Vehicle trajectory prediction method based on lane point future trajectory offset auxiliary supervision

The invention discloses a vehicle trajectory prediction method based on lane point future trajectory offset auxiliary supervision. The method comprises the following steps: obtaining historical features and lane information of each surrounding agent; performing scene coding and feature fusion; and decoding the multi-modal trajectory and the offset, and constructing an auxiliary supervision branch. According to the method, a multi-scale lane topological relation is established through sampling and neighborhood extraction, and topological structure information and semantic information of a lane are coded by using a multi-scale lane point encoder of which features are diffused along a sparse aggregation direction and an anti-sparse direction; understanding on the internal topological structure and semantic attributes of the high-precision map under different feature scales is realized, so that the scene characterization capability of an encoder is improved, and higher-precision vehicle trajectory prediction can