Surfel map editing

1. A method comprising:

obtaining initial surfel data comprising a plurality of surfels, wherein each surfel corresponds to a respective different location in an environment and each surfel represents a surface of an object;

determining a plurality of non-static surfels from the plurality of surfels;

obtaining guidance data that characterizes a prediction for static surfaces in one or more first regions of the environment, wherein each first region of the environment was obscured by the plurality of non-static surfels in the initial surfel data; and

processing the guidance data and the initial surfel data to generate final surfel data that comprises, for each first region of the environment, a plurality of predicted surfels that represent static surfaces in the first region that are partially or wholly obscured by objects represented by respective non-static surfels.

2. The method of claim 1 , wherein:

each surfel has associated data that includes a static label characterizing a predicted level of permanence of the surface represented by the surfel; and

each static surfel has a static label that satisfies a first threshold.

3. The method of claim 1 , wherein processing the guidance data and the initial surfel data to generate the final surfel data comprises:

removing the plurality of non-static surfels from the initial surfel data to generate updated surfel data; and

processing the guidance data and the initial surfel data to generate the final surfel data.

4. The method of claim 3 , wherein the final surfel data further comprises the surfels in the updated surfel data.

5. The method of claim 1 , wherein:

the initial surfel data has been generated from sensor data captured by one or more sensors on-board respective vehicles operating in the environment at respective previous time points; and

the one or more first regions in the environment where obscured from a field of view of the one or more sensors by the plurality of non-static surfels.

6. The method of claim 5 , wherein the predicted surfels represent static surfaces that have not been observed in any sensor observations of a vehicle of a plurality of vehicles operating in the environment.

7. The method of claim 1 , wherein each predicted surfel has a static label that satisfies a second threshold.

8. The method of claim 1 , wherein processing the guidance data and the updated surfel data comprises processing the guidance data and the updated surfel data using guided interpolation.

9. The method of claim 1 , wherein obtaining guidance data comprises:

determining a predicted type of the predicted static surfaces, and

obtaining guidance data corresponding to the determined predicted type.

10. The method of claim 1 , wherein each surfel in the initial surfel data comprises:

coordinates in a three-dimensional coordinate system of the environment that characterize a position of the surfel in the environment, and

a normal vector that characterizes an orientation of the surfel in the environment.

11. The method of claim 1 , wherein the initial surfel data comprises data characterizing a voxel gird, wherein each surfel in the surfel data corresponds to a different voxel in the voxel grid.

12. A system comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:

obtaining initial surfel data comprising a plurality of surfels, wherein each surfel corresponds to a respective different location in an environment and each surfel represents a surface of an object;

determining a plurality of non-static surfels from the plurality of surfels;

obtaining guidance data that characterizes a prediction for static surfaces in one or more first regions of the environment, wherein each first region of the environment was obscured by the plurality of non-static surfels in the initial surfel data; and

processing the guidance data and the initial surfel data to generate final surfel data that comprises, for each first region of the environment, a plurality of predicted surfels that represent static surfaces in the first region that are partially or wholly obscured by objects represented by respective non-static surfels.

13. The system of claim 12 , wherein:

each surfel has associated data that includes a static label characterizing a predicted level of permanence of the surface represented by the surfel; and

each static surfel has a static label that satisfies a first threshold.

14. The system of claim 12 , wherein processing the guidance data and the initial surfel data to generate the final surfel data comprises:

removing the plurality of non-static surfels from the initial surfel data to generate updated surfel data; and

processing the guidance data and the initial surfel data to generate the final surfel data.

15. The system of claim 12 , wherein:

the initial surfel data has been generated from sensor data captured by one or more sensors on-board respective vehicles operating in the environment at respective previous time points; and

the one or more first regions in the environment where obscured from a field of view of the one or more sensors by the plurality of non-static surfels.

16. The system of claim 12 , wherein processing the guidance data and the updated surfel data comprises processing the guidance data and the updated surfel data using guided interpolation.

17. One or more non-transitory computer storage media encoded with computer program instructions that when executed by a plurality of computers cause the plurality of computers to perform operations comprising:

obtaining initial surfel data comprising a plurality of surfels, wherein each surfel corresponds to a respective different location in an environment and each surfel represents a surface of an object;

determining a plurality of non-static surfels from the plurality of surfels;

obtaining guidance data that characterizes a prediction for static surfaces in one or more first regions of the environment, wherein each first region of the environment was obscured by the plurality of non-static surfels in the initial surfel data; and

processing the guidance data and the initial surfel data to generate final surfel data that comprises, for each first region of the environment, a plurality of predicted surfels that represent static surfaces in the first region that are partially or wholly obscured by objects represented by respective non-static surfels.

18. The non-transitory computer storage media of claim 17 , wherein processing the guidance data and the initial surfel data to generate the final surfel data comprises:

removing the plurality of non-static surfels from the initial surfel data to generate updated surfel data; and

processing the guidance data and the initial surfel data to generate the final surfel data.

19. The non-transitory computer storage media of claim 17 , wherein:

the initial surfel data has been generated from sensor data captured by one or more sensors on-board respective vehicles operating in the environment at respective previous time points; and

the one or more first regions in the environment where obscured from a field of view of the one or more sensors by the plurality of non-static surfels.

20. The non-transitory computer storage media of claim 17 , wherein processing the guidance data and the updated surfel data comprises processing the guidance data and the updated surfel data using guided interpolation.