Ris mode selection method based on grey correlation clustering method

Abstract

The present application relates to the technical field of data analysis, in particular to a RIS mode selection method based on a grey correlation clustering method. It comprises the following steps: S1. A base station acquires communication data of a user, and divides the communication data of the user into two types of data: numerical data with continuous values and type data with discrete values; S2. For the type data with discrete values, a grey possibility degree mapping matrix is established to directly map the discrete variables to the final mode classification; S3. For the numerical data with continuous values, a grey possibility degree function is established to first map the continuous variables to connotation classification, and then to the final mode classification; S4. A weight is set to statistically summarize the prediction scores of each mode; and S5. The optimal decision for the RIS mode is selected according to the ranking of the prediction scores. This method can accurately capture user data while reducing the complexity of calculation, and is more suitable for real-time application under a large model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, in particular to a RIS mode selection method based on a grey correlation clustering method. BACKGROUND

[0002] Reconfigurable Intelligent Surface (RIS) is an emerging wireless communication technology, and its reflection mode is one of its core characteristics. RIS is usually composed of a large number of programmable reflection units, which can dynamically adjust their phase and amplitude to achieve intelligent reconstruction of the propagation channel. When wireless signals reach the RIS surface, these reflection units will reflect, refract or scatter signals by adjusting the phase and amplitude according to pre-set algorithms or real-time feedback, in order to optimize the signal propagation environment.

[0003] In terms of application, the reflection mode of RIS has great potential in multiple fields: in complex indoor environments (such as office buildings, shopping malls, etc.), RIS can improve signal coverage and reduce blind areas. In large outdoor environments (such as city streets, stadiums, etc.), RIS can optimize signal propagation paths and enhance signal strength and stability. In future 5G and 6G networks, RIS can be used for beamforming, interference management and spectrum efficiency improvement to meet higher bandwidth and connectivity requirements. RIS can also provide more stable and efficient communication environments for Internet of Things devices, especially in device-dense environments such as smart cities and industrial automation.

[0004] Most current RIS mode selection techniques are based on engineers' experience and pre-set rules or use simple data statistics and machine learning methods to analyze user data to select modes. Simple mode selection methods based on experience generally lack flexibility and are difficult to adapt to rapidly changing environments and diverse user needs. Moreover, their simple models often fail to capture complex network characteristics, greatly limiting their accuracy. For methods that use machine learning algorithms for mode selection, the computational complexity is high, and it is difficult to achieve real-time application in large models. SUMMARY

[0005] The present application provides a RIS mode selection method based on a grey correlation clustering method, aiming to solve the technical problems existing in most current RIS mode selection techniques.

[0006] The present application provides a RIS mode selection method based on a grey correlation clustering method, comprising the following steps:

[0007] S1. The base station acquires the user's communication data and divides the user's communication data into two categories: continuous numerical data and discrete type data.

[0008] S2. For discrete value type data, establish a gray possibility mapping matrix, perform clustering of discrete variables, and directly map the discrete variables to the final pattern classification;

[0009] S3. For numerical data with continuous values, establish a gray probability function, perform clustering of continuous variables, and first map the continuous variables to the connotation classification, and then map them to the final pattern classification.

[0010] S4. Set weights and perform overall statistics on the prediction scores of various modes based on the weights;

[0011] S5. Select the optimal decision for the RIS mode based on the ranking of the predicted scores.

[0012] As a further improvement of the present invention, in step S1, the continuously valued numerical data includes distance values, SNR values, and angle values; the discrete valued data includes service priority values.

[0013] As a further improvement of the present invention, step S2 specifically includes:

[0014] Assume there is a total There are n numerical data types, and the r-th data type is defined. Continuously valued variable χ r Its connotation category Grey probability function for:

[0015]

[0016] in, For the gray range of the k-th class, for different variables χ r Number of categories Different, and each corresponding to different k of They are different; y represents the value of χ, and y∈[a,b] is the mathematical representation of y in the interval a~b. If y is not within this interval; when χ r When the value of χ is different, r The probability of choosing different categories also varies.

