US 12,260,651 B2
Machine-learned architecture for efficient object attribute and/or intention classification
Subhasis Das, San Mateo, CA (US); Oytun Ulutan, Buena Park, CA (US); Yi-Ting Lin, Foster City, CA (US); and Derek Xiang Ma, San Carlos, CA (US)
Assigned to Zoox, Inc., Foster City, CA (US)
Filed by Zoox, Inc., Foster City, CA (US)
Filed on Mar. 29, 2024, as Appl. No. 18/621,922.
Application 18/621,922 is a continuation of application No. 17/522,832, filed on Nov. 9, 2021, granted, now 11,972,614.
Prior Publication US 2024/0320985 A1, Sep. 26, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 20/58 (2022.01); G05D 1/00 (2024.01); G05D 1/249 (2024.01); G06N 3/04 (2023.01); G06N 3/08 (2023.01); G06V 40/20 (2022.01)
CPC G06V 20/58 (2022.01) [G05D 1/0221 (2013.01); G05D 1/0246 (2013.01); G05D 1/249 (2024.01); G06N 3/04 (2013.01); G06N 3/08 (2013.01); G06V 40/23 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving first sensor data associated with an object;
receiving second sensor data associated with the object;
determining, by a first machine-learned model and based at least in part on the first sensor data, a first output;
storing the first output in a cache;
retrieving, from the cache a second output previously generated by the first machine-learned model and based at least in part on the second sensor data;
aggregating, as an aggregated output, the first output and the second output;
determining, by a second machine-learned model and based at least in part on the aggregated output, an attribute associated with the object in an environment, the attribute indicating at least one of a motion of the object, a state of the object, a predicted intent of the object, or an association of the object with another object; and
controlling a vehicle based at least in part on the attribute.