Qualifying machine learning-based CSI prediction
Certain aspects of the present disclosure provide techniques for qualifying machine learning model-based channel state information (CSI) predictions. An example method generally includes receiving, from a network entity, a channel state information (CSI) prediction model for quantized CSI, calculating CSI based on downlink reference signal measurements, generating a quantized CSI difference value based a quantization of a difference between the calculated CSI and CSI predicted based on a CSI prediction model, and reporting, to the network entity, the calculated CSI and the quantized CSI difference value.
1. A method for wireless communications by a user equipment (UE), comprising:
receiving, from a network entity, a channel state information (CSI) prediction model for quantized CSI and a qualifying scheme for classifying a predicted CSI;
calculating CSI based on downlink reference signal measurements;
generating a quantized CSI difference value based on a quantization of a difference between the calculated CSI and the predicted CSI, the predicted CSI based on the CSI prediction model; and
reporting, to the network entity, the calculated CSI and the quantized CSI difference value and an indication of whether the predicted CSI is a qualified or nonqualified prediction based on the qualifying scheme and the difference between the calculated CSI and the predicted CSI.
2. The method of claim 1 , wherein a qualified prediction comprises a prediction for which the difference between is less than a threshold amount.
3. The method of claim 1 , further comprising: reporting the predicted CSI based on a determination that the predicted CSI is a qualified prediction.
4. The method of claim 1 , wherein the predicted CSI is used for early hybrid automatic repeat request (HARQ) feedback.
5. The method of claim 4 , wherein the qualifying scheme comprises a determination of a misdetection rate or a false alarm rate.
6. The method of claim 5 , wherein a qualified prediction comprises a CSI prediction made when the misdetection rate or the false alarm rate is less than a threshold amount.
7. The method of claim 5 , wherein a nonqualified prediction comprises a CSI prediction made when the misdetection rate or the false alarm rate exceeds a threshold amount.
8. The method of claim 1 , further comprising:
reporting, to the network entity, the calculated CSI and the predicted CSI; and
receiving, from the network entity, configuration information instructing the UE to report qualified CSI predictions based on the qualifying scheme.
9. The method of claim 8 , wherein the calculated CSI and the predicted CSI comprise one or more of early hybrid automatic repeat request (HARQ) acknowledgment (ACK)/negative acknowledgment (NACK) (HARQ-ACK/NACK), precoding matrix indicator (PMI), channel quality indicator (CQI), rank indicator (RI), or signal-to-noise ratio (SNR).
10. The method of claim 1 , wherein the quantized CSI difference value comprises one of a plurality of values indicating a size of the difference between the calculated CSI and the predicted CSI.
11. The method of claim 1 , wherein the difference between the calculated CSI and the predicted CSI comprises an infinite impulse response (IIR) filtered difference between the calculated CSI and the predicted CSI.
12. A method for wireless communications by a user equipment (UE), comprising:
receiving, from a network entity, a channel state information (CSI) prediction model for quantized CSI;
calculating CSI based on downlink reference signal measurements;
generating a quantized CSI difference value based on a quantization of a difference between the calculated CSI and a predicted CSI, the predicted CSI based on the CSI prediction model, wherein the quantized CSI difference value is based on an average of the difference between the calculated CSI and the predicted CSI over a time window; and
reporting, to the network entity, the calculated CSI and the quantized CSI difference value.
13. A method for wireless communications by a network entity, comprising:
transmitting, to a user equipment (UE), a channel state information (CSI) prediction model for predicting CSI, one or more reference signals for the UE to use in calculating CSI and generating one or more CSI difference values based on a calculated CSI and a predicted CSI, and a qualifying scheme for classifying the predicted CSI based on the CSI prediction model, the predicted CSI based on the CSI prediction model;
receiving, from the UE, a calculated CSI, a quantized CSI difference value, and an indication of whether the predicted CSI is a qualified or nonqualified prediction based on the qualifying scheme and a difference between the calculated CSI and the predicted CSI; and
communicating with the UE based on the calculated CSI and the quantized CSI difference value.
14. The method of claim 13 , wherein a qualified prediction comprises a prediction for which the difference is less than a threshold amount.
15. The method of claim 13 , further comprising:
receiving the predicted CSI; and
determining that the predicted CSI is a qualified prediction based on receiving the predicted CSI.
16. The method of claim 13 , wherein the predicted CSI is used for early hybrid automatic repeat request (HARQ) feedback.
