US 12,263,594 B2
Systems, robots, and methods for selecting classifiers based on context
Suzanne Gildert, Vancouver (CA); William G. Macready, West Vancouver (CA); and Thomas Mahon, Ariss (CA)
Assigned to Sanctuary Cognitive Systems Corporation, Vancouver (CA)
Filed by Sanctuary Cognitive Systems Corporation, Vancouver (CA)
Filed on Dec. 12, 2022, as Appl. No. 18/079,717.
Application 18/079,717 is a continuation of application No. 17/962,365, filed on Oct. 7, 2022.
Claims priority of provisional application 63/253,591, filed on Oct. 8, 2021.
Prior Publication US 2023/0111067 A1, Apr. 13, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 10/80 (2022.01); B25J 9/16 (2006.01)
CPC B25J 9/1661 (2013.01) [B25J 9/163 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A method of operating a robot, the method comprising:
receiving sensor data representing an environment of the robot, wherein the environment comprises a set of features;
executing a first subset of classifiers selected from a library of classifiers comprising a set of classifiers on the sensor data, wherein executing the first subset of classifiers identifies a first subset of features in the environment of the robot;
determining a first characterization of a context of the robot based on a result of executing the first subset of classifiers;
selecting a second subset of classifiers from the library of classifiers, wherein the second subset of classifiers is different from the first subset of classifiers;
executing the second subset of classifiers on the sensor data, wherein executing the second subset of classifiers identifies a second subset of features in the environment of the robot;
determining a second characterization of the context of the robot based on a result of executing the second subset of classifiers;
determining the context of the robot based at least in part on the first characterization of the context of the robot and the second characterization of the context of the robot;
identifying a task for the robot to perform based at least in part on the context of the robot;
selecting a third subset of classifiers from the library of classifiers, wherein the third subset of classifiers is different from the first subset of classifiers and the second subset of classifiers;
executing the third subset of classifiers on the sensor data, wherein executing the third subset of classifiers identifies a third subset of features in the environment of the robot;
determining a third characterization of the context of the robot based on a result of executing the third subset of classifiers; and
adjusting the context of the robot based on the third characterization of the robot.