A User Clustering Method for Ultra-Dense Networks

Abstract

This invention discloses a user clustering method for ultra-dense networks, relating to the field of mobile communication technology. The method includes the following steps: selecting the users with the largest user interference vector values ​​in the ultra-dense network as root node users, and the remaining users as branch users; calculating and determining the relative interference vector values ​​between each branch user and the multiple root node users; when the relative interference vector value between a branch user and a certain root node user is less than the relative interference vector value between that branch user and other root node users, then that branch user is assigned to an interference vector cluster formed with that root node user, thus forming multiple interference vector clusters; calculating and determining the relative interference vector value between the interference vector value of the root node user in each interference vector cluster and the mean interference vector value of all users in the cluster, and determining the user clustering result. This invention is simple, improves system throughput and frequency utilization, and has strong applicability.

Description

Technical Field

[0001] This invention relates to the field of mobile communication technology, specifically to a user clustering method for ultra-dense networks. Background Technology

[0002] Currently, 5G networks are being deployed on a large scale. While enjoying the convenience and speed of 5G, the large number of mobile devices accessing the network makes traditional base station structures unsustainable and prohibitively expensive, leading to the emergence of ultra-dense networking technology. Ultra-dense networking technology increases the density of micro base stations per unit area, achieving higher frequency reuse efficiency and thus improving system capacity in hotspot areas. However, as cell density increases, inter-cell interference intensifies, especially for users at the cell edge who are more significantly affected by interference from surrounding base stations. User clustering is an effective solution to the interference problem in ultra-dense networking. Existing clustering algorithms mostly approach user clustering from a geographical perspective, which fails to accurately represent the interference relationships between users, resulting in ineffective interference suppression and a lack of improved system performance. Furthermore, improvements in the clustering performance of existing algorithms often come at the cost of increased algorithm complexity, making the practical application of complex algorithms difficult. Summary of the Invention

[0003] This invention proposes a user clustering method for ultra-dense networks. This method only requires decision processing on the user's interference vector and relative interference value, which has low complexity. At the same time, the algorithm can effectively improve the system throughput and frequency utilization, making it easier to apply in practice.

[0004] This invention provides a user clustering method for ultra-dense networks, comprising the following steps:

[0005] Calculate the interference between users in an ultra-dense network and obtain the interference vector value for each user;

[0006] Users with larger user interference vector values ​​are designated as root node users, and the remaining users are designated as branch users.

[0007] Calculate and determine the relative interference vector values ​​between each branch user and multiple root node users;

[0008] When the relative interference vector value between a branch user and a root node user is less than the relative interference vector value between the branch user and other root node users, the branch user is classified into the interference vector cluster formed with the root node user, thus forming multiple interference vector clusters.

[0009] Calculate and determine the relative magnitude of the interference vector value of the root node user in each interference vector cluster to the mean of the interference vectors of all users in the interference vector cluster;

[0010] If the relative interference vector value of the root node user in a certain interference vector cluster is zero or less than a set threshold compared to the mean interference vector value of all users in the interference vector cluster, then all users in that interference vector cluster are grouped into one cluster.

[0011] Furthermore, when the relative interference vector value of the root node user in a certain interference vector cluster is not zero or not less than a set threshold compared with the average interference vector value of all users in the interference vector cluster, the average interference vector value of all users in each interference vector cluster is used again as the interference vector of the root node user, and the relative interference vector values ​​of each branch user and multiple root node users are recalculated and judged.

[0012] Furthermore, the calculation of interference between users in the ultra-dense network to obtain the interference vector value for each user includes the following steps:

[0013] Suppose the number of users is 1, 2, 3, ... N. The interference between any two users i and j is calculated using the cosine similarity of the interference channels between them, as follows:

[0014]

[0015] in, and These are user i in virtual community C. i and C j Channel vectors under the service;

[0016] and These are user j in virtual community C. i and C j Channel vectors under the service;

[0017] H H α represents the conjugate transpose of H; α and β represent coefficients for adjusting power and spatial dimension;

[0018] ||·|| F In this context, F = 2, which represents the L2 norm.

