Processing environmental data for vehicles

1. A computer-implemented method for processing environmental data for at least one other vehicle relative to a target vehicle, the method comprising:

processing, by a processing device using a self-attention mechanism, first data associated with the at least one other vehicle to generate a first output associated with the at least one other vehicle;

processing, by the processing device using stacked residual blocks, second data associated with the target vehicle to generate a second output associated with the target vehicle;

concatenating, by the processing device, the first output and the second output to generate a concatenated vector;

making, by the processing device, a driving decision for the target vehicle based at least in part on the concatenated vector; and

controlling the target vehicle based at least in part on the driving decision,

wherein the first data comprises a first subset of the first data associated with a first other vehicle and a second subset of the first data associated with a second other vehicle,

wherein the first subset of the first data is expressed as a first one-dimensional (1D) features vector, and wherein the second subset of the first data is expressed as a second 1D features vector,

wherein processing the first data comprises combining the first 1D features vector and the second 1D features vector into a two-dimensional (2D) matrix, and

wherein the 2D matrix comprises velocity data associated with the at least one other vehicle and position data associated with the at least one other vehicle.

2. The computer-implemented method of claim 1 , wherein the first data associated with the at least one other vehicle comprises at least one of an estimated kinematic state of the at least one other vehicle, a lane assignment, and a vehicle intent.

3. The computer-implemented method of claim 1 , wherein the second data associated with the target vehicle comprises kinematic state information and lane information.

4. The computer-implemented method of claim 3 , wherein the kinematic state information is represented in a coordinate frame of a road upon which the target vehicle is located.

5. The computer-implemented method of claim 3 , wherein the lane information is expressed by a vector of binary values, where a zero value represents no lane present and a one value represents a lane present.

6. The computer-implemented method of claim 1 , wherein each of the stacked residual blocks comprises a first convolutional unit, a rectified linear unit, and a second convolutional unit.

7. The computer-implemented method of claim 1 , wherein making the driving decision comprises determining a driving policy and a value associated with a current state of an agent in an environment.

8. The computer-implemented method of claim 7 , wherein controlling the target vehicle based at least in part on the decision comprises controlling the target vehicle based on the driving policy and a current state of an agent in an environment.

9. A system comprising:

a memory comprising computer readable instructions; and

a processing device for executing the computer readable instructions, the computer readable instructions controlling the processing device to perform operations for processing environmental data for at least one other vehicle relative to a target vehicle, the operations comprising:

processing, using a self-attention mechanism, first data associated with the at least one other vehicle to generate a first output associated with the at least one other vehicle;

processing, using stacked residual blocks, second data associated with the target vehicle to generate a second output associated with the target vehicle;

concatenating the first output and the second output to generate a concatenated vector;

making a driving decision for the target vehicle based at least in part on the concatenated vector; and

controlling the target vehicle based at least in part on the driving decision,

wherein the first data comprises a first subset of the first data associated with a first other vehicle and a second subset of the first data associated with a second other vehicle,

wherein the first subset of the first data is expressed as a first one-dimensional (1D) features vector, and wherein the second subset of the first data is expressed as a second 1D features vector,

wherein processing the first data comprises combining the first 1D features vector and the second 1D features vector into a two-dimensional (2D) matrix, and

wherein the 2D matrix comprises velocity data associated with the at least one other vehicle and position data associated with the at least one other vehicle.

10. The system of claim 9 , wherein the first data associated with the at least one other vehicle comprises at least one of an estimated kinematic state of the at least one other vehicle, a lane assignment, and a vehicle intent.

11. The system of claim 9 , wherein the second data associated with the target vehicle comprises kinematic state information and lane information.

12. A computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform operations for processing environmental data for at least one other vehicle relative to a target vehicle, the operations comprising:

processing, using a self-attention mechanism, first data associated with the at least one other vehicle to generate a first output associated with the at least one other vehicle;

processing, using stacked residual blocks, second data associated with the target vehicle to generate a second output associated with the target vehicle;

concatenating the first output and the second output to generate a concatenated vector;

making a driving decision for the target vehicle based at least in part on the concatenated vector; and

controlling the target vehicle based at least in part on the driving decision,

wherein the first data comprises a first subset of the first data associated with a first other vehicle and a second subset of the first data associated with a second other vehicle,

wherein the first subset of the first data is expressed as a first one-dimensional (1D) features vector, and wherein the second subset of the first data is expressed as a second 1D features vector,

wherein processing the first data comprises combining the first 1D features vector and the second 1D features vector into a two-dimensional (2D) matrix, and

wherein the 2D matrix comprises velocity data associated with the at least one other vehicle and position data associated with the at least one other vehicle.

13. The computer-implemented method of claim 1 , wherein first data associated with the at least one other vehicle is received from the at least one other vehicle.

14. The computer-implemented method of claim 1 , wherein the first data is received from the at least one other vehicle using vehicle-to-vehicle communication.

15. The computer-implemented method of claim 1 , wherein processing the second data comprises performing a first convolutional operation, performing an activation operation to transform results of the first convolutional operation, and performing a second convolutional operation on results of the activation operation.

16. The computer-implemented method of claim 1 , wherein the driving decision made using a driving policy and a state value.

17. The system of claim 9 , wherein first data associated with the at least one other vehicle is received from the at least one other vehicle.

18. The system of claim 9 , wherein the first data is received from the at least one other vehicle using vehicle-to-vehicle communication.

19. The system of claim 9 , wherein processing the second data comprises performing a first convolutional operation, performing an activation operation to transform results of the first convolutional operation, and performing a second convolutional operation on results of the activation operation.

20. The system of claim 9 , wherein the driving decision made using a driving policy and a state value.