Neural network identification and control of an underwater vehicle

摘要: Real-time predictive control requires a forward model that is both accurate and fast. This paper introduces two nonlinear internal memory network architectures compares their performance with Multi-layer Perceptron (MLP) augmented the use of spread encoding. The test plant single component from an Underwater Robotic Vehicle (URV), comprising thruster encased in steel frame provided buoyancy. assemblv free to move under water controlled for depth.The networks are comparable accuracy MLP but more parsimonious, resulting, faster response which makes them better suited on-line control. Although particular case study presented as focus this paper, algorithms methods developed have generic applicability.