by

requirements for the degree of OCEAN ENGINEER This thesis compares classical nonlinear control theoretic techniques with recently de-veloped neural network control methods based on the real time experimental control performance on a simple electro-mechanical system. The system has configuration-dependent inertia and velocity squared terms, which contribute a substantial nonlin-earity. The controllers being studied include PID control, sliding control, adaptive control, and Gaussian network control. Experimental results are given on the tracking performance and computation time for each controller. To establish a fair comparison among different controllers, the feedback bandwidth of each control method is chosen to be the same. The perfor-mance data for final comparison is collected using the same control frequency, i.e., the slowest pace of the four methods. These controllers are evaluated based on th