Persisting predicted objects for robustness to perception issues in autonomous driving

Patent metadata

Field	Value
Publication number	US-12043289-B2
Application number	US-202117404553-A
Country	US
Kind code	B2
Filing date	Aug 17, 2021
Priority date	Aug 17, 2021
Publication date	Jul 23, 2024
Grant date	Jul 23, 2024

Abstract

Systems and methods for operating an autonomous vehicle (AV) are provided. The method includes detecting one or more objects in an environment, predicting a first set of predicted object trajectories comprising one or more trajectories for each of the detected one or more objects, generating a plurality of candidate AV trajectories for the AV, scoring each of the candidate AV trajectories according to a cost function, using the scoring to select a final AV trajectory for execution, determining which of the predicted object trajectories affected the final AV trajectory and which did not do so, adding the predicted object trajectories that affected the final AV trajectory to a persisted prediction cache, excluding from the persisted prediction cache any predicted object trajectories that did not affect the final AV trajectory, and executing the final AV trajectory to cause the AV to move along the final AV trajectory.

First claim

The invention claimed is:

1. A method of operating an autonomous vehicle, method comprising:
selecting a first autonomous vehicle trajectory for execution by the autonomous vehicle;
detecting one or more objects in an environment of the autonomous vehicle;
obtaining a first set of predicted object trajectories comprising one or more first trajectories for each of the detected one or more objects;
comparing the first set of predicted object trajectories to the first autonomous trajectory;
determining, based on results of said comparing, that first ones of the predicted object trajectories affect the first autonomous vehicle trajectory and that second ones of the predicted object trajectories do not affect the first autonomous vehicle trajectory;
selecting at least one trajectory from the first ones of the predicted object trajectories based on an object classification, and/or an object location relative to a blind spot of a road that is unable to be perceived by the autonomous vehicle;
storing, in a persistent cache, the at least one trajectory which was selected from the first ones of the predicted object trajectories;
predicting, subsequent to said storing, a second set of predicted object trajectories comprising one or more second trajectories for each of the detected one or more objects;
using a motion planning model to generate a plurality of candidate autonomous vehicle trajectories for the autonomous vehicle, wherein the at least one trajectory which was stored in the persistent cache and the second trajectories in the second set of predicted object trajectories are used as inputs to the motion planning model for each of the candidate autonomous vehicle trajectories;
scoring each of the candidate autonomous vehicle trajectories according to a cost function; and
using the scoring to select a second autonomous vehicle trajectory for execution, wherein the second autonomous vehicle trajectory causes the autonomous vehicle to execute a maneuver.

2. The method of claim 1 , further comprising updating the persisted prediction cache with any predicted object trajectories from the second set that affected the second autonomous vehicle trajectory.

3. The method of claim 1 , wherein the at least one trajectory that is stored in the persistent cache and used as an input to the motion planning model is associated with one of the detected one or more objects for which the second set of predicted object trajectories is absent of a second trajectory.

4. The method of claim 1 , further comprising excluding from the persisted prediction cache any of the second trajectories of the second set of predicted object trajectories that did not affect the second autonomous vehicle trajectory.

5. The method of claim 4 , wherein excluding from the persisted prediction cache any of the second trajectories from the second set of predicted object trajectories that did not affect the second autonomous vehicle trajectory comprises storing, to the persisted prediction cache, only second trajectories that affected the second autonomous vehicle trajectory.

6. The method of claim 1 , further comprising comparing an age of said at least one trajectory stored in the persisted prediction cache against an age limit, and removing said at least one trajectory in the persisted prediction cache when said at least one trajectory has an age greater than the age limit.

7. The method of claim 1 , wherein said results of said comparing indicate whether the autonomous vehicle took or considered taking an action for an object associated with one or more of the predicted object trajectories.

8. A system comprising:
one or more sensors coupled to an autonomous vehicle and configured to detect one or more objects in an environment of the autonomous vehicle; and
a computing device comprising a processor and a memory, wherein the memory includes instructions that, when executed by the processor, will cause the processor to:
select a first autonomous vehicle trajectory for execution by the autonomous vehicle;
obtain a first set of predicted object trajectories comprising one or more first trajectories for each of the detected one or more objects;
compare the first set of predicted object trajectories to the first autonomous vehicle trajectory to facilitate a determination that first ones of the predicted object trajectories affect the first autonomous vehicle trajectory and second ones of the predicted object trajectories do not affect the first autonomous vehicle trajectory;
select at least one trajectory from the first ones of the predicted object trajectories based on an object classification and/or a blind spot of a road that is unable to be perceived by the autonomous vehicle;
store, in a persisted prediction cache, the at least one trajectory which was selected from the first ones of the predicted object trajectories;
predict, subsequent to said storing, a second set of predicted object trajectories comprising one or more second trajectories for each of the detected one or more objects;
use a motion planning model to generate a plurality of candidate autonomous vehicle trajectories for the autonomous vehicle, wherein the at least one trajectory which was stored in the persistent cache and the second trajectories in the second set of predicted object trajectories are used as inputs to the motion planning model for each of the candidate autonomous vehicle trajectories;
score each of the candidate autonomous vehicle trajectories according to a cost function;
use the scoring to select a second autonomous vehicle trajectory for execution,
determine which of the predicted object trajectories affected the second autonomous vehicle trajectory and which did not do so;
add the second trajectories that affected the second autonomous vehicle trajectory to the persisted prediction cache;
exclude from the persisted prediction cache any second trajectories that did not affect the second autonomous vehicle trajectory; and
using a motion planning system of the autonomous vehicle, execute the second autonomous vehicle trajectory to cause the autonomous vehicle to move along the second autonomous vehicle trajectory.

9. The system of claim 8 , wherein at least one trajectory that is stored in the persistent cache and used as an input to the motion planning model is associated with one of the detected one or more objects for which the second set of predicted object trajectories is absent of a second trajectory.

10. The system of claim 8 , further comprising instructions to exclude from the persisted prediction cache any second trajectories that did not affect the second autonomous vehicle trajectory comprises adding all of the trajectories of the second set to the persisted prediction cache, and then remove from the persisted prediction cache any predicted object trajectories of the second set that did not affect the second autonomous vehicle trajectory.

11. The system of claim 10 , wherein the instructions to exclude from the persisted prediction cache any second trajectories that did not affect the second autonomous vehicle trajectory comprise instructions to add to the persisted prediction cache only the predicted object trajectories of the second set that affected the second autonomous vehicle trajectory, and not any predicted object trajectories the second set that did not affect the second autonomous vehicle trajectory.

12. The system of claim 8 , further comprising instructions to cause the processor to compare an age of each first and second trajectory in the persisted prediction cache against an age limit.

13. The system of claim 12 , wherein the instructions to cause the processor to exclud

Assignees

• Argo Ai Llc

Inventors

• Seegmiller Neal
• Venator Ed
• Cunningham Christopher
• Barone Patrick
• Cai Xi

Classifications

• B60W60/0011

  involving control alternatives for a single driving scenario, e.g. planning several paths to avoid obstacles · CPC title

• B60W2556/15

  Data age · CPC title

• B60W2420/408

  Radar; Laser, e.g. lidar · CPC title

• B60W2420/403

  Image sensing, e.g. optical camera · CPC title

• B60W2556/10

  Historical data · CPC title