5G short message making method and system and 5G short message system

Abstract

The invention relates to the technical field of communication, and discloses a 5G short message making method and system and a 5G short message system.The making method comprises the following steps that observation information is collected, the observation information comprises state information of a 5G base station and information of a target user, the state information of the 5G base station comprises an available 5G base station and a damaged 5G base station, and the state information of the target user comprises the available 5G base station and the damaged 5G base station; the information of the target users comprises the number of the users and the position distribution of the users; graph structure data is constructed, the graph structure data comprises nodes and edges connected with the nodes, and one node represents one target user or the 5G base station. By collecting observation information such as 5G base station state and user distribution, end-to-end resource allocation learning is carried out based on a graph neural network model, a refined resource scheduling strategy is automatically generated according to actual states of a network and users, and the system can utilize residual 5G infrastructure resources to the maximum extent through intelligent scheduling, so that the resource scheduling efficiency is improved. And the availability and the coverage range of the network in a disaster environment are improved to the maximum extent.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and more specifically, to a 5G SMS production method and system, and a 5G SMS system. Background Art

[0002] Natural disasters (such as earthquakes, floods, typhoons, etc.) can damage traditional communication infrastructure and affect communication services. During disasters, it is crucial to ensure that emergency information (such as rescue information, safety tips, etc.) can be delivered to target users quickly and reliably. When part of the infrastructure is damaged, it is a major challenge to reasonably allocate the remaining resources to ensure the smooth flow of critical communications.

[0003] Natural disasters cause damage to traditional 5G communication infrastructure. Some 5G communication infrastructure may still be used, but its remote control link may be interrupted, and it is impossible to ensure that key information can be delivered to target users in a timely and accurate manner during natural disasters. Summary of the invention

[0004] The present invention provides a 5G SMS production method, system and 5G SMS system to solve the technical problem in the related art that the interruption of the remote control link in the 5G communication infrastructure will cause obstruction in the communication transmission process.

[0005] The present invention provides a 5G SMS production method, comprising the following steps:

[0006] S100: collecting observation information, where the observation information includes status information of 5G base stations and information of target users. The status information of 5G base stations includes available 5G base stations and damaged 5G base stations. The information of target users includes the number of users and location distribution of users.

[0007] S200: constructing graph structure data, where the graph structure data includes nodes and edges connecting the nodes, and one node represents a target user or a 5G base station;

[0008] There is an edge between the node representing a 5G base station and the node of an adjacent 5G base station. An adjacent 5G base station indicates a 5G base station that can communicate with each other through a short-distance communication link. There is an edge between a target user and an adjacent target user. An adjacent target user indicates a user who needs to send demand information through the target user. There is an edge between the node representing a 5G base station and all nodes representing target users. There is an edge between the nodes corresponding to the target users of the base station link connected to the same 5G base station. The node characteristics of the node representing a 5G base station are equal to the 5G base station characteristics, and the node characteristics of the node representing the target user are equal to the target user characteristics.

[0009] The observation information of each node is encoded into sequence data, which includes n sequence units. The nth sequence unit represents the observation information at the current moment, and the nsth sequence unit represents the observation information at the sth moment before the current moment.

[0010] S310, each 5G base station is configured with a total model, the total model includes a linear layer, a first hidden layer, a second hidden layer and an output layer, the sequence data are respectively input into the first hidden layer of the total model, and the first hidden state of each node is respectively output;

[0011] The graph structure data and the first hidden state of the node are input into the second hidden layer of the overall model, and the second hidden layer outputs the second hidden state; the output layer of the overall model inputs the second hidden state and outputs the initial local strategy;

[0012] S320, receiving an initial local strategy output by a model of an adjacent 5G base station through a short-range communication link;

[0013] S330, concatenating the sequence data with the initial local strategy received at the corresponding moment and inputting them into the linear layer of the overall model respectively, the linear layer of the overall model outputs sequence features, the sequence features are respectively input into the first hidden layer of the overall model, and the first hidden state of each node is respectively output;

[0014] Input the graph structure data and the first hidden state of the node into the second hidden layer of the model, and the second hidden layer outputs the second hidden state;

[0015] The output layer of the overall model inputs the second hidden state and outputs the executed local policy;

[0016] S340, parse the local strategy executed by the overall model output to obtain a 5G base station.