[0017] As a further improvement of the present invention, in step S3, a grey possibility function is established for the continuously valued numerical data, and clustering of continuous variables is performed, mapping the continuous variables to connotation classification first, specifically including:

[0018] Assume there is a total There are data types, and the d-th type is defined. Discrete variable χ d Category The gray possibility mapping matrix is:

[0019]

[0020] The matrix has N rows. column, i.e. There are N discrete values ​​that can be assigned to In each category; in the matrix, the kth (k=1,2,...,K)th row (n=1,...N) of the matrix. χ The meaning of the column element is that when χ d Pick At that time, X d The probability of belonging to the k-th class, where the element's value is less than 1. This represents the discrete-valued variable X. d Category The degree of grayness.

[0021] As a further improvement of the present invention, in step S3, after the continuous variable is mapped to the connotation classification, it is then mapped to the final pattern classification, specifically including:

[0022] Define the r-th Continuously valued variable X r The connotation category The two-step probability mapping matrix is:

[0023]

[0024] The size of the matrix is ​​K. A OK Column, K A For the final number of target categories, For variable χ r The number of connotation categories; the ξ-th (ξ=1,2,…,K) th element in the matrix. A ) line number The meaning of the column element is that when the variable χ r When a category belongs to the k-th connotation category, the probability of it belonging to the ξ-th final target category is less than 1. This represents the r-th one. Continuously valued variable X r The connotation category The two-step probability.

[0025] As a further improvement of the present invention, the specific process of setting the weights in step S4 includes:

[0026] Let there be numerical data, there are type data, for the rth continuous variable, define its weight as for the dth discrete variable, define its weight as ω d , the initial calculation formula is as follows:

[0027]

[0028] As a further improvement of the application, in the step S4, the prediction scores of each type of mode are statistically summarized by weight, specifically including:

[0029] For the mth (m = 1, …, K C ) type of mode, the score γ m predicted by the current user i to select it is calculated by the following formula:

[0030]

[0031] Wherein is the value of variable X d , and

[0032] As a further improvement of the application, the step S5 specifically includes:

[0033] If the total number of users is n, the final decision score is calculated by the following formula:

[0034]

[0035] The value of γ m is sorted in descending order, and the corresponding index value is This index sequence is the predicted RIS mode sequence of the most suitable communication demand at this time; wherein, The corresponding mode is the predicted optimal mode, and the RIS mode will be automatically set to the optimal mode.

[0036] The beneficial effects of the application are: the method can obtain and quantify user data by dividing the user data into continuous data and discrete data, realize clustering of continuous variables and discrete variables by using gray possibility function and mapping matrix, directly map discrete variables to final mode classification, first map continuous variables to connotation classification, then map them to final mode classification, and finally realize the decision of RIS mode by adaptively setting weights to statistically summarize the prediction scores of each type of mode. The method can accurately capture user data while reducing the complexity of calculation, and is more suitable for real-time application under large machine learning model. BRIEF DESCRIPTION OF DRAWINGS​

[0037] Figure 1 is a flow chart of the RIS mode selection method based on the grey correlation clustering method of the present application;

[0038] Figure 2 is a model schematic diagram of the present application for selecting the reflection mode of the reconfigurable intelligent surface by the grey correlation clustering method. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and examples.

[0040] As shown in Figure 1 , the present application provides a RIS mode selection method based on the grey correlation clustering method, RIS refers to a reconfigurable intelligent surface (Reconfigurable Intelligent Surface, RIS), and the RIS mode refers to the reflection mode of the reconfigurable intelligent surface, which specifically includes the following steps:

[0041] S1. The base station acquires the communication data of the user, and divides the communication data of the user into two types of data: numerical data with continuous values and type data with discrete values.

[0042] The base station can acquire the following communication data when transmitting data with the user: distance value, SNR value, angle value, service priority value, etc. Among them, the numerical data with continuous values includes distance value, SNR value, angle value, etc.; and the type data with discrete values includes service priority value, etc.

[0043] S2. For the type data with discrete values, a grey possibility mapping matrix (category setting corresponding to the mode of RIS) is established, the clustering of discrete variables is performed, and the discrete variables are directly mapped to the final mode classification.

