Signal synchronization for over-the-air aggregation in a federated learning framework

1. An apparatus for wireless communications, comprising:

at least one memory comprising instructions; and

at least one processor configured to execute the instructions and cause the apparatus to:

perform federated learning to generate a first set of updated model parameters corresponding to a machine learning model;

receive, from a network entity, a request for the first set of updated model parameters, wherein the request includes a resource allocation associated with an uplink channel;

determine, in response to signaling received from the network entity, a signal phase corresponding to the uplink channel;

transmit, to the network entity and in accordance with the signal phase, the first set of updated model parameters using the resource allocation on the uplink channel;

transmit a first sidelink reference signal to at least one user equipment (UE), wherein the first sidelink reference signal provides a phase synchronization among the apparatus and the at least one UE for the uplink channel;

receive, from the at least one UE, a second sidelink reference signal in response to the first sidelink reference signal;

determine a phase drift parameter based on the second sidelink reference signal; and

adjust the signal phase based on the phase drift parameter.

2. The apparatus of claim 1 , wherein the signaling includes channel state information (CSI) received from the network entity.

3. The apparatus of claim 1 , wherein the signaling includes at least one reference signal received from the network entity on a downlink channel, and wherein the signal phase is based on phase data associated with the downlink channel.

4. The apparatus of claim 1 , wherein the apparatus is configured to preserve the signal phase during a signal transmission following a signal reception.

5. The apparatus of claim 1 , wherein the at least one processor is further configured to cause the apparatus to:

receive a second set of updated model parameters from a user equipment (UE); and

combine the second set of updated model parameters with the first set of updated model parameters to yield a combined set of updated model parameters, wherein the transmitting the first set of updated model parameters includes transmitting the combined set of updated model parameters.

6. The apparatus of claim 5 , wherein the apparatus is a cluster head of a UE cluster that includes the apparatus and the UE.

7. The apparatus of claim 1 , wherein the at least one processor is further configured to cause the apparatus to:

transmit an uplink phase synchronization signal to the network entity;

receive a downlink phase synchronization signal in response to the uplink phase synchronization signal;

determine a phase drift parameter based on the downlink phase synchronization signal; and

adjust the signal phase based on the phase drift parameter.

8. The apparatus of claim 1 , wherein the at least one processor is further configured to cause the apparatus to:

transmit an uplink phase synchronization signal to the network entity; and

receive a phase drift parameter from the network entity, wherein the signal phase is based on the phase drift parameter.

9. The apparatus of claim 8 , wherein the at least one processor is further configured to cause the apparatus to:

transmit channel state feedback (CSF) in response to a channel state information reference signal (CSI-RS) received from the network entity, wherein the phase drift parameter is based on the CSF.

10. A method for wireless communications performed at a first user equipment (UE), comprising:

performing federated learning to generate a first set of updated model parameters corresponding to a machine learning model;

receiving, from a network entity, a request for the first set of updated model parameters, wherein the request includes a resource allocation associated with an uplink channel;

determining, in response to signaling received from the network entity, a signal phase corresponding to the uplink channel;

transmitting, to the network entity and in accordance with the signal phase, the first set of updated model parameters using the resource allocation on the uplink channel;

transmitting a first sidelink reference signal to at least one other user equipment (UE), wherein the first sidelink reference signal provides a phase synchronization among the first UE and the at least one other UE for the uplink channel;

receiving, from the at least one other UE, a second sidelink reference signal in response to the first sidelink reference signal;

determining a phase drift parameter based on the second sidelink reference signal; and

adjusting the signal phase based on the phase drift parameter.

11. The method of claim 10 , wherein the signaling includes channel state information (CSI) received from the network entity.

12. The method of claim 10 , wherein the signaling includes a reference signal received from the network entity on a downlink channel, and wherein the signal phase is based on phase data associated with the downlink channel.

13. The method of claim 10 , wherein the first UE is configured to preserve the signal phase during a signal transmission following a signal reception.

14. The method of claim 10 , further comprising:

receiving a second set of updated model parameters from a second UE; and

combining the second set of updated model parameters with the first set of updated model parameters to yield a combined set of updated model parameters, wherein the transmitting the first set of updated model parameters includes transmitting the combined set of updated model parameters.

15. The method of claim 14 , wherein the first UE is a cluster head of a UE cluster that includes the first UE and the second UE.

16. The method of claim 10 , further comprising:

transmitting an uplink phase synchronization signal to the network entity;

receiving a downlink phase synchronization signal in response to the uplink phase synchronization signal;

determining a phase drift parameter based on the downlink phase synchronization signal; and

adjusting the signal phase based on the phase drift parameter.

17. The method of claim 10 , further comprising:

transmitting an uplink phase synchronization signal to the network entity; and

receiving a phase drift parameter from the network entity, wherein the signal phase is based on the phase drift parameter.

18. The method of claim 17 , further comprising:

transmitting channel state feedback (CSF) in response to a channel state information reference signal (CSI-RS) received from the network entity, wherein the phase drift parameter is based on the CSF.