Mapping off-road entries for autonomous vehicles

1. A method of identifying off-road entry lane waypoints, the method comprising:

identifying, by one or more processors, from map information a polygon representative of a driveway or parking area;

identifying, by the one or more processors, a nearest lane from the map information based on the polygon;

identifying, by the one or more processors, a plurality of lane waypoints, each of the plurality of lane waypoints corresponding to a location within at least one lane;

inputting, by the one or more processors, the polygon and the plurality of lane waypoints into a model;

selecting, by the one or more processors, a lane waypoint of the plurality of lane waypoints as an off-road entry lane waypoint;

associating, by the one or more processors, the off-road entry lane waypoint with the nearest lane; and

providing, by the one or more processors, the association to an autonomous vehicle in order to allow the autonomous vehicle to use the association to control the autonomous vehicle in an autonomous driving mode.

2. The method of claim 1 , wherein each of the plurality of lane waypoints includes a heading and identifying the plurality of lane waypoints includes identifying waypoints having headings that align with a heading of the nearest lane.

3. The method of claim 1 , wherein the model is a machine learned model.

4. The method of claim 1 , further comprising:

using observations of actual road users entering a second polygon to identify a second lane waypoint;

scoring the second lane waypoint; and

training the model using the score.

5. The method of claim 4 , wherein scoring the second lane waypoint is based on a number of observations having trajectories that pass through the second lane waypoint immediately before entering the second polygon.

6. The method of claim 4 , wherein the training the model is further based on an angle difference between a heading of the second lane waypoint and a closest edge of the second polygon.

7. The method of claim 4 , wherein the training the model is further based on a distance between the second lane waypoint and a center of the second polygon.

8. The method of claim 4 , wherein the training the model is further based on a distance between the second lane waypoint and a center of the closest edge of the second polygon.

9. The method of claim 4 , wherein the training the model is further based on a distance between the second lane waypoint and an end point of the closest edge of the second polygon.

10. The method of claim 9 , wherein the end point is one of two end points of the closest edge of the second polygon that is closest to the second lane waypoint.

11. The method of claim 4 , wherein the training the model is further based on an angle difference between the heading of the second lane waypoint and a heading of a line segment from the second lane waypoint to a center of the closest edge of the second polygon.

12. The method of claim 4 , wherein the training the model is further based on a number of other lane waypoints that are closer to their respective closest edges of the second polygon.

13. The method of claim 4 , wherein the training the model is further based on a number of other a number of other candidate lanes which are closer to their respective closest edges of the second polygon.

14. The method of claim 1 , further comprising, generating a virtual lane between the off-road entry lane waypoint and the associated nearest edge of the polygon.

15. The method of claim 14 , further comprising providing the virtual lane to an autonomous vehicle in order to allow the autonomous vehicle to use the virtual to make a prediction about behavior of another road user.

16. The method of claim 1 , further comprising:

identifying, by the one or more processors, a second nearest lane based on the polygon, the second nearest lane being an opposing lane of traffic to the nearest lane;

identifying, by the one or more processors, a second plurality of lane waypoints;

inputting, by the one or more processors, the polygon and the second plurality of lane waypoints into a second model;

selecting, by the one or more processors, a lane waypoint of the plurality of lane waypoints as a second off-road entry lane waypoint;

associating, by the one or more processors, the second off-road entry lane waypoint with the nearest lane as a second association; and

providing, by the one or more processors, the second association to an autonomous vehicle in order to allow the autonomous vehicle to use the second association to control the autonomous vehicle in an autonomous driving mode.

17. A system for identifying off-road entry lane waypoints, the system comprising:

one or more processors configured to:

identify from map information a polygon representative of a driveway or parking area;

identify a nearest lane based on the polygon;

identify a plurality of lane waypoints, each of the plurality of lane waypoints corresponding to a location within at least one lane;

input, the polygon and the plurality of lane waypoints into a model;

select a lane waypoint of the plurality of lane waypoints as an off-road entry lane waypoint;

associate the off-road entry lane waypoint with the nearest lane; and

provide the association to an autonomous vehicle in order to allow the autonomous vehicle to use the association to control the autonomous vehicle in an autonomous driving mode.

18. The system of claim 17 , wherein each of the plurality of lane waypoints includes a heading and identifying the plurality of laneway points includes identifying waypoints having headings that align with a heading of the nearest lane.

19. The system of claim 17 , wherein the model is a machine learned model.

20. The system of claim 17 , wherein the one or more processors are further configured to:

using observations of actual road users entering a second polygon to identify a second lane waypoint;

scoring the second lane waypoint; and

training the model using the score.