Trajectory optimization method and system based on differentiable safety constraints in autonomous driving

Abstract

The application discloses a trajectory optimization method and system based on differentiable safety constraints in automatic driving, and relates to the technical field of automatic driving motion planning and control; constraint conditions involved in trajectory optimization in the automatic driving process are constructed, an end-to-end neural network architecture is designed based on the constructed constraint conditions, the designed neural network architecture is trained in a dual optimization mode, open-loop simulation tests are carried out based on public data sets, and closed-loop road tests are carried out based on a real vehicle computing platform; the application adopts the trajectory optimization method and system based on differentiable safety constraints in automatic driving to improve the problems of insufficient safety of trajectory planning and difficulty in constraint execution of an existing end-to-end model in a complex scene, and effectively improves the compliance and comfort of the automatic driving system in actual application.

Images