Granted Patent B2

US 12,507,121 · App. 17/129,764 · Granted Dec 23, 2025

Inventors: Hao Chen (Allen, TX); Zhou Zhou (Allen, TX); Yan Xin (Princeton, NJ); Jianzhong Zhang (Plano, TX)

Assignee: Samsung Electronics Co., Ltd.

H04W28/0942G06N3/045G06N20/00H04W28/0226H04W28/08H04W88/085

Abstract

Claims (94)

1 . A method for performing mobility load balancing, the method comprising:

receiving, at a server, current load data for a plurality of cells of a wireless communication network;

selecting, by the server, from the plurality of cells, a target cell (s), wherein a value of the current load data for the target cell exceeds a first predefined threshold;

selecting, by the server, from a neighbor cell list corresponding to the target cell, a set of neighboring cells for the target cell, wherein the selected set of neighboring cells (S) is a subset of the neighbor cell list;

calculating, by the server, a value of at least one utilization parameter for the target cell for a new load through the target cell, wherein the value of the at least one utilization parameter for the target cell comprises a plurality of values of a physical resource block (PRB) usage of the target cell and the selected set of neighboring cells;

determining, by the server, a cell individual offset (CIO) value and an electrical tilt (E-tilt) value for the target cell based on the value of the at least one utilization parameter for the target cell, wherein determining the CIO value and the E-tilt value further comprises:

determining, by the server, information about PRB usage for a union of S and s; and

determining whether the information about the PRB usage indicates a potentially curable imbalance in a distribution of network load across the cells of S and s; and

configuring one or more physical layer parameters of the target cell based on the determined CIO and E-tilt values for the target cell.

2 . The method of claim 1 , wherein the value of the at least one utilization parameter for the target cell comprises a ratio of cell edge user equipment (UE).

3 . The method of claim 1 , wherein determining the CIO value and the E-tilt value further comprises:

determining, by the server, a value of σ(S∪s), wherein σ(S∪s) comprises a measure of a standard deviation of the PRB usage for the union of S and s;

determining that the value of σ(S∪s) indicates the potentially curable imbalance in the distribution of network load across the cells of S and s by determining that the value of σ(S∪s) exceeds a second predetermined threshold; and

based on the value of σ(S∪s) exceeding the second predetermined threshold:

determining, by the server, for cells of (S∪s), a value of at least one utilization parameter;

determining, by the server, for cells of (S∪s), CIO values and E-tilt values based on the value of the at least one utilization parameter; and

configuring, for cells of (S∪s), one or more physical layer parameters based on the determined CIO and E-tilt values.

4 . The method of claim 3 , further comprising:

based on the value of σ(S∪s) being less than the second predetermined threshold, performing an iterative selection loop, performing the iterative selection loop comprising:

determining a cell of S with a highest value of a PRB metric;

selecting S , where S is a subset of S from which the cell of S with the highest value of the PRB metric has been removed;

determining a value of σ(S∪s); and

based on the value of σ(S∪s) exceeding the second predetermined threshold, selecting the cells of S for CIO and E-tilt optimization.

5 . The method of claim 4 , wherein, based on the value of σ( S ∪s) being less than the second predetermined threshold, the iterative selection loop is performed for 5.

6 . The method of claim 1 , wherein at least one of the CIO value and E-tilt value for the target cell are determined by applying the value of the at least one utilization parameter by using a deep reinforced learning (DRL) model, and further comprising:

for the target cell and a selected set of neighbor cells, reading observation data from at least one digital unit (DU) and remote radio head (RRH);

feeding observation data to a neural network to obtain actions;

determining values of rewards associated with the actions; and

determining at least one of the CIO value or E-tilt value for the target cell based on an action with a highest reward value.

7 . The method of claim 1 , wherein determining the CIO value and the E-tilt value further comprises:

determining whether the information about the PRB usage indicates the potentially curable imbalance in the distribution of network load across the cells of S and s by the information exceeding a second predetermined threshold; and

performing an iterative selection loop to remove a cell of S, based on the information not exceeding the second predetermined threshold.

8 . A server comprising:

a network interface configured to receive current load data for a plurality of cells of a wireless communication network; and

a processor operably connected to the network interface, the processor configured to:

select, from the plurality of cells, a target cell (s), wherein a value of the current load data for the target cell exceeds a first predefined threshold;

select, from a neighbor cell list corresponding to the target cell, a set of neighboring cells for the target cell, wherein the selected set of neighboring cells (S) is a subset of the neighbor cell list;

calculate, a value of at least one utilization parameter for the target cell for a new load through the target cell, wherein the value of the at least one utilization parameter for the target cell comprises a plurality of values of a physical resource block (PRB) usage of the target cell and the selected set of neighboring cells;

determine, a cell individual offset (CIO) value and an electrical tilt (E-tilt) value for the target cell based on the value of the at least one utilization parameter for the target cell, wherein to determine the CIO value and the E-tilt value, the processor is further configured to:

determine information about PRB usage for a union of S and s; and

determine whether the information about the PRB usage indicates a potentially curable imbalance in a distribution of network load across the cells of S and s; and

configure one or more physical layer parameters of the target cell based on the determined CIO and E-tilt values for the target cell.

9 . The server of claim 8 , wherein the value of the at least one utilization parameter for the target cell comprises a ratio of cell edge user equipment (UE).

