A 5G load balancing method based on neighboring cell clusters

Abstract

The present invention relates to a 5G load balancing method based on neighboring cell clusters, which solves the problems of low transfer efficiency, slow equilibrium convergence, and degradation of cell performance after load balancing in the existing 5G load balancing methods. The method includes determining a load balancing source cell; obtaining the neighboring cells of the source cell and subdividing each neighboring cell into neighboring cell clusters; performing a status evaluation value on the neighboring cell clusters according to the cell disconnection rate and congestion rate to determine the target neighboring cell cluster of the source cell; performing load transfer from the source cell to each neighboring cell within the target neighboring cell cluster and updating the load information; setting a convergence condition and performing a convergence judgment on the updated cells until the convergence condition is satisfied. The present invention sets an evaluation coefficient to optimize the utilization rate of system resources; implementing load transfer between the source cell and the target neighboring cell cluster can enable the system to reach the cell equilibrium state in the shortest possible time and provide better service quality guarantee for users.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular to a 5G load balancing method based on a neighboring cell cluster. Background Art

[0002] Load balancing is an important part of the self-organizing network SON. Without manual intervention, it automatically detects overloaded cells and selects neighboring cells with less load to transfer the load, so as to optimize the system configuration and improve the customer perception of users. Due to the large bandwidth and high cell throughput of the 5G system, load balancing is more needed than traditional 3 / 4G networks. A Chinese invention patent with the patent number ZL201410324613.0 and the name of a load balancing method based on a neighboring cell set (NSLB) solves the problem of traffic imbalance by analyzing the load space that the cell with the minimum load can accommodate, and then inversely deducing the amount of users to be transferred in the neighboring cell set. However, NSLB still has many problems. On the one hand, it is based on a single neighboring cell, with low load transfer efficiency and slow equilibrium convergence. On the other hand, it does not evaluate the cell performance of neighboring cells, but only based on the load, which may lead to a decrease in cell performance after load balancing. Summary of the Invention

[0003] The present invention mainly solves the problems of low transfer efficiency, slow equilibrium convergence, and decreased cell performance after load balancing in the existing 5G load balancing method, and provides a 5G load balancing method based on a neighboring cell cluster. The present invention starts from the source cell, constructs its neighboring cell cluster, analyzes the cell states of each neighboring cell cluster, and screens out the optimal cell for load transfer, so as to achieve the goal of fast and effective load balancing while optimizing cell performance.

[0004] The above technical problems of the present invention are mainly solved by the following technical solutions: A 5G load balancing method based on a neighboring cell cluster, characterized in that it includes the following steps,

[0005] Step 1: Obtain cell load information;

[0006] Step 2: Determine the load balancing source cell according to the maximum load space;

[0007] Step 3: Obtain the neighboring cells of the source cell that meet the load space requirements, and subdivide each neighboring cell into neighboring cell clusters according to the single-cluster load transfer amount;

[0008] Step 4: Evaluate the state of the neighboring cell cluster according to the cell drop rate and congestion rate, and determine the target neighboring cell cluster of the source cell;

[0009] Step 5: Transfer the load from the source cell to each neighboring cell in the target neighboring cell cluster and update the load information;

[0010] Step 6: Set the convergence condition, and perform convergence judgment on the updated cells until the convergence condition is met.

[0011] The present invention obtains the load space of each cell in real time, determines the source cell according to the minimum load space, determines the number of source cell neighbor clusters according to the available resource amount, and then determines the upper limit of the load that can be transferred in each neighbor cluster. Analyze the performance of each lightly loaded neighbor of the maximum load cell, and finally determine the target neighbor cluster. According to the minimization principle, finally determine the load amount transferred to the target neighbor cluster; reasonably set the evaluation coefficient to optimize the system resource utilization rate; implement load transfer between the source cell and the target neighbor cluster, which can make the system reach the cell balance state in the shortest possible time and provide better service quality guarantee for customers.

[0012] As a preferred solution, the cell load information in step 1 specifically includes n cells NBS that are adjacent to each other n ={gNB 1 , gNB 2 , …, gNB n}, and the corresponding load {Cld 1 , Cld 2 , …, Cld n}, the total number of physical resource blocks {PrT 1 , PrT 2 , …, PrT n}, the disconnection rate {DrL 1 , DrL 2 , …, DrL n} and the congestion rate {CgT 1 , CgT 2 , …, CgT n}.