17. The method of claim 16 , wherein the qualifying scheme comprises a scheme that classifies predicted CSI based on a misdetection rate or a false alarm rate.
18. The method of claim 17 , further comprising: determining that the predicted CSI comprises a CSI prediction made when the misdetection rate or the false alarm rate is less than a threshold amount based on receiving the predicted CSI.
19. The method of claim 13 , further comprising:
receiving, from the UE, the calculated CSI and the predicted CSI; and
transmitting, to the UE, configuration information instructing the UE to report qualified CSI predictions based on the qualifying scheme.
20. The method of claim 19 , wherein the calculated CSI and the predicted CSI comprise one or more of early hybrid automatic repeat request (HARQ) acknowledgment (ACK)/negative acknowledgment (NACK) (HARQ-ACK/NACK), precoding matrix indicator (PMI), channel quality indicator (CQI), rank indicator (RI), or signal-to-noise ratio (SNR).
21. The method of claim 13 , wherein the quantized CSI difference value comprises one of a plurality of values indicating a size of the difference between the calculated CSI and the predicted CSI.
22. The method of claim 13 , wherein the difference between the calculated CSI and the predicted CSI comprises an infinite impulse response (IIR) filtered difference between the calculated CSI and the predicted CSI.
23. A method for wireless communications by a network entity, comprising:
transmitting, to a user equipment (UE), a channel state information (CSI) prediction model for predicting CSI and one or more reference signals for the UE to use in calculating CSI and generating one or more CSI difference values based on a calculated CSI and a predicted CSI, the predicted CSI based on the CSI prediction model;
receiving, from the UE, a calculated CSI and a quantized CSI difference value, wherein the quantized CSI difference value comprises an average of the difference between the calculated CSI and the predicted CSI over a time window; and
communicating with the UE based on the calculated CSI and the quantized CSI difference value.
24. An apparatus for wireless communications by a user equipment (UE), comprising:
a processor configured to:
receive, from a network entity, a channel state information (CSI) prediction model for quantized CSI and a qualifying scheme for classifying a predicted CSI,
calculate CSI based on downlink reference signal measurements,
generate a quantized CSI difference value based on a quantization of a difference between the calculated CSI and the predicted CSI, the predicted CSI based on the CSI prediction model, and
report, to the network entity, the calculated CSI and the quantized CSI difference value and an indication of whether the predicted CSI is a qualified or nonqualified prediction based on the qualifying scheme and the difference between the calculated CSI and the predicted CSI; and
a memory.
25. An apparatus for wireless communications by a network entity, comprising:
a processor configured to:
transmit, to a user equipment (UE), a channel state information (CSI) prediction model for predicting CSI, one or more reference signals for the UE to use in calculating CSI and generating one or more CSI difference values based on a calculated CSI and a predicted CSI, and a qualifying scheme for classifying the predicted CSI based on the CSI prediction model, the predicted CSI based on the CSI prediction model,
receive, from the UE, a calculated CSI a quantized CSI difference value, and an indication of whether the predicted CSI is a qualified or nonqualified prediction based on the qualifying scheme and a difference between the calculated CSI and the predicted CSI, and
communicate with the UE based on the calculated CSI and the quantized CSI difference value; and
a memory.
26. An apparatus for wireless communications by a user equipment, comprising:
a processor configured to:
receive, from a network entity, a channel state information (CSI) prediction model for quantized CSI,
calculate CSI based on downlink reference signal measurements,
generating a quantized CSI difference value based on a quantization of a difference between the calculated CSI and a predicted CSI, the predicted CSI based on the CSI prediction model, wherein the quantized CSI difference value is based on an average of the difference between the calculated CSI and the predicted CSI over a time window, and
reporting, to the network entity, the calculated CSI and the quantized CSI difference value; and
a memory.
27. An apparatus for wireless communications by a network entity, comprising:
a processor configured to:
transmit, to a user equipment (UE), a channel state information (CSI) prediction model for predicting CSI and one or more reference signals for the UE to use in calculating CSI and generating one or more CSI difference values based on a calculated CSI and a predicted CSI, the predicted CSI based on the CSI prediction model,
receive, from the UE, a calculated CSI and a quantized CSI difference value, wherein the quantized CSI difference value comprises an average of the difference between the calculated CSI and the predicted CSI over a time window, and
communicate with the UE based on the calculated CSI and the quantized CSI difference value; and
a memory.