Guided neural network and its application to longitudinal dynamics identification of a vehicle

In this paper, a modified neural network approach called the Guided Neural Network is proposed for the longitudinal dynamics identification of a vehic le using the well-known gradient descent algorithm. The main contribution o f this paper is to take account of the known information about tho system i n identification and to enhance the convergence of the identification error s. In this approach, the identification is performed in two stages. First, the Guiding Network is utilized to obtain an approximate dynamic characteri stics from the known information such as nonlinear models or expert's exper iences. Then the errors between the plant and Guiding Network are compensat ed using the Compensating Network with thy gradient descent algorithm. With this approach, the convergence speed of the identification error can be en hanced and noire accurate dynamic model can be obtained. The proposed appro ach is applied to the longitudinal dynamics identification uf ii vehicle an d the resultant performance enhancement is given.