Comparison of Neural Networks Based Direct Inverse Control Systems for a Double Propeller Boat Model

Abstract

This paper presents the thorough evaluation and analysis on the direct inverse neural networks based controller systems for a double-propeller boat model. Two direct inverse controller systems that were designed with and without feedback were implemented on a double propeller boat model using two neural networks based control approaches, namely the back-propagation based neural controller (BPNN-controller) and the self-organizing maps based neural controller (SOM-controller). Then, the resulted control errors of the systems were compared. Simulation results revealed that the direct inverse control without feedback produced lower error compared to the direct inverse control with feedback. Another important finding from the study was that the SOM-controller is superior to the BPNN-controller in terms of control error and training computational cost.