Research on Robot Neural Network Motion Model
ZHANG Mingjun (College of Mechanical Electricity, Harbin Engineering University, Harbin 150001)
In this paper a theoretical research on neural network motion model of Autonomous Underwater Vehicle ( AUV ) is carried out. A not-all recurrent neural network is proposed to show the effect of series time variable. The closed neural network from output layer to input layer is given to realize its motion simulating capability. The stepperty study method is proposed also. In order to solving the study problem, integral layer was added into the output layer structure, and the method of variable standardization was designed. The result of computer-based simulations and experiment by AUV"PW45 ", proves that motion model network and performance method presented in this paper are reasonable and feasible.