[0017] Furthermore, the calculation formula of the first hidden layer is as follows:

[0018] u (t) =σ(W u X (t) +W u H (t-1) +b u )

[0019] r (t) =σ(W r X (t) +W r H (t-1) +b r )

[0020] C (t) =tanh(W c X (t) +W c r(t) ⊙H (t-1) +b c )

[0021] H (t) =(1-u (t) )⊙C (t) +u (t) ⊙H (t-1)

[0022] Among them, W u , W r , W c represents the weight parameter, b u 、b r 、b c represents the bias parameter, ⊙ represents the dot product, where X (t) represents the tth sequence unit of the sequence data, H (t) and H (t-1) They represent the t-th and t-1-th hidden states respectively. The n-th sequence hidden state is used as the node feature of the node. n ≥ t ≥ 1, where n represents the total number of sequence units of the sequence data. When t = 1, H (t) =X (t) , tanh is the hyperbolic tangent function, and σ represents the S-type function.

[0023] Furthermore, the calculation formula of the second hidden layer is as follows:

[0024]

[0025] where h i represents the hidden state of the node of the i-th node, represents the aggregation coefficient of the i-th node, K i represents the set of nodes that have edges with the i-th node, σ represents the activation function, and W h represents the state weight coefficient, x j Represents the node features of the jth node;

[0026]

[0027] Z i =W z x i

[0028] Z j =W z x j

[0029] where x i and x j Represent the node features of the i-th and j-th nodes respectively, W 1 represents the aggregation weight coefficient, represents the splicing weight coefficient, T represents transpose, exp represents the exponential function with natural constant as the base, LeakyRelu represents the modified linear unit function, K i Represents the set of nodes that have edges with the i-th node.

[0030] Furthermore, the calculation formula of the output layer is as follows:

[0031]

[0032] Where y represents the control vector of the base station represented by the vth node. A component of the control vector represents the probability value of an action. The set of all actions is called the action space, which contains all possible actions that may be selected. v represents the hidden state of the vth node, h v represents the hidden state of the vth node, ∑ v∈N h v represents the concatenation of the hidden states of all nodes, N represents the set of nodes related to the recognition results, and W y is the weight parameter, b y is the bias parameter and σ represents the S-type function.

[0033] Furthermore, the local strategy is represented as a matrix, and the value of the element in the i-th row and j-th column of the matrix represents the time slot resource allocated by the i-th base station to the j-th user. If the value is 0, it means that the i-th base station has no connection with the j-th user.

[0034] Further, in step 101, a target network is constructed, and the target network replicates the overall model;

[0035] Step 102, randomly obtain an experience from the experience pool, the experience includes the state S at time t t , action A executed at time t t , the state S at time t+1 t+1 , perform action A t Rewards received R t+1 ;

[0036] The states at time t and t+1 represent the sequence data and graph structure data at the tth and t+1th time of the total model input respectively;

[0037] Then calculate the state S t The output of the target network corresponds to A t The probability value

[0038]

[0039] Indicates S t+1The maximum probability value of the target network output;

[0040] Step 103, according to and Calculate the loss value loss and update the overall model;

[0041]

[0042] Indicates S t The following total model output corresponds to A t The probability value of

[0043] γ is the discount factor, a value between 0 and 1;

[0044] Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current overall model;

[0045] Step 105, until the number of collected experiences reaches a set threshold or the overall model converges, the step is terminated.

[0046] Furthermore, the actions in the action space are obtained by random combinations of all possible values ​​of the control parameters;

[0047] For time slot resources with continuous values, the mean is discretized into point values, and then combined with the optional values ​​of other time slot resources to obtain the action.