[0019] The interference matrix W is constructed by calculating the pairwise interference between N users. ab ,as follows:

[0020]

[0021] Define the user's interference vector as the row vector corresponding to the user in the interference matrix. Then, the interference vectors for user 1, user 2, and user N are as follows:

[0022] U1 = [0, W] ab (1, 2), ..., W ab (1, N)] (3)

[0023] U2 = [W ab (2, 1), 0, ..., W ab (2, N)] (4)

[0024] U N =[W ab (N, 1), W ab [(N, 2), ..., 0] (5)

[0025] Furthermore, the calculation and determination of the relative interference vector values ​​between each branch user and multiple root node users are as follows:

[0026] The branch user i, where i∈N, i≠a and i≠b, has a relative interference vector value d with the root node user. a，i The calculation formulas are as follows:

[0027]

[0028] Among them, U a U is the interference vector for root node user a; i Let i be the interference vector for branch user i;

[0029] U a，j This is the value at the j-th position in the root node user a;

[0030] U i，j Let be the value at the j-th position in branch user i.

[0031] Furthermore, the formula for calculating the mean interference vector of all users in the interference vector cluster is as follows:

[0032]

[0033] Where |C1| represents the number of all users in the interference vector cluster; U j Let be the interference vector for user j in the interference vector cluster; This refers to the sum of interference vectors for all users in the interference vector cluster;

[0034] The formula for calculating the relative interference vector value of the root node user in the interference vector cluster to the mean interference vector value of all users is as follows:

[0035] d a，mean =||U a -U mean ||2 (8).

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0037] This invention determines the root node based on the interference values ​​between users. It then makes a decision based on the relative interference vector values ​​between the root node and branch node users, assigning branch node users to the respective root nodes to obtain initial user clusters. Finally, it determines whether to update the root node interference vector based on the relative magnitude of the average user interference vector within the cluster and the root node's user interference vector. This algorithm uses user interference vectors that accurately describe the interference relationships between users and is also a low-complexity algorithm. It improves upon two shortcomings of existing algorithms, achieving the goal of increasing system throughput and spectrum utilization.

[0038] This invention effectively suppresses interference by analyzing the user's interference vector and the relative interference value between users, thereby reducing system interference and improving system performance. This invention only requires decision processing on the user's interference vector and the relative interference value, resulting in low complexity. Furthermore, the algorithm significantly improves system throughput and frequency utilization, making it easier to apply in practice. Attached Figure Description

[0039] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0040] Figure 1 This is a flowchart illustrating a user clustering method for ultra-dense networks according to the present invention.

[0041] Figure 2 This is a distribution diagram of base stations and users in an embodiment of a user clustering method for ultra-dense networks according to the present invention;

[0042] Figure 3 This is a comparative schematic diagram of two methods in an embodiment of a user clustering method for ultra-dense networks according to the present invention. Detailed Implementation

[0043] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. However, it should be understood that the scope of protection of the present invention is not limited to the specific implementation.

[0044] Example 1

[0045] like Figure 1 As shown, the present invention provides a user clustering method for ultra-dense networks, comprising the following steps:

[0046] Step S1: Calculate the interference between users in the ultra-dense network and obtain the interference vector value for each user;

[0047] Step S2: Select the users with larger user interference vector values ​​as root node users, and the remaining users as branch users;

[0048] Step S3: Calculate and determine the relative interference vector values ​​between each branch user and multiple root node users;

[0049] Step S4: When the relative interference vector value between a branch user and a root node user is less than the relative interference vector value between the branch user and other root node users, the branch user is classified into the interference vector cluster formed with the root node user, thus forming multiple interference vector clusters.

[0050] Step S5: Calculate and determine the relative magnitude of the interference vector value of the root node user in each interference vector cluster to the mean of the interference vector values ​​of all users in the interference vector cluster;

[0051] Step S6: When the relative interference vector value of the root node user in a certain interference vector cluster is zero or less than the average interference vector value of all users in the interference vector cluster, then all users in that interference vector cluster are grouped into one cluster.

