Intelligent car end-to-end decision making method based on space-time joint recurrent neural network

The invention discloses an intelligent car end-to-end decision making method based on the space-time joint recurrent neural network. The method comprises steps of establishing the space-time constraint neural network, establishing a space-time joint recurrent neural network training model and testing the space-time joint recurrent neural network model, from perspectives of space position constraints and time context constraints, the convolutional neural network is utilized to extract space position characteristics in the scene, the LSTMs network is utilized to capture time context characteristics in the scene, a framework of the space-time joint constraint network is constructed, the decision content is directly calculated based on an input image, the cognitive process is unified into thedecision making process, the method of simultaneously optimizing all processes can achieve better performance and simplify the system structure, and a steering wheel corner value can be accurately predicted. 本发明公开了种基于时空联合递归神经