Method and apparatus for updating maps using gradient thresholding

Patent metadata

Field	Value
Publication number	US-11592297-B2
Application number	US-201816227953-A
Country	US
Kind code	B2
Filing date	Dec 20, 2018
Priority date	Dec 20, 2018
Publication date	Feb 28, 2023
Grant date	Feb 28, 2023

Abstract

Various methods are provided for facilitating map update to an environment map using gradient thresholding. One example method may include detecting an observed feature associated with a first feature decay and generating an interpolated feature that approximates the observed feature associated with a second decay. The method also includes determining a gradient difference between the interpolated feature and a stored map feature. The stored map feature represents an encoding of the observed feature associated with a third decay associated with an environment map. The method also includes determining a relationship between the gradient difference and a feature gradient update threshold, and, based upon the relationship, updating the environment map by at least replacing the map feature representation associated with the environment map with the approximated feature representation.

First claim

What is claimed is:

1. A method for updating an environment map using gradient thresholding, the method comprising:
receiving an observed feature representation, wherein the observed feature representation includes an observed feature associated with a first feature decay;
applying the observed feature representation to a gradient thresholding neural network to generate an approximated feature representation associated with a second feature decay;
identifying a gradient difference between the approximated feature representation and a map feature representation, wherein the map feature representation represents an encoding of the observed feature associated with a third feature decay associated with an environment map;
determining a relationship between the gradient difference and a feature gradient update threshold; and
based upon the relationship, automatically updating the environment map by at least replacing the map feature representation associated with the environment map with the approximated feature representation.

2. The method of claim 1 further comprising:
determining a vehicle position; and
registering the vehicle position with the environment map by aligning an observed feature set with a stored map feature set, wherein the observed feature set comprises at least the observed feature,
further wherein the stored map feature set comprises at least the stored map feature representation, and wherein the stored map feature set is associated with the environment map.

3. The method of claim 2 further comprising identifying the map feature representation based on the registered vehicle position.

4. The method of claim 1 , wherein the second feature decay associated with the approximated feature decay comprises an interpolated feature decay between the first feature decay and the third feature decay.

5. The method of claim 1 further comprising determining the feature gradient update threshold.

6. The method of claim 1 , wherein the gradient thresholding neural network comprises a variational auto-encoder.

7. The method of claim 1 , wherein determining the relationship comprises determining whether the gradient difference exceeds the feature gradient update threshold, and wherein updating the environment map comprises updating the environment map in an instance in which the gradient difference exceeds the feature gradient update threshold.

8. An apparatus for updating an environment map using gradient thresholding, the apparatus comprising at least one processor and at least one non-transitory memory including computer program code instructions stored thereon, the computer program code instructions configured to, when executed by the at least one processor, cause the apparatus to:
receive an observed feature representation, wherein the observed feature representation includes an observed feature associated with a first feature decay;
apply the observed feature representation to a gradient thresholding neural network to generate an approximated feature representation associated with a second feature decay;
identify a gradient difference between the approximated feature representation and a map feature representation, wherein the map feature representation represents an encoding of the observed feature associated with a third feature decay associated with an environment map;
determine a relationship between the gradient difference and a feature gradient update threshold; and
based upon the relationship, automatically update, via the apparatus, the environment map by at least replacing the map feature representation associated with the environment map with the approximated feature representation.

9. The apparatus of claim 8 , wherein the computer code instructions are further configured to, when executed, cause the apparatus to:
determine a vehicle position; and
register the vehicle position with the environment map by aligning an observed feature set with a stored map feature set, wherein the observed feature set comprises at least the observed feature,
further wherein the stored map feature set comprises at least the stored map feature representation, and wherein the stored map feature set is associated with the environment map.

10. An apparatus of claim 8 , wherein the computer code instructions are further configured to, when executed, cause the apparatus to identify the map feature representation based on the registered vehicle position.

11. The apparatus of claim 8 , wherein the second feature decay associated with the approximated feature decay comprises an interpolated feature decay between the first feature decay and the third feature decay.

12. The apparatus of claim 8 , wherein the computer code instructions are further configured to, when executed, cause the apparatus to determine the feature gradient update threshold.

13. The apparatus of claim 8 , wherein the gradient thresholding neural network comprises a variational auto-encoder.

14. The apparatus of claim 8 , wherein the computer code instructions configured to cause the apparatus to determine the relationship comprises computer code instructions that, when executed, cause the apparatus to determine whether the gradient difference exceeds the feature gradient update threshold, and wherein the computer code instructions configured to cause the apparatus to update the environment map comprises computer code instructions that, when executed, cause the apparatus to update the environment map in an instance in which the gradient difference exceeds the feature gradient update threshold.

15. A computer program product for updating an environment map using gradient thresholding, the computer program product comprising at least one non-transitory computer-readable storage medium having computer-executable program code instructions stored therein, the computer-executable program code instructions comprising program code instructions for:
receiving an observed feature representation, wherein the observed feature representation includes an observed feature associated with a first feature decay;
applying the observed feature representation to a gradient thresholding neural network to generate an approximated feature representation associated with a second feature decay;
identifying a gradient difference between the approximated feature representation and a map feature representation, wherein the map feature representation represents an encoding of the observed feature associated with a third feature decay associated with an environment map;
determining a relationship between the gradient difference and a feature gradient update threshold; and
based upon the relationship, automatically updating the environment map by at least replacing the map feature representation associated with the environment map with the approximated feature representation.

16. The computer program product of claim 15 , further comprising program code instructions for:
determining a vehicle position; and
registering the vehicle position with the environment map by aligning an observed feature set with a stored map feature set, wherein the observed feature set comprises at least the observed feature,
further wherein the stored map feature set comprises at least the stored map feature representation, and wherein the stored map feature set is associated with the environment map.

17. The computer program product of claim 15 , further program code instructions for identifying the map feature representation based on the registered vehicle position.

18. The computer program product of claim 15 , wherein the sec

Assignees

• Here Global Bv

Inventors

• Viswanathan Anirudh

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06V20/56

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

• G06T7/74

  involving reference images or patches · CPC title

• G06N3/088 Primary

  Non-supervised learning, e.g. competitive learning · CPC title