Intelligent vehicle end-to-end decision-making method based on a spatial-temporal feature fusion recurrent neural network

The invention discloses an intelligent vehicle end-to-end decision-making method based on a spatial-temporal feature fusion recurrent neural network. The method comprises the steps of establishing a spatial-temporal feature fusion recurrent neural network, establishing a spatio-temporal feature fusion recurrent neural network training model; and testing the spatio-temporal feature fusion recurrentneural network model. Deep neural network, According to the feature fusion method, two or more different features can be fused; Facilitating network convergence, According to the invention, spatial-temporal feature addition is explored; spatio-temporal feature subtraction, the influence of four feature fusion methods of space-time feature multiplication and space-time feature cascading on the decision-making network is realized; an intelligent vehicle decision network experiment based on space-time feature fusion is designed to prove that a space-time feature addition method is superior to other three feature fusion