Algorithm and infrastructure for robust and efficient vehicle localization

• US 10,579,065 B2
• Filed: 11/23/2016
• Issued: 03/03/2020
• Est. Priority Date: 11/23/2016
• Status: Active Grant

• Current Assignee: Baidu USA LLC (Baidu Incorporated)
• Original Assignee: Baidu USA LLC (Baidu Incorporated)
• Inventors: Wang, Quan, Zhai, Jing
• Primary Examiner(s): Odeh, Nadeem
• Application Number:
  US15/360,882
• Publication Number:
  US 20180143647A1
• Time in Patent Office: 1,196 Days
• Field of Search: 701301, 701533, 701 27, 701448, 701469, 701523, 701514, 700264, 702150, 706 52, 382118, 3482071
• CPC Class Codes: G01C 21/30 Map- or contour-matching
• CPC Class Codes: G01C 21/3848 Data obtained from both position sensors and additional sensors
• CPC Class Codes: G01C 21/3881 Tile-based structures
• CPC Class Codes: G01S 13/86 Combinations of radar systems with non-radar systems, e.g. sonar, direction finder
• CPC Class Codes: G01S 17/06 Systems determining position data of a target
• CPC Class Codes: G01S 17/86 Combinations of lidar systems with systems other than lidar, radar or sonar, e.g. with direction finders
• CPC Class Codes: G01S 7/4808 Evaluating distance, position or velocity data
• CPC Class Codes: G05D 1/0088 characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours using knowledge based models G06N5/00
• CPC Class Codes: G05D 1/0274 using mapping information stored in a memory device navigation using map-matching G01C21/30
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Location of an autonomous driving vehicle (ADV) is determined with respect to a high definition map. On-boards sensors of the ADV obtain a 3D point cloud of objects surrounding the ADV. The 3D point cloud is organized into an ADV feature space of cells. Each cell has a median intensity value and a variance in elevation. To determine the ADV location, a coarse search of a subset of cells in the ADV feature space performed with respect to the high definition map, using a similarity metric that is based on the median intensity and variance in elevation of the candidate cell. When similarity of the first candidate cell is determined, a lookup table of similarity scores is generated and used for determining the similarity score for subsequent candidate cells. Then a fine search is performed on a small subset of candidate cells surrounding the highest similarity score cell.

Claims

1. A computer-implemented method of determining an accurate location of an autonomous driving vehicle (ADV) with respect to a high-definition (HD) map representing a real-world environment surrounding the ADV and operating the ADV, the method comprising:

determining an approximate location of the ADV using a global positioning system (GPS location of the ADV);

obtaining a section of the HD map having the GPS location of the ADV substantially at the center of the section of the HD map;

determining a first set of a plurality of candidate cells of an ADV feature space of cells surrounding the ADV, the first set representing a coarse search space of the ADV feature space, each candidate cell in the coarse search space having a median intensity and a variance in elevation, the ADV feature space determined using one or more sensors of the ADV;

for each candidate cell in the coarse search space, determining a similarity score between a feature space that surrounds the candidate cell (candidate cell feature space), and a feature space of the section of the HD map surrounding the GPS location of the ADV, using a similarity metric, wherein the similarity metric is based at least in part on the candidate cell median intensity and the candidate cell variance in elevation, wherein the similarity metric represents, for the candidate cell, a similarity of the candidate cell feature space to the HD map feature space surrounding the GPS location of the ADV;

determining a location of the ADV with respect to the HD map feature space based at least in part on the candidate cell having the highest similarity score among the candidate cells of the coarse search space; and

navigating the ADV based at least in part on the determined location of the ADV with respect to the HD map.

2. The method of claim 1, wherein the first set of the plurality of candidate cells comprise every other cell of a predetermined grid of cells surrounding the ADV.

3. The method of claim 1, wherein, for a first candidate cell in the coarse search space, determining the similarity score for the first candidate cell comprises computing the similarity score using the first candidate median intensity and the first candidate cell variance in elevation, and the method further comprises:

using information obtained from computing the similarity score of the first candidate cell to generate a lookup table of similarity scores.

