Systems and methods for estimating cuboids from LiDAR, map and image data

1. A method for operating a robotic system, comprising:

inferring, by a computing device, a first heading distribution for the object from a 3D point cloud;

obtaining, by the computing device, a second heading distribution from a vector map;

obtaining, by the computing device, a posterior distribution of a heading using the first and second heading distributions;

defining, by the computing device, a cuboid on a 3D graph using the posterior distribution; and

using the cuboid to facilitate driving-related operations of a robotic system.

2. The method according to claim 1 , wherein the first heading distribution is inferred based on at least one of a minimal bounding box, an area of the minimal bounding box, a heading direction or yaw in a frame, an area of the minimal bounding box that does not have any LiDAR points, and a distance between an edge of a bounding box at a current elevation to a most outer edge.

3. The method according to claim 1 , wherein the second heading distribution is obtained based on lane directions.

4. The method according to claim 1 , wherein the posterior distribution is obtained by performing Bayes filter operations.

5. The method according to claim 1 , wherein the defining the cuboid on the 3D graph comprises selecting a peak heading from a plurality of peak headings contained in the posterior distribution that is aligned with a visual heading of the object.

6. The method according to claim 1 , wherein the defining the cuboid on the 3D graph comprises selecting a peak heading from the posterior distribution of the heading.

7. The method according to claim 6 , wherein the defining the cuboid on the 3D graph further comprises flipping the peak heading by one hundred eighty degrees when a determination is made that the peak heading and velocity of the object are within ninety degrees of each other.

8. The method according to claim 7 , wherein the defining the cuboid on the 3D graph further comprises fitting a bounding box to the 3D point cloud that is orientated to the peak heading.

9. A system, comprising:

a processor;

a non-transitory computer-readable storage medium comprising programming instructions that are configured to cause the processor to implement a method for operating a robotic system, wherein the programming instructions comprise instructions to:

infer a first heading distribution for the object from a 3D point cloud;

obtain a second heading distribution from a vector map;

obtain a posterior distribution of a heading using the first and second heading distributions;

define a cuboid on a 3D graph using the posterior distribution; and

use the cuboid to facilitate driving-related operations of a robotic system.

10. The system according to claim 9 , wherein the first heading distribution is inferred based on at least one of a minimal bounding box, an area of the minimal bounding box, a heading direction or yaw in a frame, an area of the minimal bounding box that does not have any LiDAR points, and a distance between an edge of a bounding box at a current elevation to a most outer edge.

11. The system according to claim 9 , wherein the second heading distribution is obtained based on lane directions.

12. The system according to claim 9 , wherein the posterior distribution is obtained by performing Bayes filter operations.

13. The system according to claim 9 , wherein the cuboid is defined on the 3D graph by selecting a peak heading from a plurality of peak headings contained in the posterior distribution that is aligned with a visual heading of the object.

14. The system according to claim 9 , wherein the cuboid is defined on the 3D graph by selecting a peak heading from the posterior distribution of the heading.

15. The system according to claim 14 , wherein the cuboid is defined on the 3D graph by further flipping the peak heading by one hundred eighty degrees when a determination is made that the peak heading and velocity of the object are within ninety degrees of each other.

16. The system according to claim 14 , wherein the cuboid is defined on the 3D graph by further fitting a bounding box to the 3D point cloud that is orientated to the peak heading.

17. The non-transitory computer-readable medium of claim 16 , wherein the defining the cuboid on the 3D graph comprises selecting a peak heading from a plurality of peak headings contained in the posterior distribution that is aligned with a visual heading of the object.

18. The non-transitory computer-readable medium of claim 16 , wherein the defining the cuboid on the 3D graph comprises selecting a peak heading from the posterior distribution of the heading.

19. The non-transitory computer-readable medium of claim 18 , wherein the defining the cuboid on the 3D graph further comprises flipping the peak heading by one hundred eighty degrees when a determination is made that the peak heading and velocity of the object are within ninety degrees of each other.

20. A non-transitory computer-readable medium that stores instructions that, when executed by at least one computing device, will cause the at least one computing device to perform operations comprising:

inferring a first heading distribution for the object from a 3D point cloud;

obtaining a second heading distribution from a vector map;

obtaining a posterior distribution of a heading using the first and second heading distributions;

defining a cuboid on a 3D graph using the posterior distribution; and

using the cuboid to facilitate driving-related operations of a robotic system.