[0048] Furthermore, the reward R t+1 The calculation formula is as follows:

[0049]

[0050] Where, ΔT t represents the number of bits transmitted from time t to time t+1, Δt represents the time interval between time t and time t+1, It is a throughput indicator, measured in bits per second (bit / s), representing the data rate successfully transmitted between the base station and the user per unit time. t Represents the transmission delay indicator, S (T，L) is the target user’s satisfaction value, w 1 , w 2 , w 3 Both are weighted systems;

[0051] Satisfaction value S (T，L) The value range of is between [0, 1], and the calculation formula is as follows:

[0052]

[0053] Among them, σ is an S-type function that can compress the input to the interval [0, 1]; T ef and L ef are the expected throughput and latency reference values ​​respectively.

[0054] A 5G SMS production system executes the steps in the aforementioned 5G SMS production method, including:

[0055] Observation information collection module: responsible for collecting 5G base station status information and target user information;

[0056] Graph structure building module: builds a graph structure based on observation information, where nodes represent base stations or users, and edges represent the connection relationship between base stations / users;

[0057] Feature encoding module: encodes the observation information of each node into time series data;

[0058] Graph neural network model: includes linear layer, hidden layer and output layer, with sequence data and graph structure as input and initial local resource allocation strategy as output;

[0059] Local strategy integration module: receives the initial strategy of the neighboring base station through the short-distance link, combines its own observations with the initial strategy, and outputs the executed local strategy;

[0060] Execution strategy analysis module: Analyze the final output local strategy to obtain the specific resource allocation plan for each base station to each user.

[0061] A storage medium storing non-temporary computer-readable instructions for executing one or more steps in the aforementioned 5G SMS production method.

[0062] The beneficial effects of the present invention are:

[0063] The present invention collects observation information such as 5G base station status and user distribution, models the network topology as graph structure data, and performs end-to-end resource allocation learning based on the graph neural network model. It can automatically generate refined resource scheduling strategies according to the actual status of the network and users. Through intelligent scheduling, the system can maximize the use of remaining 5G infrastructure resources, safeguard critical communications, and maximize the availability and coverage of the network in disaster environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a flow chart of a 5G SMS production method proposed by the present invention;

[0065] Figure 2 It is a structural block diagram of a 5G SMS production system proposed by the present invention;

[0066] Figure 3 It is the architectural block diagram of the overall model proposed in the present invention.

[0067] In the figure: 101, observation information collection module; 102, graph structure construction module; 103, feature encoding module; 104, graph neural network model; 105, local strategy integration module; 106, execution strategy analysis module. DETAILED DESCRIPTION

[0068] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Various examples may omit, replace or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0069] At least one embodiment of the present invention discloses a 5G SMS production method, such as Figure 1 As shown, including:

[0070] S100: collecting observation information, where the observation information includes status information of 5G base stations and information of target users. The status information of 5G base stations includes available 5G base stations and damaged 5G base stations. The information of target users includes the number of users and location distribution of users.

[0071] Collect observation information once at the same time interval;

[0072] S200: constructing graph structure data, where the graph structure data includes nodes and edges connecting the nodes, and one node represents a target user or a 5G base station;

[0073] There is an edge between the node of the 5G base station and the node of the adjacent 5G base station, and the adjacent 5G base station refers to the 5G base stations that can communicate with each other through a short-distance communication link (D2D);

[0074] Indicates that there is an edge between the target user and the adjacent target user. The adjacent target user represents the user who needs to send demand information through the target user.

[0075] There are edges between the node representing the 5G base station and all the nodes representing the target users, and there are edges between the nodes corresponding to the target users of the base station link connected to the same 5G base station;

[0076] The node characteristics of the node representing the 5G base station are equal to the 5G base station characteristics, and the node characteristics of the node representing the target user are equal to the target user characteristics;

[0077] The observation information of each node is encoded into sequence data, which includes n sequence units. The nth sequence unit represents the observation information at the current moment, and the nsth sequence unit represents the observation information at the sth moment before the current moment.

