Conditional motion predictions

Patent metadata

Field	Value
Publication number	US-11814075-B2
Application number	US-202017003965-A
Country	US
Kind code	B2
Filing date	Aug 26, 2020
Priority date	Aug 26, 2020
Publication date	Nov 14, 2023
Grant date	Nov 14, 2023

Abstract

Among other things, techniques are described for conditional motion predictions. The techniques include generating, by a planning circuit, a set of candidate trajectories for a vehicle based on possible macro actions by the vehicle, predicting, by the planning circuit and for at least some trajectories in the set of candidate trajectories, a response by a target vehicle to the respective trajectory and a probability of the response by the target vehicle, selecting, by the planning circuit, a trajectory from the set of candidate trajectories based at least in part on the response by the target vehicle to the trajectory, the probability of the response by the target vehicle, and characteristics of the trajectory, and operating, by a control circuit, the vehicle based on the selected trajectory.

First claim

What is claimed is:

1. A vehicle, comprising:
a computer-readable media storing computer-executable instructions; and
a processor communicatively coupled to the computer-readable media, the processor configured to execute the computer executable instructions to perform operations including:
receiving at least one predefined macro action for the vehicle, the at least one predefined macro action including at least one parameter;
generating a set of candidate trajectories for the vehicle based on one or more values for the at least one parameter;
for at least some trajectories in the set of candidate trajectories,
predicting a response by a target vehicle to a respective trajectory and a probability of the response by the target vehicle, and
determining a score for the respective trajectory based on a weighted combination of:
(i) the predicted response by the target vehicle to the respective trajectory and the probability of the response by the target vehicle, and
(ii) at least one heuristic calculated for the respective trajectory, wherein calculating the at least one heuristic for the respective trajectory comprises determining a distance before a collision for the respective trajectory, and wherein a higher score is assigned to a trajectory having a greater distance before the collision relative to distances before collisions of other trajectories in the set of candidate trajectories;

selecting a trajectory from the set of candidate trajectories based at least in part on the score; and
providing data to cause operation of the vehicle based on the selected trajectory.

2. The vehicle of claim 1 , comprising:
identifying a maneuver to be performed by the vehicle on a road network; and
generating the set of candidate trajectories for performing the maneuver on the road network.

3. The vehicle of claim 2 , wherein the maneuver comprises at least one of a merge, a lane change, or a turn.

4. The vehicle of claim 1 , wherein the at least one predefined macro action for the vehicle includes a lane follow action having a parameterized velocity.

5. The vehicle of claim 1 , wherein the at least one predefined macro action for the vehicle includes a change lane action having a parameterized lane change time.

6. The vehicle of claim 1 , wherein predicting the response of the target vehicle to the trajectory comprises predicting a trajectory of the target vehicle conditioned on the trajectory of the vehicle.

7. The vehicle of claim 1 , wherein selecting the trajectory from the set of candidate trajectories comprises:
determining a score for each trajectory in the set of candidate trajectories, wherein the score is determined based at least in part on the response by the target vehicle to the trajectory, the probability of the response by the target vehicle, and characteristics of the trajectory; and
selecting the trajectory from the set of candidate trajectories based on the score.

8. The vehicle of claim 7 , wherein determining the score comprises identifying a magnitude of acceleration by the vehicle for the trajectory or by the target vehicle in response to the trajectory.

9. The vehicle of claim 7 , wherein determining the score comprises identifying a collision or near-collision by the vehicle for the trajectory or by the target vehicle in response to the trajectory.

10. The vehicle of claim 7 , wherein determining the score comprises determining a time-to-collision or distance-to-collision for the vehicle for the trajectory or for the target vehicle in response to the trajectory.

11. The vehicle of claim 7 , wherein determining the score comprises determining a distance traveled by the vehicle for the trajectory or by the target vehicle in response to the trajectory.

12. The vehicle of claim 1 , wherein the target vehicle comprises a first target vehicle, the processor configured to execute the computer executable instructions to perform operations including:
for each trajectory in the set of candidate trajectories, predicting a response by a second target vehicle to the trajectory and a probability of the response by the second target vehicle, wherein the second target vehicle is different than the first target vehicle.

13. The vehicle of claim 12 , comprising selecting the trajectory from the set of candidate trajectories based at least in part on the response by the first target vehicle to the trajectory, the probability of the response by the first target vehicle, the response by the second target vehicle to the trajectory, the probability of the response by the second target vehicle, and characteristics of the trajectory.

14. The vehicle of claim 1 , wherein the values for the at least one parameter of the at least one predefined macro action are selected based on characteristics of the vehicle or an environment of the vehicle.

15. The vehicle of claim 1 , wherein the processor is configured to execute the computer executable instructions to perform operations including:
identifying a plurality of vehicles within a proximity of the vehicle; and
filtering the plurality of vehicles to identify the target vehicle before predicting the response by the target vehicle to the respective trajectory.

16. The vehicle of claim 15 , wherein filtering the plurality of vehicles to identify the target vehicle comprises determining that the target vehicle is within at least one of a threshold distance or direction of the vehicle at one or more points along the respective trajectory.

17. The vehicle of claim 1 , wherein predicting the response by the target vehicle to the respective trajectory and the probability of the response by the target vehicle comprises inputting scene data for the target vehicle, state data for the target vehicle, and the respective trajectory into a model configured to output a predicted trajectory by the target vehicle and a probability of the predicted trajectory by the target vehicle based on the scene data, the state data, and the respective trajectory.

18. The vehicle of claim 17 , wherein each of the scene data for the target vehicle and the respective trajectory input into the model comprise a birds-eye-view image.

19. A method comprising:
receiving, by at least one processor, at least one predefined macro action for a vehicle, the at least one predefined macro action including at least one parameter;
generating, by the at least one processor, a set of candidate trajectories for the vehicle based on one or more values for the at least parameter;
for at least some trajectories in the set of candidate trajectories,
predicting, by at least one processor, a response by a target vehicle to a respective trajectory and a probability of the response by the target vehicle, and
determining a score for the respective trajectory based on a weighted combination of:
(i) the predicted response by the target vehicle to the respective trajectory and the probability of the response by the target vehicle, and
(ii) at least one heuristic calculated for the respective trajectory, wherein calculating the at least one heuristic for the respective trajectory comprises determining a distance before a collision for the respective trajectory, and wherein a higher score is assigned to a trajectory having a greater distance before the collision relative to distances before collisions of other trajectories in the set of candidate trajectories;

selecting, by the at least one processor, a trajectory from the set of candidate trajectories based at least in part on the score; and
providing data to cause operation of the vehi

Assignees

• Motional Ad Llc

Inventors

• Wolff Eric
• Katt Sammie
• Cserna Bence

Classifications

• B60W2554/4045

  Intention, e.g. lane change or imminent movement · CPC title

• B60W40/10

  related to vehicle motion · CPC title

• B60W60/0011 Primary

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

• B60W60/0027 Primary

  using trajectory prediction for other traffic participants · CPC title

• B60W30/095

  Predicting travel path or likelihood of collision · CPC title