Anomaly detection and resolution

Patent metadata

Field	Value
Publication number	US-12340282-B2
Application number	US-202017083768-A
Country	US
Kind code	B2
Filing date	Oct 29, 2020
Priority date	Oct 29, 2020
Publication date	Jun 24, 2025
Grant date	Jun 24, 2025

Abstract

Methods, apparatuses, and systems associated with anomaly detection and resolution are described. Examples can include detecting, via a sensor of a robot, an object in a path of the robot while the robot is performing a task in an environment and classifying the object as an anomaly or a non-anomaly and the environment as anomalous or non-anomalous using a machine learning model. Examples can include proceeding with the task responsive to classification of the object as a non-anomaly and the environment as non-anomalous and resolving the anomaly or the anomalous environment and proceeding with the task responsive to classification of the object as an anomaly or the environment as anomalous.

First claim

What is claimed is:

1. A method, comprising:
detecting at a processing resource and via a sensor of a robot, an object in a path of the robot while the robot is performing a task in an environment;
classifying at the processing resource the object as an anomaly or a non-anomaly and the environment as anomalous or non-anomalous using a first machine learning model;
responsive to the processing resource's classification of the object as an anomaly or the environment as anomalous using the first machine learning model, utilizing a second machine learning model to confirm whether or not the object or the environment is anomalous and what type of anomaly the object or the environment is;
proceeding with the task responsive to the processing resource's classification and confirmation of the object as a non-anomaly and the environment as non-anomalous;
determining a plurality of potential resolutions to the anomaly or the anomalous environment responsive to the processing resource's classification and confirmation of the object as an anomaly or the environment as anomalous;
selecting one of the plurality of potential resolutions utilizing a third machine learning model;
resolving the anomaly or the anomalous environment utilizing the selected potential resolution before proceeding with the task;
detecting a new object or an addition to the environment;
classifying the new object or the addition to the environment as an anomaly until a decision is made otherwise based on receipt of user instructions;
collecting data associated with the new object or the addition to the environment; and
training the first machine learning model based on the collected data and the user instructions.

2. The method of claim 1 , further comprising:
receiving at the processing resource, from the sensor, data about the task and the environment of the task while the robot is performing the task;
the processing resource, a memory resource coupled to the processing resource, or both, of the robot learning about the task and the environment of the task via the first machine learning model based on the received data; and
the processing resource, the memory resource, or both updating the first machine learning model based on the received data and data previously received from the sensor during previous task performances.

3. The method of claim 2 , wherein classifying the object as an anomaly or the environment as anomalous comprises determining, using the first machine learning model, the object or the environment deviates from an object normally detected in the path or an environment normally associated with the task based on the received data and the previously received data.

4. The method of claim 2 , further comprising the processing resource, the memory resource, or both:
receiving the first machine learning model from a cloud-based service;
augmenting the first machine learning model with data collected while performing the task; and
updating the first machine learning model while the robot is coupled to a charging station with collected data.

5. The method of claim 2 , further comprising the processing resource, the memory resource, or both learning about the task and the environment of the task and updating the first machine learning model via a cloud-based service, a local area network, or a combination thereof.

6. The method of claim 1 , wherein detecting the object in the path comprises detecting the object in the path of the robot while the robot is performing a cleaning task in the environment.

7. The method of claim 1 , wherein detecting the object in the path comprises detecting the object in the path of the robot while the robot is performing a delivery task in the environment.

8. The method of claim 1 , wherein resolving the anomaly or the anomalous environment comprises the processing resource, a memory resource coupled to the processing resource, or both instructing the robot to move the object to a different location.

9. The method of claim 1 , wherein resolving the anomaly or the anomalous environment comprises the processing resource, a memory resource coupled to the processing resource, or both communicating to a processing resource of a different robot to move the object to a different location.

10. An apparatus, comprising:
a processing resource; and
a memory resource in communication with the processing resource having instructions executable to:
detect, via a sensor of the apparatus, an object in a path of the apparatus while the apparatus is performing a cleaning task in an environment;
classify the object in the path of the apparatus as an anomaly or a non-anomaly and the environment as anomalous or non-anomalous using a first machine learning model based on historical data associated with an object previously detected by the sensor;
responsive to the classification of the object as an anomaly or the environment as anomalous using the first machine learning model, utilize a second machine learning model to confirm whether or not the object or the environment is anomalous and what type of anomaly the object or the environment is; and
determine a response to the object in the path of the apparatus based on the classification, confirmation, and the first and the second machine learning model based on historical resolution data associated with the object previously detected by the sensor, wherein the response comprises instructions executable to:
remove the object in the path of the apparatus and proceed with the cleaning task responsive to classification of the object in the path of the apparatus as a non-anomaly and the environment as non-anomalous;
determine a plurality of potential resolutions to the anomaly or the anomalous environment responsive to classification and confirmation of the object in the path of the apparatus as an anomaly or the environment as anomalous;

select one of the plurality of potential resolutions utilizing a third machine learning model;
resolve the anomaly or the anomalous environment utilizing the selected potential resolution before proceeding with the task;
detect a new object or an addition to the environment;
classify the new object or the addition to the environment as an anomaly until a decision is made otherwise based on receipt of user instructions;
collect data associated with the new object or the addition to the environment; and
train the first machine learning model based on the collected data and the user instructions.

11. The apparatus of claim 10 , further comprising the instructions executable to classify the object based on historical data of physical characteristics, environmental characteristics, or a combination thereof associated with the object previously detected by the sensor.

12. The apparatus of claim 10 , wherein the instructions executable to resolve the anomaly or the anomalous environment utilizing the selected potential resolution comprise instructions executable to:
instruct the apparatus to move the object in the path of the apparatus to a different location;
instruct a different apparatus to move the object in the path of the apparatus to the different location; or
a combination thereof.

13. The apparatus of claim 10 , further comprising the instructions executable to classify the object in the path of the apparatus using the first machine learning model based on a combination of:
the historical data associated with the object previously detected by the sensor;
historical data of physical characteristics associated with the object previously detected by the sensor; and
received user input associated with classification of the object in the path of the appara

Assignees

• Micron Technology Inc

Inventors

• Vengertsev Dmitry
• Hosseinimakarem Zahra
• Egorova Marta

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/098

  Distributed learning, e.g. federated learning · CPC title

• G06F16/285

  Clustering or classification · CPC title

• G05B19/4155

  characterised by program execution, i.e. part program or machine function execution, e.g. selection of a program · CPC title

• G05B2219/42018

  Pid learning controller, gains adapted as function of previous error · CPC title