IDENTIFYING/SIMULATING METHOD FOR NONLINEAR DIRECTION CONTROL SYSTEM OF SMALL-CALIBER TUNNEL ROBOT

摘要

PURPOSE:To attain the identification of a nonlinear direction control system by forming the input of a neural network with a tip head angle, the past pitching angle change value of a robot main body, and the offset value, and learning the each inter-neuron weight based on the difference between the present pitching angle change value and the output of the neural network. CONSTITUTION:The input of a neural network 9 consists of the inputs of a plant 8 preceding by 1-sampling and 2-sampling operations, the offset value, and the outputs of the plant 8 preceding by 1-sampling and 2-sampling operations. Meanwhile the network 9 is leant based on an output error (the difference between the output of the plant 8 and the output of the network 9)epsilon. Thus a nonlinear direction control system can be identified by means of the network 9.