METHOD AND APPARATUS FOR TRAINING CONTROL STRATEGY MODEL FOR GENERATING AUTOMATIC DRIVING STRATEGY

摘要

The present application provides a method and apparatus for training a control strategy model for generating an automatic driving strategy. The method comprises: firstly, determining, by using low-dimensional training data, a latent feature space and a strategy function defined in the latent feature space; then by taking the latent feature space as a target, supervising the training of an encoder that maps high-dimensional training data to the latent feature space; and then applying the encoder and the strategy function to a real traffic environment. That is, high-dimensional data acquired from the real traffic environment is input, so that an available automatic driving strategy can be directly obtained by using the high-dimensional data.