[0044] Suppose there are numerical data, define the rth continuous variable X r The grey possibility function of the connotation category is:

[0045]

[0046] Among them, is the grey value interval of the kth category, which is artificially defined. For different variables χ r , the number of categories is different, and the of each ​Different; y is the value of χ, y ∈ [a, b] is the mathematical representation of y in the interval a ~ b, For y is not in this interval. The meaning of this function is that when χ r The value is different, the value of the same category is different.

[0047] S3. For continuous value numerical class data, establish a gray possibility function (category setting is not limited), cluster continuous variables, and map continuous variables to connotation classification first, and then to final pattern classification.

[0048] For example, the distance of continuous variable user from the base station is mapped to its corresponding connotation classification such as far, medium and near, and then finally set the weight to map to the final pattern classification. This corresponds to the reflection mode of RIS such as 0°, 45°, 90°, etc.

[0049] Suppose there are type data, define the dth discrete value variable χ d The gray possibility mapping matrix of category is:

[0050]

[0051] The size of this matrix is N rows columns, that is There are N discrete values, which can be attributed to K categories; the meaning of the nth (n = 1, … N) row and the kth (k = 1, 2, …, K χ column element in the matrix is that when χ d Take , χ d belongs to the kth category The value of these elements is required to be less than 1, which can be given artificially according to experience, The gray possibility of discrete value variable χ d to category is represented. If statistical method is used, the calculation formula is:

[0052]

[0053] For subsequent calculation, the total number of categories of all discrete value variables is taken as the total number of patterns K C .

[0054] Two-step clustering mapping of continuous variables This step helps to map the connotation classification of continuous variables to the final target category (pattern category). The above gray possibility function only maps the variable to its connotation classification, not the final pattern classification. To get the final result, it needs to be mapped again.

[0055] Define the r-th Continuously valued variable χ r The connotation category The two-step probability mapping matrix is:

[0056]

[0057] The size of the matrix is ​​K. A OK Column, K A For the final number of target categories, For variable χ r The number of connotation categories; the ξ-th (ξ=1,2,…,K) th element in the matrix. A ) line number The meaning of the column element is that when the variable χ r When a category belongs to the k-th connotation category, the probability of it belonging to the ξ-th final target category is given. The values ​​of these elements must be less than 1 and can be given manually based on experience. This represents the r-th one. Continuously valued variable χ r The connotation category The two-step probability. If a statistical method is used, the calculation formula is:

[0058]

[0059] During algorithm execution, another K A The value can be the total number of patterns, K. C .

[0060] S4. Set weights and perform overall statistics on the prediction scores of various modes based on the weights.

[0061] The specific process of setting weights includes: for the r-th... For continuous variables, define their weights as follows: For the dth Discrete variables, with their weights defined as ω. d The initial calculation formula is as follows:

[0062]

[0063] The grey clustering coefficient calculation step yields the predicted scores for each pattern selected by user i in the current data.

[0064] For the m-th (m=1,…,K) C ( ) type pattern, predict the score Υ of the current user i's selection. m Calculated by the following formula:

[0065]

[0066] in For variable χ d Values time

[0067] S5. Select the optimal decision for the RIS mode based on the ranking of the predicted scores.

[0068] Final decision execution: If the total number of users is n, the final decision score is calculated by the following formula:

[0069]

[0070] Again to Y m The values ​​are sorted in descending order, and the corresponding index value is This index sequence is the RIS pattern sorting predicted by the algorithm as the most suitable for the current communication needs; where, The corresponding mode is the predicted optimal mode, namely the "currently most suitable RIS mode". The RIS mode will be automatically set to this optimal mode so that the base station's reflection link can cover as many users as possible and users with high demand priority.

