US 12,466,431 B2
Using sliced data to evaluate machine learning models used for autonomous vehicle operation
Shuai Zheng, Santa Clara, CA (US); Amirreza Shaban, Bellevue, WA (US); and Sachithra Hemachandra, Sunnyvale, CA (US)
Assigned to GM CRUISE HOLDINGS LLC, San Francisco, CA (US)
Filed by GM Cruise Holdings LLC, San Francisco, CA (US)
Filed on May 22, 2023, as Appl. No. 18/321,477.
Prior Publication US 2024/0391487 A1, Nov. 28, 2024
Int. Cl. B60W 60/00 (2020.01); G06N 20/00 (2019.01)
CPC B60W 60/001 (2020.02) [G06N 20/00 (2019.01); B60W 2555/20 (2020.02)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
identifying an environmental feature that is critical for an environmental condition associated with a vehicle, wherein the environmental condition is a condition in an environment where the vehicle performs an operation;
extracting, from a dataset that comprises sensor data capturing the environment, a sub-dataset that comprises sensor data capturing the environmental feature;
inputting the sub-dataset into a plurality of machine learning models;
evaluating performances of the plurality of machine learning models based on outputs of the plurality of machine learning models and a ground-truth classification of the environmental feature, each output generated by a respective machine learning model based on the sub-dataset and comprising a classification of the environmental feature, wherein the ground-truth classification of the environmental feature is a verified accurate classification of the environmental feature;
selecting a machine learning model from the plurality of machine learning models based on the performances of the plurality of machine learning models; and
controlling one or more operational behaviors of the vehicle or another vehicle by using the selected machine learning model.