Method and apparatus for machine learning based wide beam optimization in cellular network

1. A user equipment (UE) in a wireless communication system, the UE comprising:

a transceiver configured to:

receive, from a central controller (CC) via a base station (BS), information of a first beam that is used in a measurement report; and

transmit, to the CC via the BS, the measurement report including a measurement result of the first beam,

wherein:

the first beam for the BS is selected, by the CC, from a set of beams in a candidate beam pool including predetermined candidate beams, the set of beams being allocated to the BS;

consecutive measurement results corresponding to the set of beams in the candidate beam pool along with the measurement result of the first beam are preprocessed by the CC;

beam scores for the first beam based on the measurement result of the first beam are calculated by the CC; and

a second beam based on the beam scores is selected by the CC, the second beam being determined as a beam including a highest score among the set of beams in the candidate beam pool.

2. The UE of claim 1 , wherein the consecutive measurement results comprise at least one of a reference signal received power (RSRP), a reference signal received quality (RSRQ), a signal to interference and noise ratio (SINR), an identification (ID), or location information of the UEs.

3. The UE of claim 1 , wherein:

acceptability of the measurement result of the first beam is determined, by the CC, based on a threshold;

a UE connection vector based on the measurement result and the threshold is determined by the CC;

the UE connection vector to a two-dimensional connection (2D) map based on location information of the UEs is reshaped by the CC; and

the beam scores based on previous beam scores and the consecutive measurement results is calculated by the CC.

4. The UE of claim 1 , wherein:

a deep neural network to select a best beam is configured by the CC;

the best beam using the deep neural network is selected by the CC;

the preprocessed consecutive measurement results within a time window from the each of the BSs is inputted, by the CC, to the deep neural network of the BSs;

a one-hot output of the deep neural network is identified, by the CC, as an indicator for the best beam; and

weights of the deep neural network are updated, by the CC, based on the preprocessed consecutive measurement results.

5. The UE of claim 1 , wherein:

UE distribution patterns are obtained, by the CC, based on a history of measurement results from a UE;

ray-tracing data with the UE distribution patterns is combined, by the CC, to obtain scenario-specific ray-tracing data; and

a best beam based on the scenario-specific ray-tracing data is searched, by the CC, for each of the UE distribution patterns.

6. The UE of claim 5 , where:

the measurement results are classified, by the CC, into the UE distribution patterns; and

the best beam for the UE distribution patterns is selected, by the CC, based on the scenario-specific ray-tracing data.

7. The UE of claim 1 , wherein each of the BSs comprises a data unit/multi-input-multi-out (DU/MMU).

8. A central controller (CC) in a wireless communication system, the CC comprising:

a processor configured to:

select, from a set of beams in a candidate beam pool including predetermined candidate beams, a first beam for each of base stations (BSs), wherein the set of beams is allocated to the BSs; and

instruct each of the BSs to transmit, to user equipments (UEs) belonging to each of the BSs, signals with the first beam that is used in a measurement report by the UEs; and

a transceiver operably connected to the processor, the transceiver configured to:

transmit, to each of the BSs, information of the first beam;

receive, from the UEs via the BSs, the measurement report including a measurement result of the first beam,

wherein the processor is further configured to:

preprocess, for each of the BSs, consecutive measurement results that correspond to the set of beams in the candidate beam pool along with the measurement result of the first beam;

calculate, for each of the BSs, beam scores for the first beam based on the consecutive measurement results of the first beam; and

select, for each of the BSs, a second beam based on the beam scores,

wherein, for each of the BSs, the second beam is determined as a beam including a highest score among the set of beams in the candidate beam pool.

9. The CC of claim 8 , wherein the consecutive measurement results comprise at least one of a reference signal received power (RSRP), a reference signal received quality (RSRQ), a signal to interference and noise ratio (SINR), an identification (ID), or location information of the UEs.

