Granted Patent B2

US 12,395,891 · App. 17/964,597 · Granted Aug 19, 2025

Inventors: Tae Jung Kim (Daejeon, KR); Donghun Lee (Daejeon, KR); Jeehyeon Na (Daejeon, KR); Jung Mo Moon (Daejeon, KR)

Assignee: ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE

H04W28/0942H04W24/02H04W28/0861

Abstract

Claims (52)

1. A base station load balancing method in a base station, the method comprising:

updating a traffic amount of a current time by reflecting a predicted traffic amount of a future time in the traffic amount of the current time; and

determining parameters necessary for load balancing of the base station by comparing the updated traffic amount of the current time with a predetermined threshold,

wherein the updating includes predicting a traffic amount of the future time from the traffic amount of the current time using a prediction model, and

wherein a weight value of the prediction model is determined through federated learning between a service and operation management system and a plurality of base stations.

2. The method of claim 1 , further comprising:

learning a machine learning model using learning data;

updating the machine learning model; and

repeating the learning of a machine learning model and updating of the machine learning model to use the machine learning model as the predictive model.

3. The method of claim 2 , wherein the updating of the machine learning model includes:

transmitting a weight value and learning accuracy according to the learning result of the machine learning model to the service and operation management system;

receiving a global weight value determined by the service and operation management system based on the weight values and learning accuracy received from the plurality of base stations; and

updating a weight value of the machine learning model with the global weight value.

4. The method of claim 3 , wherein the global weight value is an average value of part of the weight values received from the plurality of base stations, and

the part of the weight values is randomly selected according to a standard deviation for learning accuracy provided by the plurality of base stations, or are selected in the order of high learning accuracy.

5. The method of claim 1 , wherein the updating includes

calculating the traffic amount at the current time by applying a first weight value and a second weight value to downlink physical resource block (PRB) usage and uplink PRB usage, respectively, and

the first weight value and the second weight value are determined according to a ratio of PRBs allocated to downlink and uplink in the entire PRB.

6. The method of claim 1 , wherein the updating includes applying a first weight value to the traffic amount of the current time and applying a second weight value to the predicted traffic amount of the future time, and

the first weight value is set to be greater than the second weight value.

7. The method of claim 1 , wherein the determining includes adjusting handover-related parameters so that the terminals at the edge of the base station move to another base station earlier if the updated traffic amount at the current time is greater than the threshold value.

8. A base station load balancing method for balancing a load of a plurality of base stations in a base station load control apparatus, the method comprising:

generating and managing policies necessary for a load balancing operation;

modifying the policies using a load balancing result of an overloaded base station; and

determining a weight value of a machine learning model used for predicting traffic of a future time in the plurality of base stations by performing federated learning with the plurality of base stations.

9. The method of claim 8 , wherein the determining includes:

receiving a weight value and learning accuracy according to a learning result of the machine learning model from the plurality of base stations, respectively;

calculating a global weight value based on the weight values and learning accuracy received from the plurality of base stations; and

transmitting the global weight value to the plurality of base stations so that the plurality of base stations update the weight values of the machine learning model with the global weight value.

10. The method of claim 9 , wherein the calculating includes:

selecting weight values of part of the weight values received from the plurality of base stations according to the standard deviation for the learning accuracy provided by the plurality of base stations; and

calculating an average of the selected weight values of the part as the global weight value.

11. The method of claim 10 , wherein the selecting includes:

randomly selecting the weight values of the part if a random value between 0 and 1 is less than a value corresponding to the standard deviation; and

selecting the weight values of the part in order of high learning accuracy if the random value is equal to or greater than a value calculated based on the standard deviation.

12. The method of claim 11 , wherein the value corresponding to the standard deviation is calculated based on Equation 1,

the Equation 1 is

E =min(1,δ×Std),0≤δ<1, and

wherein the δ has a value between 0 and 1, and the Std represents the standard deviation.

13. A base station load balancing apparatus for load balancing by interworking with a service and operation management system in a base station, the apparatus comprising:

a local traffic predictor that predicts a traffic amount in a future time using a prediction model; and

a load balancing processor that updates a traffic amount of a current time by reflecting the predicted traffic amount of the future time in the traffic amount of the current time, and determines parameters necessary for load balancing of the base station by comparing the updated traffic amount of the current time with a predetermined threshold,

wherein the local traffic predictor includes:

a model updater for updating a machine learning model with a global weight value determined by the service and operation management system through federated learning between a plurality of base stations and the service and operation management system; and

a model learner for learning the updated machine learning model, and wherein the prediction model is finally generated through repetition of the learning the machine learning model and updating the machine learning model.

14. The apparatus of claim 13 , wherein the model learner transmits a weight value and learning accuracy according to a learning result of the machine learning model to the service and operation management system, and

the global weight value is determined by the service and operation management system based on the weight values and learning accuracy received from the plurality of base stations.

15. The apparatus of claim 14 , wherein the global weight value is an average value of part of the weight values received from the plurality of base stations, and

the part of the weight values is randomly selected according to a standard deviation for learning accuracy provided by the plurality of base stations, or are selected in the order of high learning accuracy.

16. The apparatus of claim 13 , wherein the load balancing processor includes an overload determiner for calculating the updated traffic amount of the current time by applying a first weight value to the traffic amount of the current time and applying a second weight to the predicted traffic amount of the future time, and

the first weight value is set to be greater than the second weight value.

17. The apparatus of claim 13 , wherein the load balancing processor includes a mobility load balancing (MLB) controller for adjusting handover-related parameters so that the terminals at the edge of the base station move to another base station earlier if the updated traffic amount at the current time is greater than the threshold value.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (10)