AUTOTUNING OF FEEDBACK GAINS USING A NEURAL-NETWORK FOR A SMALL TUNNELING ROBOT

This paper describes the autotuning of feedback gains for a small tunn eling robot. We have already proposed the directional control method w herein the head angle of the control input is the sum of the deviation multiplied by feedback gain K(p) and the angular deviation multiplied by feedback gain K(a). In this paper, we used a neural network to obt ain feedback gains K(p) and K(a). The input of the neural network is a n initial deviation and an initial angular deviation. The output of th e neural network is the feedback gains K(p) and K(a). This neural netw ork learns from the deviation errors. The neural network which can be applied to any initial deviation was formed by using plural initial de viations in learning. Moreover this method can tune optimum gains to a ny design line. These results showed the validity of the proposed auto tuning method.