Validation and training service for dynamic environment perception based on local high confidence information

Patent metadata

Field	Value
Publication number	US-12082082-B2
Application number	US-202017131161-A
Country	US
Kind code	B2
Filing date	Dec 22, 2020
Priority date	Dec 22, 2020
Publication date	Sep 3, 2024
Grant date	Sep 3, 2024

Abstract

Described herein is a high confidence ground truth information service executing on a network of edge computing devices. A variety of participating devices obtain high confidence ground truth information relating to objects in a local environment. This information is communicated to the ground truth information service, where it may be verified and aggregated with similar information before being communicated as part of an acquired ground truth dataset to one or more subscribing devices. The subscribing devices use the ground truth information, as included in the ground truth dataset, to both validate and improve their supervised learning systems.

First claim

What is claimed is:

1. An edge computing device comprising:
a wireless communication device for receiving and sending data;
a data storage device for storing data;
one or more processors for executing instructions;
a memory device storing instructions thereon, which, when executed by the processor, cause the edge computing device to perform operations comprising:
receiving, via the wireless communication device, data from a plurality of devices in the local environment, the data received from each device including information identifying an object in the local environment and indicating a value for a property of the object;
for each value for a property of an object received, verifying the value satisfies a confidence threshold requirement; and
communicating an acquired ground truth dataset to a vehicle via the wireless communication device, wherein the acquired ground truth dataset includes at least one instance of a verified value for a property of an object in the local environment of the vehicle.

2. The edge computing device of claim 1 , where verifying the value satisfies a confidence threshold requirement includes confirming that meta-data, as received with the information indicating the value for the property of the object, indicates a source of information that is a trusted source of information.

3. The edge computing device of claim 1 , wherein verifying the value satisfies a confidence threshold requirement includes confirming that a confidence level, as received with the information indicating the value for the property of the object, exceeds a confidence threshold for the type of property of the object, the type of device from which the information was received, or the type of sensor from which the value of the property was derived.

4. The edge computing device of claim 1 , further comprising:
aggregating information, received from different devices in the plurality of devices, relating to the same object in the local environment;
wherein verifying information indicating the value satisfies a confidence threshold requirement includes verifying that each value for a property of the same object, or an average of all values for the same property of the object deviates by less than a threshold.

5. The edge computing device of claim 1 , wherein a device in the plurality of devices from which data is received is a vehicle, the information identifying the object in the local environment identifies the vehicle, and the value for the property of the vehicle is a value derived by one or more sensors of the vehicle and indicates one of: speed of the vehicle; direction of travel of the vehicle; position of the vehicle; or, orientation of the vehicle.

6. The edge computing device of claim 1 , wherein a device in the plurality of devices from which data is received is a stationary device that includes one or more sensors for deriving values for at least one property of an object in the local environment.

7. The edge computing device of claim 1 , further comprising:
deriving the acquired ground truth dataset by aggregating all verified values for all properties of an object in the local environment; and
wherein communicating the acquired ground truth dataset to the vehicle includes communicating the acquired ground truth dataset to the vehicle subsequent to receiving a request for an acquired ground truth dataset from the vehicle.

8. The edge computing device of claim 1 , further comprising:
deriving the acquired ground truth dataset by aggregating all verified values for all properties of an object in the local environment; and
wherein communicating the acquired ground truth dataset to the vehicle includes communicating the acquired ground truth dataset to the vehicle subsequent to having received a request, from the vehicle, to subscribe to receive acquired ground truth datasets.

9. The edge computing device of claim 1 , wherein the acquired ground truth dataset communicated to the vehicle via the wireless communication device enables the vehicle to validate a value derived for the property of the object by a supervised learning system of the vehicle.

10. A non-transitory computer-readable storage medium storing instructions, which, when executed by a processor of an edge computing device, cause the edge computing device to:
receive, via a network, data from a plurality of devices in a local environment, the data received from each device including information identifying an object in the local environment and indicating a value for a property of the object;
for each value for a property of an object received, verify the value satisfies a confidence threshold requirement; and
communicate, via the network, an acquired ground truth dataset to a vehicle, the acquired ground truth dataset including at least one instance of a verified value for a property of an object in the local environment of the vehicle.

11. The non-transitory computer-readable storage medium of claim 10 , storing additional instructions, which, when executed by the processor of the edge computing device, cause the edge computing device to:
verify the value satisfies a confidence threshold requirement by confirming that meta-data, as received with the information indicating the value for the property of the object, indicates a source of information that is a trusted source of information.

12. The non-transitory computer-readable storage medium of claim 10 , storing additional instructions, which, when executed by the processor of the edge computing device, cause the edge computing device to:
verify the value satisfies a confidence threshold requirement by confirming that a confidence level, as received with the information indicating the value for the property of the object, exceeds a confidence threshold for the type of property of the object, the type of device from which the information was received, or the type of sensor from which the value of the property was derived.

13. The non-transitory computer-readable storage medium of claim 10 , storing additional instructions, which, when executed by the processor of the edge computing to:
aggregate information, received from different devices in the plurality of devices, relating to the same object in the local environment; and
verify information indicating the value satisfies a confidence threshold requirement by verifying that each value for a property of the same object, or an average of all values for the same property of the object deviates by less than a threshold.

14. The non-transitory computer-readable storage medium of claim 10 , wherein a device in the plurality of devices from which data is received is a vehicle, the information identifying the object in the local environment identifies the vehicle, and the value for the property of the vehicle is a value derived by one or more sensors of the vehicle and indicates one of: speed of the vehicle; direction of travel of the vehicle; position of the vehicle; or, orientation of the vehicle.

15. The non-transitory computer-readable storage medium of claim 10 , wherein a device in the plurality of devices from which data is received is a stationary device that includes one or more sensors for deriving values for at least one property of an object in the local environment.

16. The non-transitory computer-readable storage medium of claim 10 , storing additional instructions, which, when executed by the processor of the edge computing to:
derive the acquired ground truth dataset by aggregating all verified values for all properties of an object in the local environment; and
communicate the acquired ground truth

Assignees

• Intel Corp

Inventors

• Geissler Florian
• Graefe Ralf
• Paulitsch Michael
• Peng Yang
• Rosales Rafael

Classifications

• G06V20/58

  Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06V20/56

  exterior to a vehicle by using sensors mounted on the vehicle · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F18/25

  Fusion techniques · CPC title