Localization of robotic vehicles

Patent metadata

Field	Value
Publication number	US-10866102-B2
Application number	US-201615390460-A
Country	US
Kind code	B2
Filing date	Dec 23, 2016
Priority date	Dec 23, 2016
Publication date	Dec 15, 2020
Grant date	Dec 15, 2020

Abstract

An example method includes determining locations of a plurality of candidate landmarks in relation to a robot based on sensor data from at least one sensor on the robot. The method further includes determining a plurality of sample sets, wherein each sample set comprises a subset of the plurality of candidate landmarks and a plurality of corresponding mapped landmarks. The method also includes determining a transformation for each sample set that relates the candidate landmarks from the subset to the corresponding mapped landmarks. The method additionally includes applying the determined transformation for each sample set to the plurality of candidate landmarks to determine a number of inliers associated with each sample set based on distances between the transformed plurality of candidate landmarks and a plurality of neighbouring mapped landmarks. The method further includes selecting a sample set from the plurality based on the number of inliers associated with each sample set. The method still further includes estimating a pose of the robot based on the selected sample set.

First claim

What is claimed is:

1. A method comprising:
determining locations of a plurality of candidate landmarks in relation to a robot based on sensor data from at least one sensor of the robot;
determining a plurality of sample sets, wherein each sample set comprises a subset of the plurality of candidate landmarks and a plurality of corresponding mapped landmarks;
for each sample set of the plurality of sample sets:
determining a transformation that relates the candidate landmarks from the subset of the plurality of candidate landmarks to the corresponding mapped landmarks of the plurality of corresponding mapped landmarks;
applying the determined transformation to the plurality of candidate landmarks; and
determining a number of inliers associated with the sample set based on distances between the transformed plurality of candidate landmarks and a plurality of neighbouring mapped landmarks;

selecting a sample set from the plurality of sample sets based on the determined number of inliers associated with each sample set from the plurality of sample sets; and
estimating a pose of the robot based on the selected sample set.

2. The method of claim 1 , further comprising navigating the robot within an environment of the robot based on the estimated pose.

3. The method of claim 1 , wherein determining the plurality of sample sets comprises identifying, based on an initial pose estimate of the robot, candidate landmarks that are within a correspondence distance threshold of corresponding mapped landmarks, wherein each sample set comprises a subset of the identified candidate landmarks and corresponding mapped landmarks.

4. The method of claim 1 , further comprising determining a pose estimation confidence associated with the estimated pose based on a ratio of the determined number of inliers associated with the selected sample set to a total number of candidate landmarks in the plurality of candidate landmarks, wherein a higher ratio of inliers to candidate landmarks indicates fewer false detections in the plurality of candidate landmarks.

5. The method of claim 1 , further comprising determining an accuracy of the estimated pose, wherein determining the accuracy comprises comparing the pose estimate to statistical data that represents past pose estimates.

6. The method of claim 1 , wherein selecting the sample set from the plurality comprises sequentially determining the number of inliers for each of the sample sets until one of the sample sets is determined to have a number of inliers that meets or exceeds an inlier threshold value.

7. The method of claim 1 , wherein determining the plurality of sample sets comprises:
selecting a first corresponding mapped landmark for each sample set; and
selecting subsequent corresponding mapped landmarks for each sample set based on positions of the subsequent corresponding mapped landmarks relative to the selected first corresponding mapped landmark.

8. The method of claim 1 , wherein determining the locations of the plurality of candidate landmarks in relation to the robot comprises receiving the locations from at least one sensor of the robot.

9. The method of claim 1 , wherein determining the locations of the plurality of candidate landmarks in relation to the robot comprises identifying signals from the at least one sensor of the robot that have an intensity greater than an intensity threshold value.

10. The method of claim 1 , further comprising:
determining a refined transformation that relates the inliers associated with the selected sample set to neighbouring mapped landmarks from the plurality of mapped landmarks; and
estimating the pose of the robot based on the refined transformation.

11. The method of claim 1 , wherein determining the plurality of sample sets comprises selecting three candidate landmarks for each sample set from the plurality of candidate landmarks based on the locations of the plurality of candidate landmarks in relation to the robot.

12. The method of claim 1 , wherein determining the transformation for each sample set that relates the candidate landmarks from the sample set to the corresponding mapped landmarks comprises estimating a shifted location and orientation of the candidate landmarks from the sample set to align with the corresponding mapped landmarks.

13. The method of claim 1 , wherein each of the plurality of candidate landmarks correspond to either a mapped landmark or a false detection, and wherein false detections comprise detected signals that are not associated with a landmark or that misrepresent the location of a landmark, the method further comprising:
concurrently with selecting the sample set from the plurality of sample sets, filtering out false detections from the plurality of candidate landmarks based on the determined number of inliers associated with each sample set.

14. The method of claim 1 , wherein selecting the sample set from the plurality of sample sets comprises:
determining that two or more sample sets result in the same number of inliers; and
responsive to determining that the two or more sample sets result in the same number of inliers, selecting a sample set from the two or more samples sets that has a lowest net distance between the transformed candidate landmarks and the mapped landmarks.

15. The method of claim 1 , wherein estimating the pose of the robot based on the selected sample set comprises estimating the pose of the robot based on a transformation of the inliers associated with the selected sample set.

16. A non-transitory computer readable medium having stored therein instructions executable by one or more processors to cause a computing system to perform the functions comprising:
determining locations of a plurality of candidate landmarks in relation to a robot based on sensor data from at least one sensor of the robot;
determining a plurality of sample sets, wherein each sample set comprises a subset of the plurality of candidate landmarks and a plurality of corresponding mapped landmarks;
for each sample set of the plurality of sample sets:
determining a transformation that relates the candidate landmarks from the subset of the plurality of candidate landmarks to the corresponding mapped landmarks of the plurality of corresponding mapped landmarks;
applying the determined transformation to the plurality of candidate landmarks; and
determining a number of inliers associated with the sample set based on distances between the transformed plurality of candidate landmarks and a plurality of neighbouring mapped landmarks;

selecting a sample set from the plurality of sample sets based on the determined number of inliers associated with each sample set from the plurality of sample sets; and
estimating a pose of the robot based on the selected sample set.

17. The non-transitory computer readable medium of claim 16 , further having stored therein a map of landmark locations, wherein the functions further comprise:
determining the plurality of corresponding mapped landmarks for each sample set based on the map of landmark locations.

18. The non-transitory computer readable medium of claim 16 , further having stored therein statistical information associated with the corresponding mapped landmarks, wherein the statistical information relates to past pose estimates of the robot, wherein the functions further comprise:
determining an accuracy of the pose estimate, wherein determining the accuracy comprises comparing the pose estimate associated with the selected sample set to the statistical information.

19.

Assignees

• X Dev Llc

Inventors

• Holz Dirk

Classifications

• G05D2111/50

  Internal signals, i.e. from sensors located in the vehicle, e.g. from compasses or angular sensors · CPC title

• G05D2111/10

  Optical signals · CPC title

• G05D2101/15

  using machine learning, e.g. neural networks · CPC title

• B25J9/162

  Mobile manipulator, movable base with manipulator arm mounted on it · CPC title

• B25J5/007

  mounted on wheels · CPC title