Intention prediction in symmetric scenarios

1. A computer-implemented method comprising:

determining, by a computing system, a symmetric scenario for a scenario;

generating, by the computing system, first training data based on a relabel of a first object in second training data for the symmetric scenario to a first ego vehicle and a relabel of a second ego vehicle in the second training data to a second object;

training, by the computing system, a first machine learning model for the scenario based on first training data generated from the second training data; and

generating, by the computing system, a prediction for the scenario based on the first machine learning model, wherein driving of a vehicle is controlled based on the prediction.

2. The computer-implemented method of claim 1 , wherein training the first machine learning model comprises:

evaluating, by the computing system, predictions generated by the first machine learning model based on labels associated with the second training data.

3. The computer-implemented method of claim 1 , wherein the prediction is provided to a planner to plan a route through an environment.

4. The computer-implemented method of claim 1 , wherein generating the first training data comprises:

relabeling, by the computing system, a first prediction associated with the symmetric scenario to a second prediction associated with the scenario.

5. The computer-implemented method of claim 1 , further comprising:

determining, by the computing system, unlabeled training data for the scenario; and

training, by the computing system, the first machine learning model to determine a transformation that converts instances of the first training data from a first feature space associated with the first training data to a second feature space associated with the unlabeled training data.

6. The computer-implemented method of claim 5 , wherein the transformation is based on minimizing differences between a first feature distribution in the first feature space and a second feature distribution in the second feature space.

7. The computer-implemented method of claim 1 , further comprising:

determining, by the computing system, unlabeled training data for the scenario; and

training, by the computing system, the first machine learning model to minimize a loss function between instances of the first training data and instances of the unlabeled training data.

8. The computer-implemented method of claim 1 , further comprising:

training, by the computing system, a second machine learning model based on labeled training data for the scenario or unlabeled training data for the scenario; and

applying, by the computing system, a portion of the first machine learning model to the second machine learning model.

9. The computer-implemented method of claim 1 , wherein the prediction is generated based on dynamics associated with an obstacle detected in an environment, and wherein the prediction includes a predicted action performed by the obstacle.

10. The computer-implemented method of claim 1 , wherein the scenario is an active merge scenario, and wherein the symmetric scenario is a passive merge scenario.

11. A system comprising:

at least one processor; and

a memory storing instructions that, when executed by the at least one processor, cause the system to perform operations comprising:

determining a symmetric scenario for a scenario;

generating first training data based on a relabel of a first object in second training data for the symmetric scenario to a first ego vehicle and a relabel of a second ego vehicle in the second training data to a second object;

training a first machine learning model for the scenario based on first training data generated from the second training data; and

generating a prediction for the scenario based on the first machine learning model, wherein driving of a vehicle is controlled based on the prediction.

12. The system of claim 11 , wherein training the first machine learning model comprises:

evaluating predictions generated by the first machine learning model based on labels associated with the second training data.

13. The system of claim 11 , wherein the prediction is provided to a planner to plan a route through an environment.

14. The system of claim 12 , wherein generating the first training data comprises:

relabeling a first prediction associated with the symmetric scenario to a second prediction associated with the scenario.

15. The system of claim 11 , the operations further comprising:

determining unlabeled training data for the scenario; and

training the first machine learning model to determine a transformation that converts instances of the first training data from a first feature space associated with the first training data to a second feature space associated with the unlabeled training data.

16. A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor of a computing system, cause the computing system to perform operations comprising:

determining a symmetric scenario for a scenario;

generating first training data based on a relabel of a first object in second training data for the symmetric scenario to a first ego vehicle and a relabel of a second ego vehicle in the second training data to a second object;

training a first machine learning model for the scenario based on first training data generated from the second training data; and

generating a prediction for the scenario based on the first machine learning model, wherein driving of a vehicle is controlled based on the prediction.

17. The non-transitory computer-readable storage medium of claim 16 , wherein training the first machine learning model comprises:

evaluating predictions generated by the first machine learning model based on labels associated with the second training data.

18. The non-transitory computer-readable storage medium of claim 16 , wherein the prediction is provided to a planner to plan a route through an environment.

19. The non-transitory computer-readable storage medium of claim 17 , wherein generating the first training data comprises:

relabeling a first prediction associated with the symmetric scenario to a second prediction associated with the scenario.

20. The non-transitory computer-readable storage medium of claim 16 , the operations further comprising:

determining unlabeled training data for the scenario; and

training the first machine learning model to determine a transformation that converts instances of the first training data from a first feature space associated with the first training data to a second feature space associated with the unlabeled training data.