4. The method of claim 3, wherein the lookup table is indexed by a range of mean intensity values and a range of variance in elevation values.

5. The method of claim 3, wherein for a second candidate cell, and subsequent candidate cells in the plurality of candidate cells, determining the similarity score for each of the second candidate cell and the subsequent candidate cells comprises looking up the similarity score in the lookup table using the median intensity and the variance in elevation for each of the second candidate cell and the subsequent candidate cells, respectively.

6. The method of claim 1, further comprising, after determining the candidate cell having a highest similarity score in the coarse search space:

determining a second set of a plurality of candidate cells surrounding the candidate cell having the highest similarity score of the coarse search space, the second set representing a fine search space within the coarse search space; and

for each candidate cell in the fine search space;

determining a fine search similarity score between a feature space that surrounds the fine search candidate cell (fine search candidate cell feature spacer), and the HD map feature space, wherein the fine search similarity score represents, for the fine search candidate cell, a similarity of the fine search candidate cell feature space to the HD map feature space;

and

wherein determining the location of the ADV location within the HD map feature space is further based at least in part on the fine search candidate cell having the highest fine search similarity score.

7. The method of claim 6, wherein determining the fine search similarity score comprises computing the fine search similarity score for each of the plurality of fine search candidate cells.

8. The method of claim 6, wherein determining the location of the ADV within the HD map feature space further comprises determining an offset between the ADV and the fine search candidate cell having the highest fine search similarity score to the HD map feature space.

9. A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations that determine an accurate location of an autonomous driving vehicle (ADV) with respect to a high-definition (HD) map representing a real-world environment surrounding the ADV, the operations comprising:

determining an approximate location of the ADV using a global positioning system (GPS location of the ADV);

obtaining a section of the HD map having the GPS location of the ADV substantially at the center of the section of the HD map;

determining a first set of a plurality of candidate cells of an ADV feature space of cells surrounding the ADV, the first set representing a coarse search space of the ADV feature space, each candidate cell in the coarse search space having a median intensity and a variance in elevation, the ADV feature space determined using one or more sensors of the ADV;

for each candidate cell in the coarse search space, determining a similarity score between a feature space that surrounds the candidate cell (candidate cell feature spacer), and a feature space of the section of HD map surrounding the GPS location of the ADV, using a similarity metric, wherein the similarity metric is based at least in part of on the candidate cell median intensity and the candidate cell variance in elevation, wherein the similarity metric represents, for the candidate cell, a similarity of the candidate cell feature space to the HD map feature space surrounding the GPS location of the ADV;

determining a location of the ADV with respect to the HD map feature space based at least in part on the candidate cell having the highest similarity score among the candidate cells of the coarse search space; and

navigating the ADV based at least in part on the determined location of the ADV with respect to the HD map.

10. The non-transitory machine-readable medium of claim 9, wherein the first set of the plurality of candidate cells comprise every other cell of a predetermined grid of cells surrounding the ADV.

11. The non-transitory machine-readable medium of claim 9, wherein, for a first candidate cell in the coarse search space, determining the similarity score for the first candidate cell comprises computing the similarity score using the first candidate median intensity and the first candidate cell variance in elevation, and the operations further comprise:

using information obtained from computing the similarity score of the first candidate cell to generate a lookup table of similarity scores.

12. The non-transitory machine-readable medium of claim 11, wherein the lookup table is indexed by a range of mean intensity values and a range of variance in elevation values.

13. The non-transitory machine-readable medium of claim 11, wherein for a second candidate cell, and subsequent candidate cells in the plurality of candidate cells, determining the similarity score for each of the second candidate cell and the subsequent candidate cells comprises looking up the similarity score in the lookup table using the median intensity and the variance in elevation for each of the second candidate cell and the subsequent candidate cells, respectively.

