Systems, robots, and methods for selecting classifiers based on context

Patent metadata

Field	Value
Publication number	US-12246452-B2
Application number	US-202217962365-A
Country	US
Kind code	B2
Filing date	Oct 7, 2022
Priority date	Oct 8, 2021
Publication date	Mar 11, 2025
Grant date	Mar 11, 2025

Abstract

The present disclosure describes systems, robots, and methods for organizing and selecting classifiers of a library of classifiers. The classifiers of the library of classifiers can be organized in a relational model, such as a hierarchy or probability model. Instead of storing, activating, or executing the entire library of classifiers at once by a robot system, computational resource demand is reduced by executing subset of classifiers to determine context, and the determined context is used as a basis for selection of another subset of classifiers. This process can be repeated, to iteratively refine context and select more specific subsets of classifiers. A selected subset of classifiers can eventually be specific to a task to be performed by the robot system, such that the robot system can take action based on output from executing such specific classifiers.

First claim

The invention claimed is:

1. A method of operation of a robot system, the robot system comprising a robot body and a robot controller, wherein the robot controller comprises at least one processor and at least one non-transitory processor-readable storage medium communicatively coupled to the at least one processor, and wherein the at least one non-transitory processor-readable storage medium stores a library of classifiers, the method comprising:
activating, by the robot controller, a first subset of classifiers from the library of classifiers;
determining, by the robot controller, a first characterization of a context of the robot body, wherein determining the first characterization of the context of the robot body includes executing, by the robot controller, the first subset of classifiers from the library of classifiers to result in each respective classifier in the first subset of classifiers providing either a positive result or a negative result;
in response to determining the first characterization of the context of the robot body, selecting, by the robot controller, a second subset of classifiers from the library of classifiers based on the first characterization of the context of the robot body, the second subset of classifiers different from the first subset of classifiers, wherein each respective classifier in the second subset of classifiers is related to at least one respective positive result of at least one respective classifier in the first subset of classifiers;
activating, by the robot controller, the second subset of classifiers;
determining, by the robot controller, at least one additional characterization of the context of the robot body, each respective additional characterization of the context of the robot body including additional characterization information to a respective immediately preceding characterization of the context of the robot body, wherein determining each respective additional characterization of the context of the robot body includes executing, by the robot controller, a respective additional subset of classifiers from the library of classifiers;
in response to determining each respective additional characterization of the context of the robot body, selecting, by the robot controller, a respective second additional subset of classifiers from the library of classifiers based on a respective additional characterization of the context of the robot body; and
activating, by the robot controller, each respective additional subset of classifiers.

2. The method of claim 1 wherein selecting, by the robot controller, a second subset of classifiers from the library of classifiers based on the first characterization of the context of the robot body includes: selecting, by the robot controller, a second subset of classifiers from the library of classifiers that are expected to provide context information of the robot body which is additional to context information provided by the first subset of classifiers.

3. The method of claim 2 wherein selecting, by the robot controller, a second subset of classifiers from the library of classifiers that are expected to provide context information of the robot body which is additional to context information provided by the first subset of classifiers comprises:
based on a relational model which defines probabilities of classifiers in the library of classifiers producing positive detection outputs based on outputs from active classifiers, selecting, by the robot controller, a second subset of classifiers from the library of classifiers that have a probability of positive detection which is within a probability range.

4. The method of claim 1 wherein selecting, by the robot controller, a second subset of classifiers from the library of classifiers based on the first characterization of the context of the robot body includes selecting, by the robot controller, a second subset of classifiers from the library of classifiers wherein each respective classifier in the second subset of classifiers is related to at least one respective classifier in the first subset of classifiers.

5. The method of claim 4 , further comprising:
selecting an additional subset of at least one classifier, wherein each respective classifier in the additional subset is unrelated to each of the classifiers in the first subset of classifiers; and
activating, by the robot controller, the additional subset of at least one classifier.

6. The method of claim 1 , wherein:
selecting, by the robot controller, a respective additional subset of classifiers from the library of classifiers comprises: selecting, by the robot controller, each additional subset of classifiers from the library of classifiers based on a relational model which defines relationships between classifiers in the library of classifiers;
the method further comprises after executing, by the robot controller, a respective additional subset of classifiers from the library of classifiers: adjusting the relational model based on outputs from executing the respective additional subset of classifiers.

7. The method of claim 6 , wherein the relational model is a probability model which defines probabilities of classifiers in the library of classifiers producing positive detection outputs based on outputs from active classifiers, and wherein adjusting the relational model based on outputs from executing the respective additional subset of classifiers comprises: increasing respective probability values for classifiers in the library of classifiers related to outputs from an immediately preceding subset of classifiers, for classifiers which produced positive detection outputs.

8. The method of claim 6 , wherein the relational model is a probability model which defines probabilities of classifiers in the library of classifiers producing positive detection outputs based on outputs from active classifiers, and wherein adjusting the relational model based on outputs from executing the respective additional subset of classifiers comprises: decreasing respective probability values for classifiers in the library of classifiers related to outputs from an immediately preceding subset of classifiers, for classifiers which produced negative detection outputs.

9. The method of claim 1 , further comprising disabling at least one classifier of the first subset of classifiers after determining the first characterization.

10. The method of claim 1 , wherein:
the at least one non-transitory processor-readable storage medium of the robot controller stores the library of classifiers remote from the robot body; and
the method further comprises accessing, by a communication interface of the robot body, the first subset of classifiers on the non-transitory processor-readable storage medium remote from the robot body.

11. The method of claim 1 , wherein:
the at least one processor of the robot controller includes a first processor carried by the robot body, and a second processor remote from the robot body;
the at least one non-transitory processor-readable storage medium of the robot controller includes a first non-transitory processor-readable storage medium carried by the robot body, and a second non-transitory processor-readable storage medium which stores the library of classifiers remote from the robot body;
executing, by the robot controller, the first subset of classifiers from the library of classifiers comprises executing, by the first processor, the first subset of classifiers from the library of classifiers.

12. The method of claim 11 , wherein activating the first subset of classifiers comprises transferring, by a communication interface, the first subset of classifiers from the second non-t

Assignees

• Sanctuary Cognitive Systems Corp

Inventors

• Gildert Suzanne
• Macready William G
• Mahon Thomas

Classifications

• B25J9/163

  learning, adaptive, model based, rule based expert control · CPC title

• B25J9/1661 Primary

  characterised by task planning, object-oriented languages · CPC title

• G06V10/809 Primary

  of classification results, e.g. where the classifiers operate on the same input data · CPC title