Systems and methods for optimizing coordination and communication resources between vehicles using models
System, methods, and other embodiments described herein relate to coordinating and optimizing prediction by a model using intermediate data and vehicle-to-vehicle (V2V) communications that improves bandwidth utilization. In one embodiment, a method includes processing acquired data, by a subject vehicle using a prediction model, to execute a vehicle task associated with navigating a driving scene. The method also includes extracting processed data according to the acquired data from intermediate activations of layers within the prediction model according to a cooperation model. The method also includes quantifying relevancy of the processed data, for a target vehicle, to select a subset using the cooperation model. The method also includes communicating the subset to a selected vehicle having the prediction model for additional prediction according to the relevancy and available resources to transmit the subset.
1. A coordination system, comprising:
a processor; and
a memory storing instructions that, when executed by the processor, cause the processor to:
process acquired data, by a subject vehicle using a prediction model, to execute a vehicle task associated with navigation of a driving scene;
extract processed data according to the acquired data from intermediate activations of layers within the prediction model according to a cooperation model;
quantify relevancy of the processed data, for a target vehicle, to select a subset associated with the intermediate activations using the cooperation model, and the relevancy adapts with trajectory estimates of the target vehicle received from a selected vehicle; and
communicate the subset to the selected vehicle having the prediction model for an additional prediction according to the relevancy and available resources to transmit the subset.
2. The coordination system of claim 1 , further including instructions to select the processed data according to expected observations by the selected vehicle of the target vehicle using the prediction model, wherein the subject vehicle observes the target vehicle from a view different than the selected vehicle.
3. The coordination system of claim 1 , wherein the instructions to communicate the subset further include instructions to communicate cluster weights from the prediction model and internal representations of the target vehicle using a Gaussian model, wherein the Gaussian model includes mean values computed for the trajectory estimates of the target vehicle.
4. The coordination system of claim 1 , wherein the instructions to quantify the relevancy further include instructions to compute a statistical distribution having a mean and a variance of polynomial trajectories associated with the target vehicle.
5. The coordination system of claim 1 , further including instructions to:
prioritize the processed data from the intermediate activations using a greedy model that recomputes the relevancy after forming a data packet having a portion of the processed data in a frame; and
optimize bandwidth between the subject vehicle and the selected vehicle for the available resources using the frame.
6. The coordination system of claim 1 , further including instructions to:
receive the trajectory estimates of the target vehicle from the selected vehicle by the subject vehicle using vehicle-to-vehicle (V2V) communications.
7. The coordination system of claim 1 , further including instructions to train the prediction model by the selected vehicle using the subset.
8. The coordination system of claim 1 , wherein the selected vehicle processes the subset within the prediction model without training.
9. The coordination system of claim 1 , wherein the intermediate activations represent one of image data, controller area network (CAN) data, past trajectories, velocity, and angular velocity of the subject vehicle partially processed by the prediction model.
10. A non-transitory computer-readable medium comprising:
instructions that when executed by a processor cause the processor to:
process acquired data, by a subject vehicle using a prediction model, to execute a vehicle task associated with navigating a driving scene;
extract processed data according to the acquired data from intermediate activations of layers within the prediction model according to a cooperation model;
quantify relevancy of the processed data, for a target vehicle, to select a subset associated with the intermediate activations using the cooperation model, and the relevancy adapts with trajectory estimates of the target vehicle received from a selected vehicle; and
communicate the subset to the selected vehicle having the prediction model for an additional prediction according to the relevancy and available resources to transmit the subset.
11. The non-transitory computer-readable medium of claim 10 , further including instructions to select the processed data according to expected observations by the selected vehicle of the target vehicle using the prediction model, wherein the subject vehicle observes the target vehicle from a view different than the selected vehicle.
12. A method comprising:
processing acquired data, by a subject vehicle using a prediction model, to execute a vehicle task associated with navigating a driving scene;
extracting processed data according to the acquired data from intermediate activations of layers within the prediction model according to a cooperation model;
quantifying relevancy of the processed data, for a target vehicle, to select a subset associated with the intermediate activations using the cooperation model, and the relevancy adapts with trajectory estimates of the target vehicle received from a selected vehicle; and
communicating the subset to the selected vehicle having the prediction model for an additional prediction according to the relevancy and available resources to transmit the subset.
13. The method of claim 12 , further comprising selecting the processed data according to expected observations by the selected vehicle of the target vehicle using the prediction model, wherein the subject vehicle observes the target vehicle from a view different than the selected vehicle.
14. The method of claim 12 , wherein communicating the subset further includes communicating cluster weights from the prediction model and internal representations of the target vehicle using a Gaussian model, wherein the Gaussian model includes mean values computed for the trajectory estimates of the target vehicle.
15. The method of claim 12 , wherein quantifying the relevancy further includes computing a statistical distribution having a mean and a variance of polynomial trajectories associated with the target vehicle.
16. The method of claim 12 , further comprising:
prioritizing the processed data from the intermediate activations using a greedy model that recomputes the relevancy after forming a data packet having a portion of the processed data in a frame; and
optimizing bandwidth between the subject vehicle and the selected vehicle for the available resources using the frame.
17. The method of claim 12 , further comprising:
receiving the trajectory estimates of the target vehicle from the selected vehicle by the subject vehicle using vehicle-to-vehicle (V2V) communications.
18. The method of claim 12 , further comprising training the prediction model by the selected vehicle using the subset.
19. The method of claim 12 , wherein the selected vehicle processes the subset within the prediction model without training.
20. The method of claim 12 , wherein the intermediate activations represent one of image data, controller area network (CAN) data, past trajectories, velocity, and angular velocity of the subject vehicle partially processed by the prediction model.