US 12,472,976 B2
Multi-level optimization framework for behavior prediction in autonomous driving
Yu Cao, Sunnyvale, CA (US); and Ang Li, Sunnyvale, CA (US)
Assigned to APOLLO AUTONOMOUS DRIVING USA LLC, Sunnyvale, CA (US)
Filed by Apollo Autonomous Driving USA LLC, Sunnyvale, CA (US)
Filed on Dec. 22, 2022, as Appl. No. 18/145,344.
Prior Publication US 2024/0208533 A1, Jun. 27, 2024
Int. Cl. B60W 60/00 (2020.01); B60W 30/09 (2012.01); B60W 40/00 (2006.01); G06N 5/022 (2023.01)
CPC B60W 60/001 (2020.02) [B60W 30/09 (2013.01); B60W 40/00 (2013.01); G06N 5/022 (2013.01); B60W 2554/404 (2020.02)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
generating a first variant of a first machine learning (ML) model, the first variant associated with an initial hyperparameter value;
determining a prediction metric for the first variant of the first ML model, the prediction metric indicating an accuracy of a behavior prediction for the first variant of the first ML model;
generating an estimated simulation metric for the first variant of the first ML model by applying a second ML model to the prediction metric corresponding to the first variant of the first ML model; and
identifying a first hyperparameter associated with a second variant of the first ML model, the second variant of the first ML model having a corresponding prediction metric and a corresponding estimated simulation metric that meet a first predetermined criteria, wherein the second variant of the first ML model is used by an autonomous driving vehicle (ADV) to predict a behavior of an obstacle.