[0013] As a preferred solution, step 2 specifically includes the following steps:

[0014] (2-1): For NBS n ={gNB 1 , gNB 2 , …, gNB n}, calculate the mathematical expectation of the load Calculate the load space CSp of each cell i =Cld av -Cld i ;

[0015] (2-2): Select the cell gNB i with the minimum CSp s as the source cell for load balancing. If there are multiple minimum values, take the first one.

[0016] Calculate the mathematical expectation of the computing load, find the difference between the mathematical expectation and the load of each cell, obtain the load space of each cell, and select the cell with the minimum load space as the source cell for load balancing. The blood transfusion expectation of the load is the ratio of the sum of the loads of all cells to the number of cells.

[0017] As a preferred solution, step three specifically includes the following steps:

[0018] (3-1): Arrange the neighboring cells of the source cell gNB i whose load space CSp s >0 in descending order of load space to obtain its neighboring cell list The number of neighboring cells s m ∈(0, n - 1], calculate the sum of the load spaces of each neighboring cell in the neighboring cell list

[0019] (3-2): Calculate the number of neighboring cell clusters where abs() represents the absolute value function and ceil() represents the ceiling function; calculate the single-cluster load transfer amount Δ sc = abs(CSp s ) / N NCC ;

[0020] (3-3): Starting from the first position of the neighboring cell list NCL s , divide it into NCC sub-lists NCL s ={NCL 1 , NCL 2 ,…NCL NCC}, the number of cells in the corresponding sub-lists are {NC 1 , NC 2 ,…NC NCC}, where, NC 1 +NC 2 +…+NC NCC = s m , and the sum of the load spaces of the cells in the sub-list NCL j , j≠NCC Then the sum of the load spaces of the cells in the last sub-list NCL NCC Each sub-list is a neighboring cell cluster.

[0021] Obtain the neighboring cells of the source cell whose load space is positive, that is, the load space that can still accommodate resources, and arrange them in descending order of load space to obtain the neighboring cell list. Calculate the sum of the load spaces of each neighboring cell in the neighboring cell list, determine the number of neighboring cell clusters of the source cell according to the sum of the load spaces, and then determine the single-cluster load transfer amount, thereby dividing the cells in the neighboring cell list into neighboring cell clusters.

[0022] As a preferred solution, step four specifically includes the following steps:

[0023] (4-1): Set the evaluation coefficient For each cell gNB in NBS n ={gNB 1 ,gNB 2 ,…,gNB n}, calculate the cell status evaluation value i

[0024] (4-2): Calculate the status evaluation value of each sub-table NCL s of NCL j ,j = 1,2…,NCC where Sta j represents the status evaluation value of any cell gNB j in the sub-table NCL j 1 ;

[0025] (4-3): Select a cluster NCL 1 ,NCL 2 ,…NCL NCC with the minimum as the target neighbor cell cluster of the source cell gNB t s t ;

[0026] Reasonably set the evaluation coefficient to calculate the status evaluation value of each cell, and then obtain the status evaluation value of each neighbor cell cluster, and select the cluster with the minimum status evaluation value as the target neighbor cell cluster.

[0027] As a preferred solution, step five specifically includes the following steps:

[0028] (5-1): Set each cell in the target neighbor cell cluster NCL t to be where NC t is the number of cells in the target neighbor cell cluster, set the source cell gNB s and the user load in the common coverage area of each cell gNB t in NCL m Then gNB s transfers the load to where min() represents the minimum value function; transfer the load to

[0029] (5-2): Update the load of the source cell gNB s ​ Update the target neighbor cell cluster NCL t For each cell gNB m in it, the load Cld m = Cld m + Δ m * PrT s / PrT m .

[0030] Determine the load transfer amount of each cell from the load space of each cell in the target neighbor cell cluster, the user load amount in the common coverage area, and the single-cluster load transfer amount according to the minimization principle, perform load transfer from the source cell to each cell, and update the loads of the source cell and each cell in the neighbor cell cluster after the transfer.

