US 12,466,432 B2
Autonomous vehicle driving path label generation for machine learning models
Congrui Hetang, Mountain View, CA (US); and Ningshan Zhang, Mountain View, CA (US)
Assigned to Waymo LLC, Mountain View, CA (US)
Filed by Waymo LLC, Mountain View, CA (US)
Filed on May 9, 2022, as Appl. No. 17/740,215.
Prior Publication US 2023/0356748 A1, Nov. 9, 2023
Int. Cl. B60W 60/00 (2020.01); G05B 13/02 (2006.01)
CPC B60W 60/0011 (2020.02) [G05B 13/0265 (2013.01)] 21 Claims
OG exemplary drawing
 
1. A system to generate a set of training data for training a machine learning model, comprising:
a memory device; and
a processing device, operatively coupled to the memory device, to:
generate training input comprising scene data of a scene representing an environment that is proximate to an autonomous vehicle (AV) and that includes one or more obstacles obstructing an initial AV driving path; and
generate target output for the training input, wherein to generate the target output, the processing device is to:
identify a set of input data comprising a roadgraph identifying an intermediate AV driving path related to the scene representing the environment proximate to the AV, the intermediate AV driving path reflecting a modification to the initial AV driving path to avoid the one or more obstacles obstructing the initial AV driving path;
perform, using the set of input data, a path adjustment operation that identifies one or more candidate AV driving paths based on the intermediate AV driving path and determines a cost value for each of the one or more candidate AV driving paths; and
identify, among the one or more candidate AV driving paths, a final AV driving path having a cost value that satisfies an evaluation criterion, wherein the final AV driving path is included in the set of training data as the target output paired with the training input comprising the scene data identifying the scene.