14. The non-transitory machine-readable medium of claim 9, further comprising, after determining the candidate cell having a highest similarity score in the coarse search space:

determining a second set of plurality of candidate cells surrounding the candidate cell having the highest similarity score of the coarse search space, the second set representing a fine search space within the coarse search space; and

for each candidate cell in the fine search space;

determining a fine search similarity score between a feature space that surrounds the fine search candidate cell (fine search candidate cell feature spacer), and the HD map feature space;

and

wherein determining the location of the ADV location within the HD map feature space is further based at least in part on the fine search candidate cell having the highest fine search similarity score.

15. The non-transitory machine-readable medium of claim 14, wherein determining the fine search similarity score comprises computing the fine search similarity score for each of the plurality of fine search candidate cells.

16. The non-transitory machine-readable medium of claim 14, wherein determining the location of the ADV within the HD map feature space further comprises determining an offset between the ADV and the fine search candidate cell having the highest fine search similarity score to the HD map feature space.

17. A data processing system, comprising:

a processor; and

a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations that determine an accurate location of an autonomous driving vehicle (ADV) with respect to a high-definition (HD) map representing a real-world environment surrounding the ADV, the operations including;

determining an approximate location of the ADV using a global positioning system (GPS location of the ADV);

obtaining a section of the HD map having the GPS location of the ADV substantially at the center of the section of the HD map;

determining a first set of a plurality of candidate cells of an ADV feature space of cells surrounding the ADV, the first set representing a coarse search space of the ADV feature space, each candidate cell in the coarse search space having a median intensity and a variance in elevation, the ADV feature space determined using one or more sensors of the ADV;

for each candidate cell in the coarse search space, determining a similarity score between a feature space that surrounds the candidate cell (candidate cell feature spacer), and a feature space of the section of the HD map surrounding the GPS location of the ADV, using a similarity metric, wherein the similarity metric is based at least in part on the candidate cell median intensity and the candidate cell variance in elevation, wherein the similarity score represents, for the candidate cell, a similarity of the candidate cell feature space to the HD map feature space surrounding the GPS location of the ADV;

determining a location of the ADV with respect to the HD map feature space based at least in part on the candidate cell having the highest similarity score among the candidate cells of the coarse search space; and

navigating the ADV based at least in part on the determined location of the ADV with respect to the HD map.

18. The data processing system of claim 17, wherein the first set of the plurality of candidate cells comprise every other cell of a predetermined grid of cells surrounding the ADV.

19. The data processing system of claim 17, wherein, for a first candidate cell in the coarse search space, determining the similarity score for the first candidate cell comprises computing the similarity score using the first candidate median intensity and the first candidate cell variance in elevation, and the operations further comprise:

using information obtained from computing the similarity score of the first candidate cell to generate a lookup table of similarity scores.

20. The data processing system of claim 19, wherein the lookup table is indexed by a range of mean intensity values and a range of variance in elevation values.

21. The data processing system of claim 19, wherein for a second candidate cell, and subsequent candidate cells in the plurality of candidate cells, determining the similarity score for each of the second candidate cell and the subsequent candidate cells comprises looking up the similarity score in the lookup table using the median intensity and the variance in elevation for each of the second candidate cell and the subsequent candidate cells, respectively.

22. The data processing system of claim 17, the operations further comprising, after determining the candidate cell having a highest similarity score in the coarse search space:

determining a second set of a plurality of candidate cells surrounding the candidate cell having the highest similarity score of the coarse search space, the second set representing a fine search space within the coarse search space; and

for each candidate cell in the fine search space;

determining a fine search similarity score between a feature space that surrounds the fine search candidate cell (fine search candidate cell feature spacer), and the HD map feature space, wherein the fine search similarity score represents, for the fine search candidate cell, a similarity of the fine search candidate cell feature space to the HD map feature space;

and

wherein determining the location of the ADV location within the HD map feature space is further based at least in part on the fine search candidate cell having the highest fine search similarity score.

23. The data processing system of claim 22, wherein determining the fine search similarity score comprises computing the fine search similarity score for each of the plurality of fine search candidate cells.

24. The data processing system of claim 22, wherein determining the location of the ADV within the HD map feature space further comprises determining an offset between the ADV and the fine search candidate cell having the highest fine search similarity score to the HD map feature space.