Rank indicator (RI) and channel quality indicator (CQI) estimation using a multi-layer perceptron (MLP)
View Patent ↗An apparatus and method are provided for using an MLP algorithm to map channel features to an RI and/or a CQI for CSI feedback. The method includes estimating a channel in the communication network for a signal; extracting at least one channel feature related to the estimated channel; determining RI and CQI pairs; inputting, to an MLP network, the extracted at least one channel feature and the RI and CQI pairs; receiving, for each of the RI and CQI pairs, an output of the MLP network, wherein the outputs of the MLP network indicate throughput or spectral efficiency for the electronic device; and selecting an RI and CQI pair of the RI and CQI pairs based on the received outputs.
1. A method performed by an electronic device in a communication network, the method comprising:
estimating a channel in the communication network for a signal;
extracting at least one channel feature related to the estimated channel;
determining rank indicator (RT) and channel quality indicator (CQI) combinations;
inputting, to a multi-layer perceptron (MLP network, the extracted at least one channel feature and the RI and CQI combinations;
receiving, for each of the RI and CQI combinations, an output of the MLP network, wherein the outputs of the MIT network indicate throughput or spectral efficiency for the electronic device; and
selecting an RI and CQI combination of the RI and CQI combinations based on the received outputs.
2. The method of claim 1 , wherein selecting the RI and CQI combination of the RI and CQI combinations based on the received outputs comprises selecting an RI and CQI combination of the RI and CQI combinations corresponding to a best output of the received outputs.
3. The method of claim 1 , further comprising receiving an indication of resources based on the selected RI and CQI combination.
4. The method of claim 3 , further comprising transmitting the signal using the indicated resources.
5. The method of claim 3 , further comprising processing the signal using the indicated resources.
6. The method of claim 1 , further comprising:
calculating weights for the estimated channel; and
applying the weights to the received outputs, prior to selecting the RI and CQI combination of the RI and CQI combinations corresponding to the best output of the received outputs.
7. The method of claim 6 , wherein the weights include a function of an estimated probability of an acknowledgement (ACK).
8. The method of claim 6 , wherein the RI and CQI combination of the RI and CQI combinations are selected using:
(
RI
*
(
s
)
,
CQ
I
*
(
s
)
)
=
arg
max
(
RI
,
CQI
)
Q
*
(
s
,
RI
,
CQI
)
*
W
(
RI
,
CQI
)
∀
s
∈
𝕊
,
where Q*(s,RI,CQI) is an estimated Q-function from the MLP network when a state s is observed and RI,CQI is used and W(RI,CQI) is a function of RI and CQI to weight a resulting Q-function and modify the MLP network decision if a predetermined condition is met.
9. The method of claim 8 , further comprising controlling a block error rate (BLER) by adjusting W(RI,CQI).
10. The method of claim 1 , wherein the MLP network is modeled offline using at least one of an unconstrained case for maximizing throughput or a constrained case in consideration of an additional constraint on performance.
11. The method of claim 10 , wherein the additional constraint on performance includes a block error rate (BLER) constraint.
12. The method of claim 1 , wherein the MLP network is trained offline using at least one of an unconstrained case for maximizing throughput or unconstrained case in consideration of an additional constraint on performance.
13. The method of claim 1 , wherein the at least one channel feature related to the estimated channel includes at least one of an average signal-to-noise ratio (SNR), an estimated delay spread, and an estimated Doppler frequency.
14. The method of claim 1 , wherein the outputs of the MLP network are based on at least one of a number of ports for the signal, a number of receive antennas, a channel state information (CSI) reporting period, and a feedback delay.
15. The method of claim 1 , wherein the RI and CQI combinations are included in a subset of RI and CQI combinations selected from all possible RI and CQI combinations.
16. An electronic device for use in a communication network, the electronic device comprising:
a transceiver;
a multi-layer perceptron (MLP) network; and
a processor configured to:
estimate a channel in the communication network for a signal,
extract at least one channel feature related to the estimated channel,
determine rank indicator (RI) and channel quality indicator (CQI) combinations,
input, to the MLP network, the extracted at least one channel feature and the RI and CQI combinations,
receive, for each of the RI and CQI combinations, an output of the MLP network, wherein the outputs of the MU′ network indicate throughput or spectral efficiency for the electronic device, and
select an RI and CQI combination of the RI and CQI combinations based on the received outputs.
17. The electronic device of claim 16 , wherein processor is further configured to selecting an RI and CQI combination of the RI and CQI combinations corresponding to a best output of the received outputs.
18. The electronic device of claim 16 , wherein the processor is further configured to transmit, via the transceiver, the signal using resources that are indicated to the electronic device based on the selected RI and CQI combination.
19. The electronic device of claim 16 , wherein the processor is further configured to process the signal using resources that are indicated to the electronic device based on the selected RI and CQI combination.
20. The electronic device of claim 16 , wherein the processor is further configured to:
calculate weights for the estimated channel, and
apply the weights to the received outputs, prior to selecting the RI and CQI combination of the RI and CQI combinations corresponding to the best output of the received outputs.