[0052] Step S7: When the relative interference vector value of the root node user in a certain interference vector cluster is not zero or not less than the set threshold, the relative interference vector value of the mean interference vector of all users in the interference vector cluster is used as the interference vector of the root node user again, and the relative interference vector values ​​of each branch user and multiple root node users are recalculated and judged.

[0053] In step S1, the interference between users in the ultra-dense network is calculated to obtain the interference vector value for each user. This includes the following steps: Assuming the number of users is 1, 2, 3, ... N, the interference between any two users i and j is calculated using the cosine similarity of the interference channels between the two users, as follows:

[0054]

[0055] in, and These are user i in virtual community C. i and C j Channel vectors under the service;

[0056] and These are user j in virtual community C. i and C j Channel vectors under the service;

[0057] H H α represents the conjugate transpose of H; α and β represent coefficients for adjusting power and spatial dimension;

[0058] ||·|| F In this context, F = 2, which represents the L2 norm.

[0059] The interference matrix W is constructed by calculating the pairwise interference between N users. ab ,as follows:

[0060]

[0061] Define the user's interference vector as the row vector corresponding to the user in the interference matrix. Then, the interference vectors for user 1, user 2, and user N are as follows:

[0062] U1 = [0, W] ab (1, 2), ..., W ab (1, N)] (3)

[0063] U2 = [W ab (2, 1), 0, ..., W ab (2, N)] (4)

[0064] U N =[W ab (N, 1), W ab [(N, 2), ..., 0] (5)

[0065] In step S3, the relative interference vector values ​​between each branch user and multiple root node users are calculated and determined as follows:

[0066] Branch user i, where i∈N, i≠a and i≠b, has a relative interference vector value d with the root node user. a，i The calculation formulas are as follows:

[0067]

[0068] Among them, U a The interference vector for the root node user; U i Let i be the interference vector for branch user i;

[0069] U a，j This is the value at the j-th position in the root node user a;

[0070] U i，j Let be the value at the j-th position in branch user i.

[0071] In step S5, the formula for calculating the mean of the interference vectors of all users in the interference vector cluster is:

[0072]

[0073] Where |C1| represents the number of all users in the interference vector cluster; U j Let ∑ be the interference vector of user j in the interference vector cluster; j∈C1 U j This refers to the sum of interference vectors for all users in the interference vector cluster;

[0074] The formula for calculating the relative interference vector value of the root node user in the interference vector cluster to the mean interference vector value of all users is as follows:

[0075] d a，mean =||U a -U mean ||2 (8).

[0076] The following describes the specific implementation methods in conjunction with specific examples.

[0077] 1. A simulation system was used to simulate a scenario involving 6 base stations and 36 users. Figure 2 It is a map showing the distribution of base stations and users;

[0078] 2. Calculate the interference value between all users in this scenario according to equation (1) to obtain the interference vector value of each user.

[0079] 3. Taking a binary cluster as an example, the two users with larger interference vector values ​​among the 36 users are taken as root node users, namely the first root node user and the second root node user.

[0080] If the relative vector value of a branch user to the first root node user is less than its relative vector value to the first root node user, then the branch user is classified into the first interference vector cluster under the first root node user; if the relative vector value of a branch user to the second root node user is less than its relative vector value to the second root node user, then the branch user is classified into the second interference vector cluster under the second root node user.

[0081] Calculate and determine the relative interference vector values ​​of the root node users in the first and second interference vector clusters to the mean interference vector value of all users in the interference vector cluster. When the relative interference vector value of the root node users in the first and second interference vector clusters to the mean interference vector value of all users in the interference vector cluster is zero or less than a set threshold, the user clustering result is obtained, that is, the users are divided into two clusters, namely the users contained in the first interference vector cluster and the users contained in the second interference vector cluster.

[0082] The user clustering method of the present invention yields a binary clustering result for 36 users.

[0083] A comparison algorithm is introduced, namely the existing graph theory-based user clustering method, to perform binary clustering on 36 users using the same system parameters and simulation conditions.