10. The CC of claim 8 , wherein the processor is further configured to:

determine whether the measurement result of the first beam is acceptable based on a threshold;

identify a UE connection vector based on the determined measurement result and the threshold;

reshape the UE connection vector to a two-dimensional connection (2D) map based on location information of the UEs; and

calculate, for each of the BSs, the beam scores based on previous beam scores and the consecutive measurement results.

11. The CC of claim 8 , wherein the processor is further configured to:

configure, for each of the BSs, a deep neural network to generate a score on each beam for each cell and to select a best beam;

select, for each of the BSs, the best beam using the deep neural network;

input, to the deep neural network of the BSs, the preprocessed consecutive measurement results within a time window from the each of the BSs;

identify a one-hot output of the deep neural network as an indicator for the best beam; and

update, for each of the BSs, weights of the deep neural network based on the preprocessed consecutive measurement results.

12. The CC of claim 8 , wherein the processor is further configured to:

obtain, for each of the BSs, UE distribution patterns based on a history of measurement results from a UE;

combine, for each of the BSs, ray-tracing data with the UE distribution patterns to obtain scenario-specific ray-tracing data; and

search a best beam based on the scenario-specific ray-tracing data for each of the UE distribution patterns.

13. The CC of claim 12 , where the processor is further configured to:

classify the measurement results into the UE distribution patterns; and

select the best beam for the UE distribution patterns based on the scenario-specific ray-tracing data.

14. The CC of claim 8 , wherein each of the BSs comprises a data unit/multi-input-multi-out (DU/MMU).

15. A method of a central controller (CC) in a wireless communication system, the method comprising:

selecting, from a set of beams in a candidate beam pool including predetermined candidate beams, a first beam for each of base stations (BSs), wherein the set of beams is allocated to the BSs;

instructing each of the BSs to transmit, to user equipments (UEs) belonging to each of the BSs, signals with the first beam that is used in a measurement report by the UEs;

transmitting, to each of the BSs, information of the first beam;

receiving, from the UEs via the BSs, the measurement report including a measurement result of the first beam;

preprocessing, for each of the BSs, consecutive measurement results that correspond to the set of beams in the candidate beam pool along with the measurement result of the first beam;

calculating, for each of the BSs, beam scores for the first beam based on the consecutive measurement results of the first beam; and

selecting, for each of the BSs, a second beam based on the beam scores,

wherein, for each of the BSs, the second beam is determined as a beam including a highest score among the set of beams in the candidate beam pool.

16. The method of claim 15 , wherein the consecutive measurement results comprise at least one of a reference signal received power (RSRP), a reference signal received quality (RSRQ), a signal to interference and noise ratio (SINR), an identification (ID), or location information of the UEs.

17. The method of claim 15 , further comprising:

determining whether the measurement result of the first beam is acceptable based on a threshold;

identifying a UE connection vector based on the determined measurement result and the threshold;

reshaping the UE connection vector to a two-dimensional connection (2D) map based on location information of the UEs; and

calculating, for each of the BSs, the beam scores based on previous beam scores and the consecutive measurement results.

18. The method of claim 15 , wherein:

configuring, for each of the BSs, a deep neural network to select a best beam;

generating, for each of the BSs, a score on each beam for each cell;

selecting, for each of the BSs, the best beam using the deep neural network;

inputting, to the deep neural network of the BSs, the preprocessed consecutive measurement results within a time window from the each of the BSs;

identifying a one-hot output of the deep neural network as an indicator for the best beam; and

updating for each of the BSs, weights of the deep neural network based on the preprocessed consecutive measurement results.

19. The method of claim 15 , further comprising:

obtaining, for each of the BSs, UE distribution patterns based on a history of measurement results from a UE;

combining, for each of the BSs, ray-tracing data with the UE distribution patterns to obtain scenario-specific ray-tracing data; and

searching a best beam based on the scenario-specific ray-tracing data for each of the UE distribution patterns.

20. The method of claim 19 , further comprising:

classifying the measurement results into the UE distribution patterns; and

selecting the best beam for the UE distribution patterns based on the scenario-specific ray-tracing data.