Temporal domain overlapped prediction with UE model monitoring enhancement
This document discloses a method and an apparatus to perform using a predicted sequence of beam measurement outputs for at least one observation window associated with at least one prediction window providing reference signal received power measurements to predict a beam prediction output; identifying that a prediction window of the at least one prediction window is overlapped with at least a portion of an observation window of the at least one observation window over a time period; and comparing the beam prediction output during the overlapped portion with measurements of the observation window during the overlapped portion to perform long term real time monitoring to determine at least one prediction output for use in a next time instant, wherein the determining is based on a predicted reference signal received power of the at least one prediction output being above a threshold.
1 . An apparatus, comprising:
at least one processor; and
at least one non-transitory memory storing instructions, that when executed by the at least one processor, cause the apparatus at least to:
use a sequence of beam measurements including reference signal received power measurements for at least one observation window to predict a beam prediction output for at least one prediction window using an artificial intelligence/machine learning (AI/ML) model,
wherein the beam prediction output includes beam identifications or predicted reference signal received power of a subset of beams associated with the beam prediction output;
identify that a prediction window of the at least one prediction window is overlapped with at least a portion of an observation window of the at least one observation window over a time period;
compare the beam prediction output during the overlapped portion with measurements of the observation window during the overlapped portion to perform long term real time monitoring to determine at least one of a prediction accuracy of the AI/ML model or a predicted event,
wherein the predicted event is based on a predicted reference signal received power being above a threshold.
2 . The apparatus of claim 1 , wherein the at least one observation window comprises at least one sliding observation window of identification reference signal received power, and wherein the at least one prediction window comprises at least one sliding prediction window.
3 . The apparatus of claim 2 , wherein the identifying and comparing is using network configuration information from a communication network comprising a channel state information reference signal associated with the overlapped portion of the observation window.
4 . The apparatus of claim 3 , wherein the network configuration information comprises an indication of a size of the at least one sliding prediction window, and provides a configuration for the apparatus to predict one of downlink reception beam pair or downlink transmission beam pair as a new serving beam.
5 . The apparatus of claim 3 , wherein the network configuration information provides a configuration for the apparatus to predict one of downlink transmission beam identification or subset of downlink transmission beam identifications or one of downlink transmission beam identification as new serving beam.
6 . The apparatus of claim 1 , wherein the prediction window is overlapped with at least one next observation window over the time period, and
wherein the comparing is comprising comparing the beam prediction output of more than one observation window with the prediction window during the overlapped portion to perform the long term real time monitoring.
7 . The apparatus of claim 1 , wherein the at least one non-transitory memory storing instructions that when executed by the at least one processor, cause the apparatus at least to:
for the beam prediction output, obtain at least one downlink transmission beam identification reference signal received power or at least one downlink transmission and reception beam pair identification reference signal received power provided by machine learning.
8 . The apparatus of claim 7 , wherein the machine learning is configured to determine whether a predicted new serving downlink transmission beam or downlink transmission and reception beam pair will be configured for at least one of a physical uplink shared channel, physical uplink control channel, physical downlink shared channel, or a physical downlink control channel.
9 . The apparatus of claim 4 , wherein the one of a downlink reception beam pair or a downlink transmission beam pair is from at least one of a downlink reception beam pair observation window or downlink transmission beam pair observation window.
10 . The apparatus of claim 1 , wherein the reference signal received power measurements are provided between time point t-N and time point t, wherein the identifying that the prediction window is overlapped is performed after time point t, and wherein predicting the beam prediction output is performed after time point t.
11 . The apparatus of claim 1 , wherein the sequence of beam measurements are associated with sliding windows.
12 . The apparatus of claim 1 , wherein the at least one non-transitory memory storing instructions that when executed by the at least one processor cause the apparatus at least to:
receive from a communication network information comprising an indication comprising a channel state information reference signal according to overlapped portions of at least one sliding prediction window of time or a concurrent at least one observation window of time.
13 . The apparatus of claim 12 , wherein the at least one non-transitory memory storing instructions that when executed by the at least one processor cause the apparatus at least to:
perform training and run an inference to obtain a reference signal received power of a set of reference signal received power measurement beam predictions.
14 . The apparatus of claim 1 , wherein the beam prediction output is based on measuring layer 1 reference signal received power measurements.
15 . The apparatus of claim 14 , wherein the layer 1 reference signal and assistant information comprising beam identifications, user equipment positions, and a line of sight or non-line of sight indication are used as input to the AI/ML model.
16 . The apparatus of claim 1 , wherein the sequence of beam measurements is of reference signal received power over time.
17 . The apparatus of claim 10 , wherein the comparing the beam prediction output with measurements of the observation window during the overlapped portion is performed after time point t using the sequence of beam measurements from the at least one observation window.
18 . A method, comprising:
using a sequence of beam measurements including reference signal received power measurements for at least one observation window to predict a beam prediction output for at least one prediction window using an artificial intelligence/machine learning (AI/ML) model,
wherein the beam prediction output includes beam identifications or predicted reference signal received power of a subset of beams associated with the beam prediction output;
identifying that a prediction window of the at least one prediction window is overlapped with at least a portion of an observation window of the at least one observation window over a time period; and
comparing the beam prediction output during the overlapped portion with measurements of the observation window during the overlapped portion to perform long term real time monitoring to determine at least one of a prediction accuracy of the AI/ML model or a predicted event,
wherein the predicted event is based on a predicted reference signal received power of being above a threshold.
19 . The method of claim 18 , wherein the at least one observation window comprises at least one sliding observation window of identification reference signal received power, and wherein the at least one prediction window comprises at least one sliding prediction window.
20 . The method of claim 19 , wherein the identifying and comparing is using network configuration information from a communication network comprising a channel state information reference signal associated with the overlapped portion of the observation window.