[0084] 4. Using system throughput as a performance indicator, the comparison results are as follows: Figure 3 As shown. For central users (corresponding to the 85%–90% position of the CDF curve), the method of this invention has higher system throughput compared to the comparison algorithm; especially for edge users (corresponding to the 10% position of the CDF curve), the advantage of the method of this invention in improving system throughput is even more obvious. According to Figure 3 As can be seen from the content, the user clustering method in this invention has a significant advantage in improving system throughput performance.

[0085] This invention determines the root node based on the interference values ​​between users. It then makes a decision based on the relative interference vector values ​​between the root node and branch node users, assigning branch node users to the respective root nodes to obtain initial user clusters. Finally, it determines whether to update the root node interference vector based on the relative magnitude of the average user interference vector within the cluster and the root node's user interference vector. This algorithm uses user interference vectors that accurately describe the interference relationships between users and is a low-complexity algorithm, making it easier to apply in practice. It improves upon both shortcomings of existing algorithms, achieving the goal of increasing system throughput and spectrum utilization. The performance improvement is particularly significant for users located at the cell edge.

[0086] Finally, it should be noted that the above-disclosed embodiment is only one specific embodiment of the present invention. However, the embodiments of the present invention are not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A user clustering method for ultra-dense networks, characterized in that, Includes the following steps: Calculate the interference between users in an ultra-dense network and obtain the interference vector value for each user; Users with larger user interference vector values ​​are designated as root node users, and the remaining users are designated as branch users. Calculate and determine the relative interference vector values ​​between each branch user and multiple root node users; When the relative interference vector value between a branch user and a root node user is less than the relative interference vector value between the branch user and other root node users, the branch user is classified into the interference vector cluster formed with the root node user, thus forming multiple interference vector clusters. Calculate and determine the relative magnitude of the interference vector value of the root node user in each interference vector cluster to the mean of the interference vectors of all users in the interference vector cluster; If the relative interference vector value of the root node user in a certain interference vector cluster is zero or less than a set threshold compared to the mean interference vector value of all users in the interference vector cluster, then all users in that interference vector cluster are grouped into one cluster.

2. The user clustering method for ultra-dense networks according to claim 1, characterized in that: If the relative interference vector value of the root node user in a certain interference vector cluster is not zero or not less than a set threshold compared with the mean interference vector value of all users in the interference vector cluster, then the mean interference vector value of all users in each interference vector cluster is used again as the interference vector of the root node user, and the relative interference vector values ​​of each branch user and multiple root node users are recalculated and judged.

3. The user clustering method for ultra-dense networks according to claim 1, characterized in that: The calculation of interference between users in an ultra-dense network to obtain the interference vector value for each user includes the following steps: Let the number of users be 1, 2, 3, ... N The cosine similarity between any two users is calculated using the interference channel between them. i and users j The interference between them is as follows: （1） in, and They are users In virtual communities and Channel vectors under the service; and They are users j In virtual communities and Channel vectors under the service; express H The conjugate transpose of; and This represents a coefficient that adjusts power and spatial dimension; In this context, F=2 represents the L2 norm; Calculate separately N The interference between each pair of users constitutes an interference matrix. W ab ,as follows: （2） Define the user's interference vector as the row vector corresponding to the user in the interference matrix. Then, the interference vectors for user 1, user 2, and user N are as follows: （3） （4） （5）。 4. The user clustering method for ultra-dense networks according to claim 3, characterized in that: The relative interference vector values ​​between each branch user and multiple root node users are calculated and determined as follows: For branch users i ,in 'b' is the root user, and it is related to the root user. Relative interference vector value The calculation formulas are as follows: （6） in, For root node users a Interference vector; For branch users i Interference vector; For root node users a In the middle of j The value of each position; For branch users i In the middle of j The value at each position.

5. The user clustering method for ultra-dense networks according to claim 4, characterized in that: The formula for calculating the mean interference vector of all users in the interference vector cluster is: （7） in, The number of all users in the interference vector cluster; For users in the interference vector cluster j Interference vector; This refers to the sum of interference vectors for all users in the interference vector cluster; The formula for calculating the relative interference vector value of the root node user in the interference vector cluster to the mean interference vector value of all users is as follows: （8）。

Images