Systems and methods to track vehicles proximate perceived by an autonomous vehicle

Patent metadata

Field	Value
Publication number	US-10229510-B2
Application number	US-201816039864-A
Country	US
Kind code	B2
Filing date	Jul 19, 2018
Priority date	Mar 30, 2017
Publication date	Mar 12, 2019
Grant date	Mar 12, 2019

Abstract

The present disclosure provides systems and methods for tracking vehicles or other objects that are perceived by an autonomous vehicle. A vehicle filter can employ a motion model that models the location of the tracked vehicle using a vehicle bounding shape and an observation model that generates an observation bounding shape from sensor observations. A dominant vertex or side from each respective bounding shape can be identified and used to update or otherwise correct one or more predicted shape locations associated with the vehicle bounding shape based on a shape location associated with the observation bounding shape.

First claim

What is claimed is:

1. A computer-implemented method for tracking vehicles that are perceived by an autonomous vehicle, the method comprising:
obtaining, by one or more computing devices included in the autonomous vehicle, state data descriptive of a previous state of a tracked vehicle that is perceived by the autonomous vehicle, wherein the previous state of the tracked vehicle comprises at least one previous location of the tracked vehicle or a vehicle bounding shape associated with the tracked vehicle;
predicting, by the one or more computing devices, one or more first locations for a first portion of the vehicle bounding shape based at least in part on the state data;
obtaining, by the one or more computing devices, observation data descriptive of one or more sensor observations;
determining, by the one or more computing devices, a second location of a second portion of an observation bounding shape associated with the tracked vehicle based at least in part on the observation data; and
determining, by the one or more computing devices, at least one estimated current location of the first portion of the vehicle bounding shape associated with the tracked vehicle based at least in part on a comparison of the one or more first locations for the first portion of the vehicle bounding shape to the second location of the second portion of the observation bounding shape.

2. The computer-implemented method of claim 1 , wherein determining, by the one or more computing devices, the at least one estimated current location of the first portion of the vehicle bounding shape associated with the tracked vehicle comprises assigning, by the one or more computing devices, a respective weight to each of the one or more first locations based at least in part on the comparison of the one or more first locations for the vehicle bounding shape to the second location of the observation bounding shape, the respective weight assigned to each first location indicative of an amount of agreement between such first location and the second location of the observation bounding shape.

3. The computer-implemented method of claim 1 , wherein predicting, by the one or more computing devices, the one or more first locations for the first portion of the vehicle bounding shape comprises:
predicting, by the one or more computing devices, one or more first shape locations for the vehicle bounding shape; and
respectively identifying, by the one or more computing devices, a nearest unoccluded vertex of each of the one or more first shape locations.

4. The computer-implemented method of claim 3 , wherein predicting, by the one or more computing devices, the one or more first locations for the first portion of the vehicle bounding shape further comprises, for each of the one or more first shape locations:
identifying, by the one or more computing devices, a dominant side that includes the nearest unoccluded vertex and corresponds to a length of the tracked vehicle.

5. The computer-implemented method of claim 3 , wherein respectively identifying, by the one or more computing devices, the nearest unoccluded vertex of each of the one or more first shape locations comprises:
obtaining, by the one or more computing devices, an occlusion map that describes occluded and unoccluded areas; and
for each of the one or more first shape locations, iteratively analyzing, by the one or more computing devices, in order of proximity to the autonomous vehicle, one or more nearest vertices of such bounding shape relative to the occlusion map until an unoccluded vertex is identified.

6. The computer-implemented method of claim 3 , wherein the method further comprises:
when the respective nearest unoccluded vertex cannot be identified for one of the first shape locations:
selecting, by the one or more computing devices, a nearest occluded vertex; and
increasing, by the one or more computing devices, a covariance value associated with the vehicle bounding shape.

7. The computer-implemented method of claim 1 , wherein:
obtaining, by the one or more computing devices, the observation data descriptive of the one or more sensor observations comprises obtaining, by the one or more computing devices, the observation data descriptive of a plurality of observed points; and
determining, by the one or more computing devices, the second location of the second portion of the observation bounding shape comprises fitting, by the one or more computing devices, the observation bounding shape around the plurality of observed points such that the fitted observation bounding shape includes all of the plurality of observed points and minimizes a collective distance from the plurality of points to respective closest sides of the observation bounding shape.

8. The computer-implemented method of claim 1 , wherein determining, by the one or more computing devices, the second location of the second portion of the observation bounding shape associated with the tracked vehicle comprises:
identifying, by the one or more computing devices, the second vertex that corresponds to the first vertex of the vehicle bounding polygon; and
selecting, by the one or more computing devices, the second side for the observation bounding shape that includes the second vertex and is closest in orientation to a predicted heading associated with at least one of the one or more first shape locations.

9. The computer-implemented method of claim 1 , wherein:
the vehicle bounding shape has a first size defined by one or more first values respectively for one or more size parameters;
the observation bounding shape has a second size defined by one or more second values respectively for the one or more size parameters; and
the method further comprises:
comparing, by the one or more computing devices, the second value for each size parameter to the first value for such size parameter; and
for at least one size parameter for which the second value is greater than the first value, updating, by the one or more computing devices, the first value for such size parameter to equal the second value.

10. The computer-implemented method of claim 9 , wherein the vehicle bounding shape comprises a vehicle bounding rectangle, the observation bounding shape comprises an observation bounding rectangle, and the one or more size parameters comprise a width and a length.

11. The computer-implemented method of claim 1 , wherein the first portion of the vehicle bounding shape comprises a first vertex or a first side of the vehicle bounding shape.

12. The computer-implemented method of claim 11 , wherein the second portion of the observation bounding shape comprises a second vertex or a second side of the observation bounding shape.

13. A computer system, comprising:
one or more processors; and
one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computer system to perform operations, the operations comprising:
predicting, by the one or more computing devices, one or more first shape locations for a vehicle bounding shape associated with a tracked vehicle that is perceived by an autonomous vehicle;
identifying, by the one or more computing devices, a first dominant portion for each of the one or more first shape locations predicted for the vehicle bounding shape;
obtaining, by the one or more computing devices, observation data descriptive of one or more sensor observations;
determining, by the one or more computing devices, a second shape location for an observation bounding shape associated with the tracked vehicle based at least in part on the observation data;
iden

Assignees

• Uber Technologies Inc

Inventors

• Becker Brian C
• Bagnell J Andrew
• Venkatraman Arunprasad
• Lakshmanan Karthik

Classifications

• G06V20/56 Primary

  exterior to a vehicle by using sensors mounted on the vehicle · CPC title

• G06T7/60 Primary

  Analysis of geometric attributes · CPC title

• G06V10/457

  by analysing connectivity, e.g. edge linking, connected component analysis or slices · CPC title

• G08B13/19608

  Tracking movement of a target, e.g. by detecting an object predefined as a target, using target direction and or velocity to predict its new position · CPC title

• G06K9/00791

  Physics · mapped topic