uncertainties like unknown external disturbances and unknown system parameters. First, the controller was designed for single input single output (SISO) systems and then it was generalized for a certain class of multi input multi output systems. Stability proof was given using Lyapunov Stability Criteria and finally, the theory was supported by simulation and experimental results. The Neuro - Sliding Mode Controller proposed in this thesis consists of a one layered neural network whose activation functions are linear. The main working principle of the controller is minimizing a cost function which is determined from the requirements of the Lyapunov Stability Criteria and Sliding Mode Control Theory. The major contribution of this work is that, different from the similar works in the field, the stability of the overall control system was shown by analyzing the properties of the cost function introduced to the neural network for minimization. Two different experimental setups were used for SISO and MEVIO cases respectively. For the SISO case, the position of an electrical motor that actuates a linear servo - drive was controlled. For the MEVIO case, a system consisting of two piezoelectric actuators connected to each other via a load cell, which was used for force measurement, was used. In this system, the position of one actuator and the internal force created were controlled simultaneously. Both experiments were successful and supported the theory.