Lane Decision Algorithm for Active Avoidance of Intelligent Vehicle Based on Improved Back Propagation Neural Network

Abstract

In view of the traditional intelligent vehicle lane decision algorithm is lack of flexibility, and slow convergence speed of traditional back propagation neural network algorithm, the training time is long, easy to fall into local minimum values and without guiding network structure theory, by studying the traditional improving methods of back propagation neural network algorithm, introducing auxiliary weights adjustment parameters and contraction coefficient, abate sawtooth phenomenon, speed up the convergence speed and reduce the training time, and to some extent, improve the accuracy of intelligent vehicle lane decision for active avoidance. Through the synthetic judging three different lanes static target decision, compare the improved back propagation algorithm with the traditional algorithm in the actual lane decision algorithm for active avoidance accuracy and convergence time.