[0031] As a preferred solution, step six specifically includes the following steps:

[0032] Set an acceptable equilibrium state β g , a convergence target value ε, and an upper limit Nm of the number of convergence times th ; For NBS n = {gNB 1 , gNB 2 , …, gNB n}, calculate the load balancing factor The number of convergence times Nm L = Nm L + 1, verify whether the condition or the condition Nm L ≥ Nm th is satisfied. If it is satisfied, end. If it is not satisfied, repeat steps two to six until this condition is satisfied.

[0033] Therefore, the advantages of the present invention are as follows: Implement to obtain the load space of each cell, determine the source cell according to the minimum load space, obtain the subdivision number of the neighbor cell cluster based on the total load space of the lightly loaded neighbor cells, and calculate the load transfer amount of each neighbor cell cluster; Obtain the current performance state of the cell according to the cell index statistics, and then obtain the target neighbor cell cluster according to the minimization principle; Set an evaluation coefficient to optimize the system resource utilization rate; Implement load transfer between the source cell and the target neighbor cell cluster, which can make the system reach the cell equilibrium state in the shortest possible time and provide better service quality guarantee for users. Brief Description of the Drawings

[0034] Figure 1 is a flowchart of the present invention;

[0035] Figure 2 is a comparison chart of the convergence efficiency between the method of the present invention and other algorithms;

[0036] Figure 3It is a comparison chart of the offloading process of the maximum overloaded cell between the method of the present invention and other algorithms;

[0037] Figure 4 It is a comparison chart of the cell throughput between the method of the present invention and other algorithms. Specific embodiments

[0038] The technical solution of the present invention will be further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0039] Embodiment:

[0040] A 5G load balancing method based on neighbor cell clusters in this embodiment includes the following steps:

[0041] Step 1: Obtain cell load information; specifically, it includes n cells NBS that are adjacent to each other n ={gNB 1 , gNB 2 , …, gNB n}, and the corresponding loads {Cld 1 , Cld 2 , …, Cld m}, total number of physical resource blocks {PrT 1 , PrT 2 , …, PrT n}, disconnection rate {DrL 1 , DrL 2 , …, DrL m} and congestion rate {CgT 1 , CgT 2 , …, CgT n}.

[0042] Step 2: Determine the load balancing source cell according to the maximum load space; specifically, it includes the following steps:

[0043] (2-1): For NBS n ={gNB 1 , gNB 2 , …, gNB n}, calculate the mathematical expectation of the load Calculate the load space CSp of each cell i =Cld av -Cld i ;

[0044] (2-2): Select the cell gNB i with the minimum CSp s as the load balancing source cell. If there are multiple minimum values, the first one can be taken.

[0045] Step 3: Obtain the neighboring cells of the source cell that meet the load space requirements, and subdivide each neighboring cell into neighboring cell clusters according to the single-cluster load transfer amount; specifically, it includes the following steps:

[0046] (3-1): Arrange the neighboring cells of the source cell gNB with a load space CSp i >0 in descending order of load space to obtain its neighboring cell list s For the number of neighboring cells s ∈(0, n - 1], calculate the sum of the load spaces of each neighboring cell in the neighboring cell list m

[0047] (3-2): Calculate the number of neighboring cell clusters where abs() represents the absolute value function and ceil() represents the ceiling function; calculate the single-cluster load transfer amount Δ sc = abs(CSp s ) / N NCC ;

[0048] (3-3): Starting from the first position of the neighboring cell list NCL s , subdivide it into NCC sub-lists NCL s = {NCL 1 , NCL 2 , … NCL NCC}, and the number of cells in the corresponding sub-lists are {NC 1 , NC 2 , … NC NCC}, where NC 1 + NC 2 + … + NC NCC = s m , and the sum of the load spaces of the cells in the sub-list NCL j , j ≠ NCC Then the sum of the load spaces of the cells in the last sub-list NCL NCC

[0049] Step 4: Perform a status evaluation value on the neighboring cell clusters according to the cell disconnection rate and congestion rate, and determine the target neighboring cell cluster of the source cell; specifically, it includes the following steps:

[0050] (4-1): Set the evaluation coefficient For each cell gNB in NBS n = {gNB 1 , gNB 2 , …, gNB n}, calculate the cell status evaluation value i ​​​

[0051] (4-2): Calculate NCL s for each sub-table NCL j , j = 1, 2…, the state evaluation value of NCC where Sta j represents the state evaluation value of any small cell gNB j in the sub-table NCL j ;

[0052] (4-3): Select a cluster NCL 1 , NCL 2 , … NCL NCC with the smallest as the target neighbor cell cluster of the source cell gNB t . s

[0053] Step Five: Transfer the load from the source cell to each neighbor cell in the target neighbor cell cluster and update the load information; specifically, it includes the following steps:

[0054] (5-1): Set each cell in the target neighbor cell cluster NCL t to be respectively where NC t is the number of cells in the target neighbor cell cluster, set the source cell gNB s and the common coverage area user load of each cell gNB t in NCL m Then, gNB s transfers the load respectively to where min() represents the minimum value function; transfer the load to

[0055] (5-2): Update the load of the source cell gNB s Update the load Cld of each cell gNB t in the target neighbor cell cluster NCL m = Cld m + Δ m * PrT m / PrT s . m .

[0056] Step Six: Set the convergence condition, and perform convergence judgment on the updated cells until the convergence condition is met. Specifically, set the acceptable equilibrium state β 0 , the convergence target value ε, and the upper limit of the convergence times Nm th ; For NBS​​n = {gNB 1 , gNB 2 , …, gNB n}, calculate the load balancing factor Convergence times Nm L = Nm L + 1, verify whether the condition or condition Nm L ≥ Nm th is satisfied. If satisfied, end. If not satisfied, repeat steps one to six until the condition is met.

[0057] The following uses an example to illustrate this embodiment. Taking n = 8 as an example, the cell load situation of the 5G system is shown in Table 1:

[0058] Table 1 Services carried out

[0059] Cell Initial Load Total Number of Resource Blocks Drop Rate Congestion Rate Remarks <![CDATA[gNB 1 > 0.14 273 0.018 0.024 Mutually Adjacent Cells <![CDATA[gNB 2 > 0.25 273 0.021 0.019 Mutually Adjacent Cells <![CDATA[gNB 3 > 0.33 273 0.013 0.027 Mutually Adjacent Cells <![CDATA[gNB 4 > 0.17 273 0.034 0.033 Mutually Adjacent Cells <![CDATA[gNB 5 > 0.73 273 0.046 0.028 Mutually Adjacent Cells <![CDATA[gNB 6 > 0.46 273 0.019 0.022 Mutually Adjacent Cells <![CDATA[gNB 7 > 0.28 273 0.024 0.021 Mutually Adjacent Cells <![CDATA[gNB 8 > 0.39 273 0.011 0.012 Mutually Adjacent Cells

[0060] The basic data is shown in Table 2:

[0061] Table 2 Basic data

[0062]

[0063] A 5G load balancing method based on neighbor cell clusters includes the following steps:

[0064] Step 1: Obtain cell load information;

[0065] Step 2: Determine the load balancing source cell according to the maximum load space;

[0066] (2-1): For NBS n = {gNB 1 , gNB 2 , …, gNB n}, calculate the mathematical expectation of the load Calculate the load space of each cell,

[0067] CSp i = Cld av - Cld i

[0068] = {0.204, 0.094, 0.014, 0.174, -0.386, -0.116, 0.064, -0.046};

[0069] (2-2): The cell gNB i with the minimum CSp s＝5 = -0.39 is used as the load balancing source cell.

[0070] Step 3: Obtain the neighboring cells of the source cell that meet the load space requirements, and subdivide each neighboring cell into neighboring cell clusters according to the single-cluster load transfer volume;

[0071] (3-1): Arrange the neighboring cells of the source cell gNB with load space CSp i >0 in descending order of load space to obtain its neighboring cell list, s The number of its neighboring cells is s

[0072]

[0073] =5, calculate the sum of its load spaces, m

[0074]

[0075] (3-2): Calculate the number of neighboring cell clusters,

[0076]

[0077] Calculate the single-cluster load transfer volume,

[0078]

[0079] (3-3): Starting from the first position of the neighboring cell list NCL s , subdivide it into NCC = 3 sub-lists,

