ニューラルネットワークによる潜水艇の運動の同定

概要

論文の詳細を見る
A structure of neural network which includes two kinds of recurrent connections, i. e., from the output layer to the input layer and from the hidden layer to the input layer, is proposed to represent the dynamics of a plant. The inputs consist of state variables of the plant and control signals, and the outputs are the state variables at the next time step. On the advantage of neural network, the property of the dynamic system with significant non-linearity can be represented by neural network based on learning of I/O data of the plant. A learning process to construct a network covering a wide area of the input domain is introduced and tested by applying this structure to the longitudinal motion of a test-bed vehile of crusing type underwater robots. It is shown that the system identification of high quality can be accomplished through the proposed simple scheme.