US 11,893,457 B2Granted

Integrating simulated and real-world data to improve machine learning models

AssigneeInternational Business Machines Corporation

InventorsGerman H Flores, Mu Qiao, Divyesh Jadav

Life of this patent

Estimating the term…

Loading patent…

Abstract

Techniques for data integration and labeling are provided. Training real-world signal data is collected for a physical environment, where the training real-world signal data comprises at least one of (i) coordinate information or (ii) a direction to move. Simulated signal data is generated for a first portion of the physical environment, and an aggregate data set is generated comprising the training real-world signal data and the simulated signal data. A machine learning (ML) model is trained using the aggregate data set. A first real-world data point is received, where the first real-world data point does not include coordinate information, and the first real-world data point is labeled based at least in part on coordinate information of the aggregate data set.

Claims12

Dependent claims

A method, comprising:

collecting training real-world signal data for a physical environment, wherein the training real-world signal data comprises at least one of (i) coordinate information or (ii) a direction to move;
in response to classifying a first region of the physical environment as having high signal quality, based on determining that real-world signal data for the first region satisfies a first set of criteria, generating simulated signal data for the first region, wherein:
classifying the first region of the physical environment as having high signal quality comprises:
selecting a subset of the training real-world signal data that was collected in the first region of the physical environment;
and
evaluating the subset of the training real-world signal data to determine that the subset of the training real-world signal data satisfies the first set of criteria, and
the first set of criteria comprises at least one member selected from a group consisting of (i) a minimum number of beacons reflected in the subset of the training real-world signal data, (ii) a minimum average signal strength reflected in the subset of the training real-world signal data, and (iii) a maximum signal variance reflected in the subset of the training real-world signal data;
in response to classifying a second region of the physical environment as having poor signal quality, based on determining that real-world signal data for the second region fails to satisfy a second set of criteria:
refraining from generating simulated signal data for the second region;
and
requesting additional training real-world signal data from the second region;
in response to classifying a third region of the physical environment as having adequate signal quality, based on determining that real-world signal data for the third region fails to satisfy the first set of criteria and satisfies the second set of criteria:
refraining from generating simulated signal data for the third region;
and
refraining from requesting additional training real-world signal data from the third region;
generating an aggregate data set comprising the training real-world signal data and the simulated signal data;
training a machine learning (ML) model using the aggregate data set;
receiving, at runtime, a first real-world data point, wherein the first real-world data point does not include coordinate information;
labeling the first real-world data point based at least in part on coordinate information of the aggregate data set;
generating a next direction to move by evaluating the first real-world data point using the trained ML model;
and
returning the next direction to move to a device that transmitted the first real-world data point.

19 more elements

One or more computer-readable storage media collectively containing computer program code that, when executed by operation of one or more computer processors, performs an operation comprising:

collecting training real-world signal data for a physical environment, wherein the training real-world signal data comprises at least one of (i) coordinate information or (ii) a direction to move;
in response to classifying a first region of the physical environment as having high signal quality, based on determining that real-world signal data for the first region satisfies a first set of criteria, generating simulated signal data for the first region, wherein:
classifying the first region of the physical environment as having high signal quality comprises:
selecting a subset of the training real-world signal data that was collected in the first region of the physical environment;
and
evaluating the subset of the training real-world signal data to determine that the subset of the training real-world signal data satisfies the first set of criteria, and
the first set of criteria comprises at least one member selected from a group consisting of (i) a minimum number of beacons reflected in the subset of the training real-world signal data, (ii) a minimum average signal strength reflected in the subset of the training real-world signal data, and (iii) a maximum signal variance reflected in the subset of the training real-world signal data;
in response to classifying a second region of the physical environment as having poor signal quality, based on determining that real-world signal data for the second region fails to satisfy a second set of criteria:
refraining from generating simulated signal data for the second region;
and
requesting additional training real-world signal data from the second region;
in response to classifying a third region of the physical environment as having adequate signal quality, based on determining that real-world signal data for the third region fails to satisfy the first set of criteria and satisfies the second set of criteria:
refraining from generating simulated signal data for the third region;
and
refraining from requesting additional training real-world signal data from the third region;
generating an aggregate data set comprising the training real-world signal data and the simulated signal data;
training a machine learning (ML) model using the aggregate data set;
receiving, at runtime, a first real-world data point, wherein the first real-world data point does not include coordinate information;
labeling the first real-world data point based at least in part on coordinate information of the aggregate data set;
generating a next direction to move by evaluating the first real-world data point using the trained ML model;
and
returning the next direction to move to a device that transmitted the first real-world data point.

19 more elements

A system comprising:

one or more computer processors;
and
one or more memories collectively containing one or more programs which when executed by the one or more computer processors performs an operation, the operation comprising:
collecting training real-world signal data for a physical environment, wherein the training real-world signal data comprises at least one of (i) coordinate information or (ii) a direction to move;
in response to classifying a first region of the physical environment as having high signal quality, based on determining that real-world signal data for the first region satisfies a first set of criteria, generating simulated signal data for the first region, wherein:
classifying the first region of the physical environment as having high signal quality comprises:
selecting a subset of the training real-world signal data that was collected in the first region of the physical environment;
and
evaluating the subset of the training real-world signal data to determine that the subset of the training real-world signal data satisfies the first set of criteria, and
the first set of criteria comprises at least one member selected from a group consisting of (i) a minimum number of beacons reflected in the subset of the training real-world signal data, (ii) a minimum average signal strength reflected in the subset of the training real-world signal data, and (iii) a maximum signal variance reflected in the subset of the training real-world signal data;
in response to classifying a second region of the physical environment as having poor signal quality, based on determining that real-world signal data for the second region fails to satisfy a second set of criteria:
refraining from generating simulated signal data for the second region;
and
requesting additional training real-world signal data from the second region;
in response to classifying a third region of the physical environment as having adequate signal quality, based on determining that real-world signal data for the third region fails to satisfy the first set of criteria and satisfies the second set of criteria:
refraining from generating simulated signal data for the third region;
and
refraining from requesting additional training real-world signal data from the third region;
generating an aggregate data set comprising the training real-world signal data and the simulated signal data;
training a machine learning (ML) model using the aggregate data set;
receiving, at runtime, a first real-world data point, wherein the first real-world data point does not include coordinate information;
labeling the first real-world data point based at least in part on coordinate information of the aggregate data set;
generating a next direction to move by evaluating the first real-world data point using the trained ML model;
and
returning the next direction to move to a device that transmitted the first real-world data point.

22 more elements

Dependencies are inferred from the claim text.