10 . The server of claim 8 , wherein to determine the CIO value and the E-tilt value, the processor is further configured to:

determine a value of σ(S∪s), wherein σ(S∪s) comprises a measure of a standard deviation of the PRB usage for the union of S and s;

determine that the value of σ(S∪s) indicates the potentially curable imbalance in the distribution of network load across the cells of S and s by determining that the value of σ(S∪s) exceeds a second predetermined threshold; and

based on the value of σ(S∪s) exceeding the second predetermined threshold:

determine, for cells of (S∪s), a value of at least one utilization parameter;

determine, for cells of (S∪s), CIO values and E-tilt values based on the value of the at least one utilization parameter; and

configure, for cells of (S∪s), one or more physical layer parameters based on the determined CIO and E-tilt values.

11 . The server of claim 10 , wherein the processor is further configured to:

based on the value of σ(S∪s) being less than the second predetermined threshold, perform an iterative selection loop, wherein to perform the iterative selection loop, the processor is further configured to:

determine a cell of S with a highest value of a PRB metric;

select S , where S is a subset of S from which the cell of S with the highest value of the PRB metric has been removed;

determine a value of σ( S ∪s); and

based on the value of σ( S ∪s) exceeding the second predetermined threshold, select the cells of S for CIO and E-tilt optimization.

12 . The server of claim 11 , wherein, based on the value of σ( S ∪s) being less than the second predetermined threshold, the iterative selection loop is performed for S .

13 . The server of claim 8 , wherein at least one of the CIO value and E-tilt value for the target cell are determined by applying the value of the at least one utilization parameter by using a deep reinforced learning (DRL) model, and the processor is further configured to:

for the target cell and a selected set of neighbor cells, read observation data from at least one digital unit (DU) and remote radio head (RRH);

feed observation data to a neural network to obtain actions;

determine values of rewards associated with the actions; and

determine at least one of the CIO value or E-tilt value for the target cell based on an action with a highest reward value.

14 . The server of claim 13 , wherein the observation data comprises, PRB usage data, a ratio of edge users and throughput data.

15 . A non-transitory computer-readable medium comprising program code, which when executed by a processor of a server, causes the server to:

receive, via a network interface of the server, current load data for a plurality of cells of a wireless communication network;

select, from the plurality of cells, a target cell (s), wherein a value of the current load data for the target cell exceeds a first predefined threshold;

select, from a neighbor cell list corresponding to the target cell, a set of neighboring cells for the target cell, wherein the selected set of neighboring cells (S) is a subset of the neighbor cell list;

calculate, a value of at least one utilization parameter for the target cell for a new load through the target cell, wherein the value of the at least one utilization parameter for the target cell comprises a plurality of values of a physical resource block (PRB) usage of the target cell and the selected set of neighboring cells;

determine, a cell individual offset (CIO) value and an electrical tilt (E-tilt) value for the target cell based on the value of the at least one utilization parameter for the target cell, wherein the program code for determining the CIO value and the E-tilt value comprises program code, which, when executed by the processor, cause the server to:

determine information about PRB usage for a union of S and s; and

determine whether the information about the PRB usage indicates a potentially curable imbalance in a distribution of network load across the cells of S and s; and

configure one or more physical layer parameters of the target cell based on the determined CIO and E-tilt values for the target cell.

16 . The non-transitory computer-readable medium of claim 15 , wherein the information about the PRB usage for the union of S and s comprises a measure of a statistic of the PRB usage for the union of S and s; and

wherein the value of the at least one utilization parameter for the target cell comprises a ratio of cell edge user equipment (UE).

17 . The non-transitory computer-readable medium of claim 15 , wherein the program code for determining the CIO value and the E-tilt value comprises program code, which, when executed by the processor, cause the server to:

determine a value of o (SUs), wherein σ(S∪s) comprises a measure of a standard deviation of the PRB usage for the union of S and s;

determine that the value of σ(S∪s) indicates the potentially curable imbalance in the distribution of network load across the cells of S and s by determining that the value of σ(S∪s) exceeds a second predetermined threshold; and

based on the value of σ(S∪s) exceeding the second predetermined threshold:

determine, for cells of (S∪s), a value of at least one utilization parameter;

determine, for cells of (S∪s), CIO values and E-tilt values based on the value of the at least one utilization parameter; and

configure, for cells of (S∪s), one or more physical layer parameters based on the determined CIO and E-tilt values.

18 . The non-transitory computer-readable medium of claim 17 , further comprising program code, which when executed by the processor, cause the server to:

based on the value of σ(S∪s) being less than the second predetermined threshold, perform an iterative selection loop, wherein performing the iterative selection loop comprises:

determining a cell of S with a highest value of a PRB metric;

selecting S , where S is a subset of S from which the cell of S with the highest value of the PRB metric has been removed;

determining a value of σ( S ∪s); and

based on the value of σ( S ∪s) exceeding the second predetermined threshold, selecting the cells of S for CIO and E-tilt optimization.

19 . The non-transitory, computer-readable medium of claim 18 ,

wherein based on the value of σ( S ∪s) being less than the second predetermined threshold, the iterative selection loop is performed for S .

20 . The non-transitory, computer-readable medium of claim 15 , wherein at least one of the CIO value and E-tilt value for the target cell are determined by applying the value of the at least one utilization parameter by using a deep reinforced learning (DRL) model, and further comprising program code, which when executed by the processor, cause the server to:

for the target cell and a selected set of neighbor cells, read observation data from at least one digital unit (DU) and remote radio head (RRH);

feed observation data to a neural network to obtain actions;

determine values of rewards associated with the actions; and

determine at least one of the CIO value or E-tilt value for the target cell based on an action with a highest reward value.

Assignments (1)

Continuity (2)

References Cited (53)