Development of Self-Organizing Maps Neural Networks Based Control System for a Boat Model

Abstract

This paper describes the development of a controller system for a developed double-propeller boat model using the unsupervised learning neural networks, namely the Self-Organizing Maps (SOM). The performance characteristics of the proposed SOM-based controller are then compared with that of the well-known Back-propagation Neural Networks (BPNN)-based controller through a direct inverse control scheme. Experimental results showed that the SOM-based controller can produce a low error, even lower than that of the widely used BPNN-based controller. Furthermore, the computational cost of the SOM-based controller is found to be more than 700 times faster than that of the BPNN-based controller. These findings suggest that the utilization of the proposed SOM-based controller for the control of a boat is highly effective.