[0071] Most existing models use FPGA to control the RIS (Reflection System) to select the reflection mode, such as 90°, 0°, 45°, etc. For example... Figure 2 As shown, this invention selects the reflection mode of the reconfigurable smart surface (RIS) using a grey relational clustering method. After selection by the grey relational clustering algorithm, the blue mode in the figure is selected as the optimal RIS mode in the current state, thereby enabling the reflection link to cover as many users as possible, as well as users with high priority needs. The selection method of this invention is highly flexible and can adapt to rapidly changing environments and diverse user needs. While ensuring accurate capture of user data, it reduces computational complexity and is more suitable for real-time applications under large machine learning models.

[0072] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A RIS pattern selection method based on grey relational clustering, characterized in that, Includes the following steps: S1. The base station acquires the user's communication data and divides the user's communication data into two categories: continuous numerical data and discrete type data. S2. For discrete value type data, establish a gray possibility mapping matrix, perform clustering of discrete variables, and directly map the discrete variables to the final pattern classification; S3. For numerical data with continuous values, establish a gray probability function, perform clustering of continuous variables, and first map the continuous variables to the connotation classification, and then map them to the final pattern classification. S4. Set weights and perform overall statistics on the prediction scores of various modes based on the weights; S5. Select the optimal decision for the RIS mode based on the ranking of the predicted scores; Step S2 specifically includes: Assume there is a total Given a set of numerical data, define the r-th continuously valued variable χ. r Grey probability function for its intensional category k for: in, k = 1, 2, ..., Kχ r , For the gray range of the k-th class, for different variables χ r The number of categories Kχ r Different, and each corresponding to different k of They are different; y represents the value of χ, and y∈[a,b] is the mathematical representation of y in the interval a~b. If y is not within this interval; when χ r When the value of χ is different, r The probability of choosing different categories also varies; In step S3, a grey possibility function is established for the continuously valued numerical data, and clustering of continuous variables is performed, mapping the continuous variables to connotation classification. Specifically, this includes: Assume there is a total Given a data type, define the d-th discrete variable χ. d The gray possibility mapping matrix for category k is: in, k = 1, 2, ..., Kχ d ; The matrix has N rows and Kχ. d Column, i.e. There are N discrete values ​​that can be assigned to Kχ. d In each category; the meaning of the element in the nth row and kth column of the matrix is, where n = 1, ..., N, k = 1, 2, ..., K χ When χ d Pick When, χ d The probability of belonging to the k-th class, where the element's value is less than 1. This represents the discrete variable χ. d The gray likelihood of category k, where k = 1, 2, ..., Kχ d ; In step S3, after the continuous variable is mapped to the connotation classification, it is then mapped to the final pattern classification, specifically including: Define the r-th continuously valued variable χ r The two-step possibility degree mapping matrix of the connotation category k is: The size of the matrix is ​​K. A line Kχ r Column, K A Kχ represents the final number of target categories. r For variable χ r The number of connotation categories; the meaning of the element in the ξ-th row and k-th column of the matrix is, when the variable χ r When a category belongs to the k-th connotation category, the probability of it belonging to the ξ-th final target category is less than 1. This represents the r-th continuously variable χ. r The two-step probability of the connotation category k; where k = 1, 2, ..., Kχ r ξ=1,2,…,K A ; In step S4, the specific process of setting the weights includes: Assume there is a total Numerical data types, with Given data of type r, define the weight of the r-th continuous variable as follows: For the d-th discrete variable, its weight is defined as ω. d The initial calculation formula is as follows: in, In step S4, the prediction scores of various modes are statistically analyzed based on weights, specifically including: For the m-th pattern, predict the score Υ for the current user i when selecting it. m Calculated by the following formula: in For variable χ d Values time 2. The RIS pattern selection method based on grey relational clustering as described in claim 1, characterized in that, In step S1, the continuously measured numerical data includes distance values, SNR values, and angle values. Discrete value type data includes business priority values.

3. The RIS pattern selection method based on grey relational clustering as described in claim 1, characterized in that, Step S5 specifically includes: If the total number of users is n, the final decision score is calculated using the following formula: Again to Y m The values ​​are sorted in descending order, and the corresponding index value is... This index sequence is a sorting of RIS modes that best predict the current communication needs; where... The corresponding mode is the predicted optimal mode, and the RIS mode will be automatically set to this optimal mode.

Images