System and method of actor-based map attribute generation

1 . A computer-implemented method that, when executed by data processing hardware, causes the data processing hardware to perform operations comprising:

gathering sensor data via a sensor system of a host vehicle that is an autonomous vehicle;

processing perception data via a perception system of the host vehicle;

maintaining onboard map data of the host vehicle;

determining a position of the host vehicle;

evaluating the sensor data, the perception data, and the onboard map data via an online mapping module comprising an actor-based reasoner module;

determining, via the actor-based reasoner module, one or more detected map attributes based on the perception data concerning behavior and trajectories of one or more road actors, the one or more detected map attributes including geometrical road attributes and semantic road attributes;

fusing the one or more detected map attributes with existing map attributes;

updating the onboard map data with the one or more detected map attributes;

controlling the host vehicle based on the updated onboard map data; and

determining, via the actor-based reasoner module, inferred lane segments over a moving time window by: (i) evaluating the trajectories of the one or more road actors, (ii) computing left and right normalized perpendicular vectors with respect to a motion vector derived from the trajectories, (iii) projecting along the normalized perpendicular vectors to form the inferred lane segments, and (iv) connecting the inferred lane segments to previously inferred segments.

2 . The method of claim 1 , further comprising determining quality of the existing map attributes based on the one or more detected map attributes.

3 . The method of claim 2 , further comprising communicating the one or more detected map attributes to a cloud database for use by other vehicles.

4 . The method of claim 1 , wherein the online mapping module further comprises an online map generator module comprising a segmentizer configured to generate lane segments and a connectivity graph by detecting shared segments, a classifier configured to classify map elements in the connectivity graph, and a scene builder and publisher configured to fuse the one or more detected map attributes with the onboard map data and output a high-definition map for consumption by a planning module of the host vehicle.

5 . The method of claim 1 , wherein fusing the one or more detected map attributes with the existing map attributes further comprises providing high-definition map data configured for a planning module of the host vehicle.

6 . The method of claim 1 , wherein the actor-based reasoner module comprises a connectivity inference module configured to generate a connectivity sequence from a connectivity graph produced by a segmentizer module that evaluates the onboard map data and the position of the host vehicle.

7 . The method of claim 1 , further comprising detecting anomalies of the onboard map data by comparing the detected semantic road attributes with existing semantic attributes of a road and reporting the anomalies to a cloud system.

8 . The method of claim 1 , further comprising detecting anomalies of the onboard map data by comparing the detected geometric road attributes with existing geometric attributes of a road and reporting the anomalies to a cloud system.

9 . The method of claim 1 , further comprising cancelling, via the actor-based reasoner module, lane offset by performing actor-perception alignment when a distance between a closest perception point and an inferred edge point is less than or equal to a first threshold and performing actor-actor alignment when a distance between a closest aligned inferred point and the inferred edge point is less than or equal to a second threshold.

10 . A system comprising:

data processing hardware; and

memory hardware in communication with the data processing hardware, the memory hardware storing instructions that, when executed on the data processing hardware, cause the data processing hardware to perform operations comprising:

gathering sensor data via a sensor system of a host vehicle that is an autonomous vehicle;

processing perception data via a perception system of the host vehicle;

maintaining onboard map data of the host vehicle;

determining a position of the host vehicle;

evaluating the sensor data, the perception data, and the onboard map data via an online mapping module comprising an actor-based reasoner module;

determining, via the actor-based reasoner module, one or more detected map attributes based on the perception data concerning behavior and trajectories of one or more road actors, the one or more detected map attributes including geometrical road attributes and semantic road attributes;

fusing the one or more detected map attributes with existing map attributes;

updating the onboard map data with the one or more detected map attributes;

controlling the host vehicle based on the updated onboard map data; and

determining, via the actor-based reasoner module, inferred lane segments over a moving time window by: (i) evaluating the trajectories of the one or more road actors, (ii) computing left and right normalized perpendicular vectors with respect to a motion vector derived from the trajectories, (iii) projecting along the normalized perpendicular vectors to form the inferred lane segments, and (iv) connecting the inferred lane segments to previously inferred segments.

11 . The system of claim 10 , further comprising determining quality of the existing map attributes based on the one or more detected map attributes.

12 . The system of claim 11 , further comprising communicating the one or more detected map attributes to a cloud database for use by other vehicles.

13 . The system of claim 10 , further comprising cancelling, via the actor-based reasoner module, lane offset by performing actor-perception alignment when a distance between a closest perception point and an inferred edge point is less than or equal to a first threshold and performing actor-actor alignment when a distance between a closest aligned inferred point and the inferred edge point is less than or equal to a second threshold.

14 . A vehicle management system comprising:

a sensor system;

a perception system;

an onboard map module;

an online mapping module, comprising:

an actor-based reasoner module, and

an online map generator module;

data processing hardware; and

memory hardware in communication with the data processing hardware, the memory hardware storing instructions that, when executed on the data processing hardware, cause the data processing hardware to perform operations comprising:

gathering sensor data via the sensor system of a host vehicle that is an autonomous vehicle;

processing perception data via the perception system of the host vehicle;

maintaining, via the onboard map module, onboard map data of the host vehicle;

determining a position of the host vehicle;

evaluating the sensor data, the perception data, and the onboard map data via the online mapping module;

determining, via the actor-based reasoner module, one or more detected map attributes based on the perception data concerning behavior and trajectories of one or more road actors, the one or more detected map attributes including geometrical road attributes and semantic road attributes;

fusing, via the online map generator module, the one or more detected map attributes with existing map attributes;

updating the onboard map data with the one or more detected map attributes;

controlling the host vehicle based on the updated onboard map data; and

determining, via the actor-based reasoner module, inferred lane segments over a moving time window by: (i) evaluating the trajectories of the one or more road actors, (ii) computing left and right normalized perpendicular vectors with respect to a motion vector derived from the trajectories, (iii) projecting along the normalized perpendicular vectors to form the inferred lane segments, and (iv) connecting the inferred lane segments to previously inferred segments.

15 . The vehicle management system of claim 14 , further comprising determining quality of the existing map attributes based on the one or more detected map attributes.

16 . The vehicle management system of claim 15 , further comprising communicating the one or more detected map attributes to a cloud database for use by other vehicles.

17 . The vehicle management system of claim 14 , further comprising cancelling, via the actor-based reasoner module, lane offset by performing actor-perception alignment when a distance between a closest perception point and an inferred edge point is less than or equal to a first threshold and performing actor-actor alignment when a distance between a closest aligned inferred point and the inferred edge point is less than or equal to a second threshold.