[0080] NCL s ={NCL 1 ,NCL 2 ,…NCL NCC}

[0081] ={{gNB 1},{gNB 4},{gNB 2 ,gNB 7 ,gNB 3}}

[0082] The number of cells in the corresponding sub-lists are {NC 1 ,NC 2 ,…NC NCC}={1,1,3}, and the sum of the load spaces of the cells in the sub-list NCL j (j≠NCC) Then the sum of the load spaces of the cells in the last sub-list NCL NCC

[0083]

[0084] ​​Step 4: Evaluate the state of the neighboring cell clusters based on the cell drop rate and congestion rate, and determine the target neighboring cell clusters of the source cell;

[0085] (4-1): Evaluation coefficient For each cell gNB n ={gNB 1 ,gNB 2 ,…,gNB n} in NBS i , calculate the cell state evaluation value,

[0086]

[0087] (4-2): Calculate the state evaluation value of each sub-table NCL s of NCL j , j = 1, 2…, NCC,

[0088]

[0089] (4-3): Select a cluster NCL 1 , NCL 2 ,…NCL NCC with the minimum as the target neighboring cell cluster {gNB 1} of the source cell gNB s . 1}

[0090] Step 5: Transfer the load from the source cell to each neighboring cell in the target neighboring cell cluster and update the load information;

[0091] (5-1): If there is only the cell gNB 1 in the neighboring cell cluster NCL 1 , then transfer the load from the source cell gNB s＝5 to the cell gNB 1 . Set the user load volume in their common coverage area Then the load volume s＝5 transferred from gNB 1 to gNB

[0092] (5-2): Update the load of the source cell gNB s ,

[0093]

[0094] Update the load of the cell gNB t in the target neighboring cell cluster NCL 1 ,

[0095]

[0096] Step 6: Set the convergence condition, and perform convergence judgment on the updated cell until the convergence condition is met.

[0097] For NBS n ={gNB 1 , gNB 2 , …, gNB n}, calculate the load balancing factor.

[0098]

[0099] The number of convergence times Nm L = Nm L +1 = 1, and it does not meet the condition abs(β - β g ) = 0.1206 ≤ ε = 0.01% or the condition Nm L ≥ Nm th . Repeat Steps 1 to 6 until this condition is met.

[0100] The above method is verified by simulation experiments below. The NCCLB method of the present invention and the previous NSLB method are simulated on the MATLAB platform. The basic data information is shown in Tables 1 and 2 above. A certain number of users are randomly scattered and random services are configured. The obtained results are respectively shown in Figures 2 to 4 as follows.

[0101] As Figure 2 shown in the load balancing convergence process of the two methods. NSLB converges after about 107 cycles. NCCLB adopts neighboring cell cluster matching, with a faster load transfer speed and higher convergence efficiency, and approximately experiences 71 cycles;

[0102] As Figure 3 shown, NCCLB takes only a little more than half the time of NSLB to overload, and can optimize the system configuration faster. However, correspondingly, it has slightly higher requirements for system resources;

[0103] As Figure 4 shown in the cell throughput curves of the NCCLB and NSLB algorithms. The former can select the target neighboring cell cluster with low disconnection rate and congestion rate and good performance state by evaluating the system, so that the overall throughput of the cell can be stably at a high level, while NSLB is relatively random and the algorithm itself has weak control over cell performance.

[0104] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Those skilled in the art of the present invention can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, but will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A 5G load balancing method based on neighbor cell clusters, characterized in that: It includes the following steps, Step 1: Obtain cell load information; Step 2: Calculate the difference between the load mathematical expectation and the load of each cell to obtain the load space of each cell, and select the cell with the smallest load space to determine the load balancing source cell; Step 3: Obtain the neighbor cells of the source cell that satisfy the load space being greater than zero, calculate the sum of the load spaces of each neighbor cell, determine the number of neighbor cell clusters of the source cell according to the sum of the load spaces, and then determine the single-cluster load transfer amount. Subdivide each neighbor cell according to the single-cluster load transfer amount; Step 4: Calculate the cell status evaluation value = evaluation coefficient * cell disconnection rate + (1 - evaluation coefficient) * cell congestion rate, where the evaluation coefficient ∈ [0, 1]; Step 5: Transfer the load from the source cell to each neighbor cell within the target neighbor cell cluster and update the load information; Step 6: Set the convergence condition, and perform a convergence judgment on the updated cells until the convergence condition is met.