[0078] S310, the overall model includes a linear layer, a first hidden layer, a second hidden layer and an output layer, the sequence data are respectively input into the first hidden layer of the overall model, and the first hidden state of each node is respectively output;

[0079] Inputting the graph structure data and the first hidden state of the node into the second hidden layer of the overall model, and the second hidden layer outputs the second hidden state;

[0080] The output layer of the overall model inputs the second hidden state and outputs the initial local policy;

[0081] Each 5G base station is configured with a master model;

[0082] S320, receiving an initial local strategy output by a model of an adjacent 5G base station through D2D;

[0083] S330, concatenating the sequence data with the initial local strategy received at the corresponding moment and inputting them into the linear layer of the overall model respectively, the linear layer of the overall model outputs sequence features, the sequence features are respectively input into the first hidden layer of the overall model, and the first hidden state of each node is respectively output;

[0084] Input the graph structure data and the first hidden state of the node into the second hidden layer of the model, and the second hidden layer outputs the second hidden state;

[0085] The output layer of the overall model inputs the second hidden state and outputs the executed local policy;

[0086] S340, parsing the executed local strategy output by the total model to obtain a 5G base station;

[0087] Based on the local decision made after analysis, the 5G SMS publisher transmits key information to the target user through the corresponding base station. The key information includes rescue information, safety tips, etc.

[0088] In one embodiment of the present invention, the calculation formula of the first hidden layer is as follows:

[0089] u (t) =σ(W u X (t) +W u H (t-1) +b u )

[0090] r (t) =σ(W r X (t) +Wr H (t-1) +b r )

[0091] C (t) =tanh(W c X (t) +W c r (t) ⊙H (t-1) +b c )

[0092] H (t) =(1-u (t) )⊙C (t) +u (t) ⊙H (t-1)

[0093] Among them, W u , W r , W c represents the weight parameter, b u 、b r 、b c represents the bias parameter, ⊙ represents the dot product, where X (t) represents the tth sequence unit of the sequence data, H (t) and H (t-1) They represent the t-th and t-1-th hidden states respectively. The n-th sequence hidden state is used as the node feature of the node. n ≥ t ≥ 1, where n represents the total number of sequence units of the sequence data. When t = 1, H (t) =X (t) , tanh is the hyperbolic tangent function, σ represents the S-type function;

[0094] In one embodiment of the present invention, the sequence units are encoded into features through feature engineering before being input into the second hidden layer.

[0095] In one embodiment of the present invention, the calculation formula of the second hidden layer is as follows:

[0096]

[0097] where h i represents the hidden state of the node of the i-th node, represents the aggregation coefficient of the i-th node, K i represents the set of nodes that have edges with the i-th node, σ represents the activation function, and W h represents the state weight coefficient, x j Represents the node features of the jth node;

[0098]

[0099] Z i =Wz x i

[0100] Z j =W z x j

[0101] where x i and x j Represent the node features of the i-th and j-th nodes respectively, W 1 represents the aggregation weight coefficient, represents the splicing weight coefficient, T represents transpose, exp represents the exponential function with natural constant as the base, LeakyRelu represents the modified linear unit function, K i Represents the set of nodes that have edges with the i-th node;

[0102] In one embodiment of the present invention, the calculation formula of the output layer is as follows:

[0103]

[0104] Where y represents the control vector of the base station represented by the vth node. A component of the control vector represents the probability value of an action. The set of all actions is called the action space, which contains all possible actions that may be selected. v represents the hidden state of the vth node, h v represents the hidden state of the vth node, Σ v∈N h v represents the concatenation of the hidden states of all nodes, N represents the set of nodes related to the recognition results, and W y is the weight parameter, b y is the bias parameter, σ represents the S-type function;

[0105] In one embodiment of the present invention, the local strategy is represented as a matrix, and the value of the element in the i-th row and j-th column of the matrix represents the time slot resources allocated by the i-th base station to the j-th user. If the value is 0, it means that the i-th base station has no connection with the j-th user.

[0106] Among them, the reasonable allocation of time slot resources directly determines the data transmission time and throughput between the base station and the user. The allocation of time slot resources is directly related to key performance indicators such as latency and throughput, and is the main reference for evaluating the efficiency of wireless resource utilization;

[0107] In one embodiment of the present invention, the actions in the action space are obtained by randomly combining possible values ​​of all control parameters.

