Method and system for forecasting reactions of other road users in autonomous driving
System and methods for causing movement of an autonomous vehicle. The methods comprise: determining candidate trajectories of the autonomous vehicle using candidate actions forecasted for a virtual doppelganger and reactive actions predicted for a non-Ego actor that is located in proximity to the autonomous vehicle; selecting first candidate trajectories of the candidate trajectories in which the autonomous vehicle has an influence on a response of the non-Ego actor; using the first candidate trajectories for the autonomous vehicle to refine the reactive actions that were predicted for the non-Ego actor; using the reactive actions that were refined to select one of the first candidate trajectories as a selected trajectory for the autonomous vehicle to follow; and causing an automation subsystem of the autonomous vehicle to move the autonomous vehicle along the selected trajectory.
1. A method for causing movement of an autonomous vehicle, comprising:
determining candidate trajectories of the autonomous vehicle using candidate actions forecasted for a virtual doppelganger that is co-located with the autonomous vehicle and reactive actions predicted for a non-Ego actor that is located in sensor range to the autonomous vehicle;
selecting first candidate trajectories of the candidate trajectories in which the autonomous vehicle has an influence on a response of the non-Ego actor;
using the first candidate trajectories for the autonomous vehicle to refine the reactive actions that were predicted for the non-Ego actor;
using the reactive actions that were refined to select one of the first candidate trajectories as a selected trajectory for the autonomous vehicle to follow; and
causing an automation subsystem of the autonomous vehicle to move the autonomous vehicle along the selected trajectory.
2. The method according to claim 1 , further comprising scoring the candidate trajectories of the autonomous vehicle based on a level of probability that the non-Ego actor will execute one or more of the reactive actions that were refined.
3. The method according to claim 2 , wherein the selected trajectory for the autonomous vehicle comprises a candidate trajectory having a highest score.
4. The method according to claim 2 , wherein said scoring is further based on an estimated value of the candidate trajectory should it be executed, given that the non-Ego actor executes a particular trajectory.
5. The method according to claim 1 , wherein each of the candidate actions forecasted for the virtual doppelganger corresponds to a possible modality of a plurality of possible modalities predicted for the non-Ego actor and is based on at least one possible trajectory for the non-Ego actor.
6. The method according claim 5 , wherein the plurality of possible modalities comprise high-level categories of actions, at least some of which are associated with a plurality of possible trajectories.
7. The method according to claim 1 , wherein each of the reactive actions predicted for a non-Ego actor is reactive to one of the candidate actions forecasted for the virtual doppelganger.
8. A system, comprising:
a processor;
a non-transitory computer-readable storage medium comprising programming instructions that are configured to cause the processor to implement a method for managing data, wherein the programming instructions comprise instructions to:
determine candidate trajectories of the autonomous vehicle using candidate actions forecasted for a virtual doppelganger that is co-located with the autonomous vehicle and reactive actions predicted for a non-Ego actor that is located in sensor range to the autonomous vehicle;
select first candidate trajectories of the candidate trajectories in which the autonomous vehicle has an influence on a response of the non-Ego actor;
use the first candidate trajectories for the autonomous vehicle to refine the reactive actions that were predicted for the non-Ego actor;
use the reactive actions that were refined to select one of the first candidate trajectories as a selected trajectory for the autonomous vehicle to follow; and
cause an automation subsystem of the autonomous vehicle to move the autonomous vehicle along the selected trajectory.
9. The system according to claim 8 , wherein the programming instructions further comprise instructions to score the candidate trajectories of the autonomous vehicle based on a level of probability that the non-Ego actor will execute one or more of the reactive actions that were refined.
10. The system according to claim 9 , wherein the selected trajectory for the autonomous vehicle comprises a candidate trajectory having a highest score.
11. The system according to claim 9 , wherein said scoring is further based on an estimated value of the candidate trajectory should it be executed, given that the non-Ego actor executes a particular trajectory.
12. The system according to claim 8 , wherein each of the candidate actions forecasted for the virtual doppelganger corresponds to a possible modality of a plurality of possible modalities predicted for the non-Ego actor and is based on at least one possible trajectory for the non-Ego actor.
13. The system according claim 12 , wherein the plurality of possible modalities comprise high-level categories of actions, at least some of which are associated with a plurality of possible trajectories.
14. The system according to claim 8 , wherein each of the reactive actions predicted for a non-Ego actor is reactive to one of the candidate actions forecasted for the virtual doppelganger.
15. A non-transitory computer-readable medium that stores instructions that, when executed by at least one computing device, will cause the at least one computing device to perform operations comprising:
determining candidate trajectories of the autonomous vehicle using candidate actions forecasted for a virtual doppelganger that is co-located with the autonomous vehicle and reactive actions predicted for a non-Ego actor that is located in sensor range to the autonomous vehicle;
selecting first candidate trajectories of the candidate trajectories in which the autonomous vehicle has an influence on a response of the non-Ego actor;
using the first candidate trajectories for the autonomous vehicle to refine the reactive actions that were predicted for the non-Ego actor;
using the reactive actions that were refined to select one of the first candidate trajectories as a selected trajectory for the autonomous vehicle to follow; and
causing an automation subsystem of the autonomous vehicle to move the autonomous vehicle along the selected trajectory.
16. The non-transitory computer-readable medium according to claim 15 , wherein the at least one computing device is further caused to score the candidate trajectories of the autonomous vehicle based on a level of probability that the non-Ego actor will execute one or more of the reactive actions that were refined.
17. The non-transitory computer-readable medium according to claim 16 , wherein the selected trajectory for the autonomous vehicle comprises a candidate trajectory having a highest score.
18. The non-transitory computer-readable medium according to claim 16 , wherein said candidate trajectories are scored further based on an estimated value of the candidate trajectory should it be executed, given that the non-Ego actor executes a particular trajectory.
19. The non-transitory computer-readable medium according to claim 15 , wherein each of the candidate actions forecasted for the virtual doppelganger corresponds to a possible modality of a plurality of possible modalities predicted for the non-Ego actor and is based on at least one possible trajectory for the non-Ego actor.
20. The non-transitory computer-readable medium according claim 19 , wherein the plurality of possible modalities comprise high-level categories of actions, at least some of which are associated with a plurality of possible trajectories.