2. A 5G load balancing method based on neighbor cell clusters according to claim 1, characterized in that The cell load information in the first step specifically includes n cells NBS that are adjacent to each other n ={gNB 1 , gNB 2 , …, gNB n}, and the corresponding loads {Cld 1 , Cld 2 , …, Cld n}, total number of physical resource blocks {PrT 1 , PrT 2 , …, PrT n}, disconnection rate {DrL 1 , DrL 2 , …, DrL n} and congestion rate {CgT 1 , CgT 2 , …, CgT n}.

3. A 5G load balancing method based on neighbor cell clusters according to claim 2, characterized in that The specific steps of Step 2 include the following steps: (2-1): For NBS n ={gNB 1 , gNB 2 , …, gNB n}, calculate the mathematical expectation of the load Calculate the load space CSp of each cell i = Cld av - Cld i ; (2-2): Select CSp i The cell gNB with the minimum value s , as the source cell for load balancing.

4. A 5G load balancing method based on neighbor cell clusters according to claim 3, characterized in that The specific steps of Step 3 include the following steps: (3-1): The source cell gNB i >0 with a load space CSp s sorts the neighboring cells in descending order of the load space to obtain its neighboring cell list The number of neighboring cells s m ∈ (0, n - 1], calculate the sum of the load spaces of each neighboring cell in the neighboring cell list (3-2): Calculate the number of neighboring cell clusters where abs() represents the absolute value function and ceil() represents the ceiling function; calculate the single-cluster load transfer amount Δ sc = abs(CSp s ) / NCC; (3-3): Starting from the first position of the neighbor cell list NCL s , divide it into NCC sub-lists NCL s = {NCL 1 , NCL 2 , … NCL NCC}, and the number of cells in the corresponding sub-lists are {NC 1 , NC 2 , … NC NCC}, where NC 1 + NC 2 + … + NC NCC = s m , and the sum of the cell load spaces in the sub-list NCL j , j ≠ NCC Then the sum of the cell load spaces in the last sub-list NCL NCC ​ 5. A 5G load balancing method based on neighbor cell clusters according to claim 4, characterized in that The specific steps of Step 4 include the following steps: (4-1): Set the evaluation coefficient For NBS n = {gNB 1 , gNB 2 , …, gNB n}, for each cell gNB i , calculate the cell status evaluation value (4-2): Calculate NCL s for each sub-table NCL j , j = 1, 2…, the status evaluation value of NCC where Sta j represents the status evaluation value of any cell gNB j in the sub-table NCL j ; (4-3): Select one with the minimum 1 , NCL 2 , … NCL NCC from the sub-tables NCL as the target neighbor cell cluster of the source cell gNB t s s .

6. A 5G load balancing method based on neighbor cell clusters according to claim 5, characterized in that the The specific steps of Step 5 include the following steps: (5-1): Set the target neighbor cell cluster NCL t Each cell within is Among them, NC t is the number of cells within the target neighbor cell cluster. Set the source cell gNB s and NCL t The user load in the common coverage area of each cell gNB within m is Then, gNB s respectively transfers the load to The transferred load Among them, min() represents the minimum value function; transfer the load to The transferred load (5-2): Update source cell gNB s Load Update target neighbor cell cluster NCL t Each cell gNB in m The load Cld of m = Cld m + Δ m * PrT s / PrT m .

7. A 5G load balancing method based on neighbor cell clusters according to claim 6, characterized in that The specific steps of Step 6 include the following steps: Set the acceptable equilibrium state β 0 , the convergence target value ε, and the upper limit of the number of convergence times Nm th ; For NBS n = {gNB 1 , gNB 2 , …, gNB n}, calculate the load balancing factor The number of convergence times Nm L = Nm L + 1, verify whether the condition or the condition Nm L ≥ Nm th is satisfied. If satisfied, end. If not satisfied, repeat steps one to six until the condition is met.

Images