Codebook embedding generation and processing

1. A method for wireless communication by a wireless node, comprising:

receiving an embedding representing a characterization associated with a beam, wherein the characterization comprises array gain measurements or codebook beam characterization metrics;

providing the embedding to a machine learning (ML) model;

generating one or more communication parameters for communication using the beam via the ML model based on the embedding;

communicating using the one or more communication parameters; and

training an encoder configured to generate embeddings using the characterization associated with the beam.

2. The method of claim 1 , wherein receiving the embedding comprises receiving a look-up table indicating the embedding.

3. The method of claim 1 , wherein:

the ML model comprises a rotation transformer decoder;

generating the one or more communication parameters comprises determining, via the rotation transformer decoder, a rotation of the beam for the communication based on the embedding;

the one or more communication parameters comprises a rotated beam in accordance with the determination; and

the method further comprises providing a rotation instruction to the rotation transformer decoder wherein the rotation of the beam is determined based on the rotation instruction.

4. The method of claim 1 , wherein:

the method further comprises generating another embedding based on a characterization of another beam;

the embedding and the other embedding are provided to a pointwise difference decoder; and

generating the one or more communication parameters comprises predicting a pointwise difference between the beam and the other beam based on the embedding and the other embedding via the pointwise difference decoder.

5. The method of claim 1 , wherein:

the embedding is provided to a reference signal receive power (RSRP) decoder;

generating the one or more communication parameters comprises determining, via the RSRP decoder, an RSRP for each of multiple synchronization signal blocks (SSBs) based on the embedding; and

the method further comprises providing an indication of an orientation of the wireless node, wherein the RSRP is determined based on the orientation.

6. The method of claim 1 , wherein the characterization comprises a spherical array gain.

7. The method of claim 1 , wherein the array gain measurements correspond to array gain measurements per beam on a sphere, and the codebook beam characterization metrics correspond to codebook beam characterization metrics per beam as an auxiliary input for the embedding.

8. A method for wireless communication, comprising:

receiving a characterization associated with a beam, wherein the characterization comprises array gain measurements or codebook beam characterization metrics;

generating an embedding based on the characterization;

providing the embedding to a wireless node; and

training an autoencoder using a federated learning model,

wherein the embedding is generated using an encoder of the autoencoder.

9. The method of claim 8 , wherein providing the embedding comprises providing a look-up table indicating the embedding associated with the beam.

10. The method of claim 8 , wherein:

the characterization comprises a spherical array gain;

the method further comprises converting the spherical array gain to a graph; and

the embedding is generated using a graph convolution network based on the graph.

11. The method of claim 8 , further comprising receiving one or more training codebooks for training of the autoencoder.

12. The method of claim 8 , wherein the array gain measurements correspond to array gain measurements per beam on a sphere, and the codebook beam characterization metrics correspond to codebook beam characterization metrics per beam as an auxiliary input for the embedding, wherein the characterization is received as a noise augmented input, and wherein the embedding is generated using a de-noising auto-encoder based on the noise augmented input.

13. The method of claim 8 , wherein generating the embedding comprises down sampling the characterization using a Fourier transform.

14. The method of claim 8 , wherein the characterization comprises a spherical array gain.

15. An apparatus for wireless communication by a wireless node, comprising:

a memory comprising executable instructions; and

one or more processors configured to execute the executable instructions and cause the apparatus to:

receive an embedding representing a characterization associated with a beam, wherein the characterization comprises array gain measurements or codebook beam characterization metrics;

provide the embedding to a machine learning (ML) model;

generate one or more communication parameters for communication using the beam via the ML model based on the embedding;

communicate using the one or more communication parameters, and

train an encoder configured to generate embeddings using the characterization associated with the beam.

16. The apparatus of claim 15 , wherein, in causing the apparatus to receive the embedding, the one or more processors are configured to cause the apparatus to receive a look-up table indicating the embedding.

17. The apparatus of claim 15 , wherein:

the ML model comprises a rotation transformer decoder;

in causing the apparatus to generate the one or more communication parameters, the one or more processors are configured to cause the apparatus to determine, via the rotation transformer decoder, a rotation of the beam for the communication based on the embedding;

the one or more communication parameters comprises a rotated beam in accordance with the determination; and

the one or more processors are further configured to cause the apparatus to provide a rotation instruction to the rotation transformer decoder, wherein the rotation of the beam is determined based on the rotation instruction.

18. The apparatus of claim 15 , wherein:

the one or more processors are further configured to cause the apparatus to:

generate another embedding based on a characterization of another beam; and

provide the embedding and the other embedding to a pointwise difference decoder; and

in causing the apparatus to generate the one or more communication parameters, the one or more processors are configured to cause the apparatus to predict a pointwise difference between the beam and the other beam based on the embedding and the other embedding via the pointwise difference decoder.

19. The apparatus of claim 15 , wherein:

the one or more processors are further configured to cause the apparatus to provide the embedding to a reference signal receive power (RSRP) decoder;

in causing the apparatus to generate the one or more communication parameters, the one or more processors are configured to cause the apparatus to determine, via the RSRP decoder, an RSRP for each of multiple synchronization signal blocks (SSBs) based on the embedding; and

the one or more processors are further configured to cause the apparatus to provide an indication of an orientation of the wireless node, wherein the RSRP is determined based on the orientation.

20. The apparatus of claim 15 , wherein the characterization comprises a spherical array gain.

21. The apparatus of claim 15 , wherein the array gain measurements correspond to array gain measurements per beam on a sphere, and the codebook beam characterization metrics correspond to codebook beam characterization metrics per beam as an auxiliary input for the embedding.

22. An apparatus for wireless communication, comprising:

a memory comprising executable instructions; and

one or more processors configured to execute the executable instructions and cause the apparatus to:

receive a characterization associated with a beam, wherein the characterization comprises array gain measurements or codebook beam characterization metrics;

generate an embedding based on the characterization;

provide the embedding to a wireless node, and

train an autoencoder using a federated learning model,

wherein to generate the embedding the one or more processors are configured to cause the apparatus to generate the embedding using an encoder of the autoencoder.

23. The apparatus of claim 22 , wherein in causing the apparatus to provide the embedding, the one or more processors are configured to cause the apparatus to provide a look-up table indicating the embedding associated with the beam.

24. The apparatus of claim 22 , wherein:

the characterization comprises a spherical array gain;

the one or more processors are further configured to cause the apparatus to convert the spherical array gain to a graph; and

in causing the apparatus to generate the embedding, the one or more processors are configured to cause the apparatus to generate the embedding using a graph convolution network based on the graph.

25. The apparatus of claim 22 , wherein the one or more processors are further configured to cause the apparatus to receive one or more training codebooks for training of the autoencoder.

26. The apparatus of claim 22 , wherein the array gain measurements correspond to array gain measurements per beam on a sphere, and the codebook beam characterization metrics correspond to codebook beam characterization metrics per beam as an auxiliary input for the embedding,

wherein in causing the apparatus to receive the characterization, the one or more processors are configured to cause the apparatus to receive the characterization as a noise augmented input, and

wherein in causing the apparatus to generate the embedding, the one or more processors are configured to cause the apparatus to generate the embedding using a de-noising auto-encoder based on the noise augmented input.

27. The apparatus of claim 22 , wherein in causing the apparatus to generate the embedding, the one or more processors are configured to cause the apparatus to down sample the characterization using a Fourier transform.

28. The apparatus of claim 22 , wherein the characterization comprises a spherical array gain.