Automatic driving algorithm training method and device

The invention relates to an automatic driving algorithm training method and device. The method comprises the following steps: generating a road simulation environment based on test road information; constructing a multi-scene driving data set according to the historical traffic scene data by using a preset orthogonal arrangement method; according to the multi-scene driving data set, an automatic driving algorithm is optimized in a road simulation environment, an optimized driving algorithm is obtained, and space-time distribution characteristics of traffic participants and simulation track data of a vehicle are obtained; and projecting the spatial and temporal distribution characteristics of the traffic participants to a test road in real time, obtaining real trajectory data of the vehicle, and training an optimized driving algorithm according to the real trajectory data and the simulation trajectory data to obtain a fully trained automatic driving algorithm. The method enables the automatic driving algorithm