Global path planning method for smart driving cars based on extension neural network

Abstract

Current global path planning algorithms for smart driving cars are capable of effective path type output, but some of them are expensive to develop and are not conducive to maintenance and iteration. In this paper, based on an extension neural network (ENN), the basic-elements model in Extenics is used to construct the information-elements involved in global path planning for smart driving cars as the input of the neurons; the maximum and minimum values of the quantity values of the input information-elements are used as the double weight values, and the midpoint of the interval formed by them is used as the midpoint of weight value interval, and the initial output function of the extension neurons is constructed based on the midpoint extension distance; The input and output layers are connected by double weight values to build the overall structure of the ENN for global path planning of smart driving cars. The ENN model is trained by collecting relevant data, adjusting the weight values and midpoints of weight value intervals of the ENN for incorrect predictions, and replacing the initial weight values and midpoints of weight value intervals to form a new output function. The trained model can predict path types quickly and accurately, showing that ENN can be used to optimize the efficiency and accuracy of the global path planning algorithm for smart driving cars in terms of output path types.