| CPC B60W 60/0027 (2020.02) [G05D 1/0221 (2013.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01); B60W 2552/10 (2020.02); B60W 2554/4042 (2020.02); B60W 2554/4043 (2020.02); B60W 2554/4044 (2020.02); B60W 2554/4045 (2020.02)] | 20 Claims |
|
1. A method performed by one or more computers, the method comprising:
obtaining features of an agent travelling on a roadway;
identifying one or more unmapped U-turn regions in a vicinity of the agent on the roadway based on a respective first score for each unmapped U-turn region, wherein:
each unmapped U-turn region comprises a candidate lane segment in a road graph of the region of the roadway,
the candidate lane segment corresponds to one or more target lane segments in the road graph on an opposite side of the roadway relative to the candidate lane segment, and
the respective first score for each unmapped U-turn region represents a likelihood that agents will make a U-turn from the candidate lane segment into one of the one or more target lane segments that corresponds to the candidate lane segment;
generating, for each of the unmapped U-turn regions and from at least the features of the agent, a respective likelihood score that represents a likelihood that the agent intends to make an unmapped U-turn at the unmapped U-turn region;
selecting, based on the respective likelihood scores, one or more of the unmapped U-turn regions;
for each selected unmapped U-turn region, providing data specifying a candidate future trajectory in which the agent makes the unmapped U-turn at the selected unmapped U-turn region as a possible future trajectory for the agent; and
controlling an autonomous vehicle based on the data specifying the candidate future trajectory of the agent.
|