Learning based speed planner for autonomous driving vehicles

A non-transitory computer-readable medium programmed with executable instructions that, when executed by a processing system, perform operations for efficient speed planning in an autonomous driving vehicle (ADV), the operations comprising:

• generating a grid of cells representing an area that encompasses a path that the ADV is to following in autonomous driving mode in view of one or more obstacles to the ADV on the path, wherein grid cells coinciding with the one or more obstacles are set to a first value, and all other grid cells are set to a second value;
• generating a feature vector, F′, from the grid of cells;
• selecting, from a plurality of feature vectors, a feature vector FMODEL that most closely corresponds to F′ and determining a speed model, WMODEL corresponding to the selected feature vector FMODEL;
• determining an initial speed v0 for navigating the ADV along the path in autonomous driving mode using feature vector F′ and speed model WMODEL;
• and
• navigating the path, by the ADV, in autonomous driving mode, at the initial speed of v0.

A system comprising a processing system having at least one hardware processor, the processing system coupled to a memory programmed with executable instructions that, when executed by the processing system perform operations for efficient speed planning in an autonomous driving vehicle (ADV), the operations comprising:

• generating a grid of cells representing an area that encompasses a path that the ADV is to following in autonomous driving mode in view of one or more obstacles to the ADV on the path, wherein grid cells coinciding with the one or more obstacles are set to a first value, and all other grid cells are set to a second value;
• generating a feature vector, F′, from the grid of cells;
• selecting, from a plurality of feature vectors, a feature vector FMODEL that most closely corresponds to F′ and determining a speed model, WMODEL corresponding to the selected feature vector FMODEL;
• determining an initial speed v0 for navigating the ADV along the path in autonomous driving mode using feature vector F′ and speed model WMODEL;
• and
• navigating the path, by the ADV, in autonomous driving mode, at the initial speed of v0.