US 12,269,510 B2
Rare scenario handling for autonomous vehicles
Burkay Donderici, Burlingame, CA (US)
Assigned to GM Cruise Holdings LLC, San Francisco, CA (US)
Filed by GM Cruise Holdings LLC, San Francisco, CA (US)
Filed on Oct. 6, 2022, as Appl. No. 17/961,401.
Prior Publication US 2024/0116539 A1, Apr. 11, 2024
Int. Cl. B60W 60/00 (2020.01); G06N 20/00 (2019.01)
CPC B60W 60/0015 (2020.02) [B60W 60/0013 (2020.02); G06N 20/00 (2019.01)] 18 Claims
OG exemplary drawing
 
1. A method comprising:
determining, based on likelihood values associated with a set of available autonomous vehicle (AV) scene data, a target likelihood value range for identifying relevant rare scenarios, the set of available AV scene data describing scenarios involving AVs and surrounding objects;
generating a known AV scenario based on the set of available AV scene data;
modifying the known AV scenario to generate a new AV scenario;
generating, using a machine learning model trained based on the set of available AV scene data, a likelihood value indicating a probability of the new AV scenario occurring;
determining that the likelihood value is within the target likelihood value range;
in response to determining that the likelihood value is within the target likelihood value range, initiating a computer-generated simulation of the new AV scenario;
capturing a set of synthetic AV scene data from the computer-generated simulation of the new AV scenario;
training a machine learning model based on the set of synthetic AV scene data; and
implementing the machine learning model to direct operation of an AV.