Use of prior maps for estimation of lane boundaries
View Patent ↗Disclosed herein are methods and systems for using prior maps for estimation of lane boundaries or other features within an environment. An example method may include receiving a location of a plurality of detected points on a roadway in an environment of an autonomous vehicle, determining, from a prior map of the roadway, a location of a plurality of reference points from a boundary marker on the roadway that correspond to the detected points on the roadway, determining distances between the detected points and the corresponding reference points based on the location of the detected points in the environment and the location of the reference points from the prior map of the roadway, determining a confidence buffer representing a threshold amount of variation associated with the prior map based at least in part on the distances between the detected points and the corresponding reference points, selecting one or more of the detected points such that the distance between a selected detected point and a corresponding reference point is less than the confidence buffer, and using the selected points to direct the autonomous vehicle along the roadway.
1. A computer-implemented method, comprising:
receiving a location of a plurality of detected points on a roadway in an environment of an autonomous vehicle;
determining, from a prior map of the roadway, a location of a plurality of reference points from a boundary marker on the roadway that correspond to the detected points on the roadway;
determining distances between the detected points and the corresponding reference points based on the location of the detected points in the environment and the location of the reference points from the prior map of the roadway;
determining a confidence buffer representing a threshold amount of variation associated with the prior map based at least in part on the distances between the detected points and the corresponding reference points;
selecting one or more of the detected points such that the distance between a selected detected point and a corresponding reference point is less than the confidence buffer, wherein the selected points are a current representation of the boundary marker on the roadway; and
using the selected points to direct the autonomous vehicle along the roadway.
2. The method of claim 1 , wherein determining the location of the plurality of reference points comprises selecting a plurality of points from a continuous curve corresponding to the boundary marker on the prior map.
3. The method of claim 1 , wherein determining the confidence buffer comprises:
determining a confidence metric associated with a level of confidence in the prior map;
periodically updating the confidence metric based at least in part on distances between detected points and corresponding reference points; and
determining the confidence buffer as a function of the confidence metric.
4. The method of claim 1 , wherein determining the confidence buffer comprises assigning a greater weight to detected points that were detected more recently.
5. The method of claim 1 , wherein determining the confidence buffer comprises assigning a weight to detected points based on a distance of the detected points to a sensor which detected the points.
6. The method of claim 1 , further comprising determining the confidence buffer based on a percentage of detected points that were not selected.
7. The method of claim 1 , further comprising:
determining a trend in the distances between detected points and corresponding reference points; and
based on a distance between a given detected point and a corresponding reference point following the trend, selecting the given detected point.
8. A vehicle, comprising:
a computing system configured to:
determine a location of a plurality of detected points on a roadway in an environment based at least in part on sensor data from at least one sensor on the vehicle;
determine, from a prior map of the roadway, a location of a plurality of reference points from a boundary marker on the roadway that correspond to the detected points on the roadway;
determine distances between the detected points and the corresponding reference points based on the location of the detected points in the environment and the location of the reference points from the prior map of the roadway;
determine a confidence buffer representing a threshold amount of variation associated with the prior map based at least in part on the distances between the detected points and the corresponding reference points;
select one or more of the detected points such that the distance between a selected detected point and a corresponding reference point is less than the confidence buffer, wherein the selected points are a current representation of the boundary marker on the roadway; and
use the selected points to direct the vehicle along the roadway.
9. The vehicle of claim 8 , wherein the computing system is configured to determine the location of the plurality of reference points by selecting a plurality of points from a continuous curve corresponding to the boundary marker on the prior map.
10. The vehicle of claim 8 , wherein the computing system is configured to determine the confidence buffer by:
determining a confidence metric associated with a level of confidence in the prior map;
periodically updating the confidence metric based at least in part on distances between detected points and corresponding reference points; and
determining the confidence buffer as a function of the confidence metric.
11. The vehicle of claim 8 , wherein the computing system is configured to determine the confidence buffer by assigning a greater weight to detected points that were detected more recently.
12. The vehicle of claim 8 , wherein the computing system is configured to determine the confidence buffer by assigning a weight to detected points based on a distance of the detected points to the vehicle.
13. The vehicle of claim 8 , wherein the computing system is configured to determine the confidence buffer based on a percentage of detected points that were not selected.
14. The vehicle of claim 8 , wherein the computing system is further configured to:
determine a trend in the distances between detected points and corresponding reference points; and
based on a distance between a given detected point and a corresponding reference point following the trend, select the given detected point.
15. A non-transitory computer readable medium having stored therein instructions, that when executed by a computing system, cause the computing system to perform functions comprising:
receiving a location of a plurality of detected points on a roadway in an environment of an autonomous vehicle;
determining, from a prior map of the roadway, a location of a plurality of reference points from a boundary marker on the roadway that correspond to the detected points on the roadway;
determining distances between the detected points and the corresponding reference points based on the location of the detected points in the environment and the location of the reference points from the prior map of the roadway;
determining a confidence buffer representing a threshold amount of variation associated with the prior map based at least in part on the distances between the detected points and the corresponding reference points;
selecting one or more of the detected points such that the distance between a selected detected point and a corresponding reference point is less than the confidence buffer, wherein the selected points are a current representation of the boundary marker on the roadway; and
using the selected points to direct the autonomous vehicle along the roadway.
16. The non-transitory computer readable medium of claim 15 , wherein determining the location of the plurality of reference points comprises selecting a plurality of points from a continuous curve corresponding to the boundary marker on the prior map.
17. The non-transitory computer readable medium of claim 15 , wherein determining the confidence buffer comprises:
determining a confidence metric associated with a level of confidence in the prior map;
periodically updating the confidence metric based at least in part on distances between detected points and corresponding reference points; and
determining the confidence buffer as a function of the confidence metric.
18. The non-transitory computer readable medium of claim 15 , wherein determining the confidence buffer comprises assigning a greater weight to detected points that were detected more recently.
19. The non-transitory computer readable medium of claim 15 , wherein determining the confidence buffer comprises assigning a weight to detected points based on a distance of the detected points to a sensor which detected the points.
20. The non-transitory computer readable medium of claim 15 , further comprising instructions, that when executed by a computing system, cause the computing system to perform a function comprising:
determining the confidence buffer based on a percentage of detected points that were not selected.