[0108] For time slot resources with continuous values, the mean is discretized into point values, and then combined with the optional values ​​of other time slot resources to obtain the action.

[0109] In one embodiment of the present invention, the step of training the overall model includes:

[0110] Step 101, constructing a target network, the target network replicates the overall model;

[0111] Step 102, randomly obtain an experience from the experience pool, the experience includes the state S at time t t , action A executed at time t t , the state S at time t+1 t+1 , perform action A t Rewards received R t+1 ;

[0112] The states at time t and t+1 represent the sequence data and graph structure data at the tth and t+1th time of the total model input respectively;

[0113] Then calculate the state S t The output of the target network corresponds to A t The probability value

[0114]

[0115] Indicates S t+1 The maximum probability value of the target network output;

[0116] Step 103, according to and Calculate the loss value loss and update the overall model;

[0117]

[0118] Indicates S t The following total model output corresponds to A t The probability value of

[0119] γ is the discount factor, a value between 0 and 1;

[0120] Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current overall model;

[0121] Step 105, until the number of collected experiences reaches a set threshold or the overall model converges, the step is terminated.

[0122] In one embodiment of the present invention, the experience is derived from the record of historical action execution.

[0123] In one embodiment of the present invention, the reward R t+1 The calculation formula is as follows:

[0124]

[0125] Where, ΔT t represents the number of bits transmitted from time t to time t+1, Δt represents the time interval between time t and time t+1, It is a throughput indicator, measured in bits per second (bit / s), representing the data rate successfully transmitted between the base station and the user per unit time. t Represents the transmission delay indicator, S (T，L) is the target user’s satisfaction value, w 1 , w 2 , w 3 Both are weighted systems;

[0126] Satisfaction value S (T，L) The value range of is between [0, 1], and the calculation formula is as follows:

[0127]

[0128] Among them, σ is an S-type function that can compress the input to the interval [0, 1]; T ef and L ef are the expected throughput and latency reference values ​​respectively.

[0129] A 5G SMS production system, comprising:

[0130] Observation information collection module: responsible for collecting 5G base station status information (available / damaged) and target user information (quantity, location distribution);

[0131] Graph structure building module: builds a graph structure based on observation information, where nodes represent base stations or users, and edges represent the connection relationship between base stations / users;

[0132] Feature encoding module: encodes the observation information of each node (base station / user) into time series data;

[0133] Graph neural network model: includes linear layer, hidden layer and output layer, with sequence data and graph structure as input and initial local resource allocation strategy as output;

[0134] Local strategy integration module: receives the initial strategy of the neighboring base station through the short-distance link, combines its own observations with the initial strategy, and outputs the executed local strategy;

[0135] Execution strategy analysis module: Analyze the final output local strategy to obtain the specific resource allocation plan for each base station to each user.

[0136] According to the above-mentioned 5G SMS production method and system, the 5G SMS system adapted thereto includes the following modules:

[0137] User interface module: provides a user-friendly SMS editing interface, supports multimedia content (pictures, videos, voice, etc.) insertion, input of recipient numbers or selection of contacts in the address book;

[0138] Encoding module: encodes the multimedia text message content edited by the user, performs format conversion, compression and other processing;

[0139] 5G access module: Sends encoded SMS data through the 5G wireless access network, providing high-bandwidth and low-latency transmission based on 5G NR technology;

[0140] 5G core network module: IMS subsystem, processing multimedia signaling and session control;

[0141] SMS service center module: temporarily store and forward multimedia SMS;

[0142] Receiving module: The target 5G terminal receives and decodes multimedia SMS data.

[0143] At least one embodiment of the present disclosure provides a storage medium storing non-temporary computer-readable instructions for executing one or more steps in the aforementioned 5G SMS production method.

[0144] The computer program may be stored / distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope.

[0145] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A 5G SMS production method, characterized in that: The following steps are involved: S100: collecting observation information, where the observation information includes status information of 5G base stations and information of target users. The status information of 5G base stations includes available 5G base stations and damaged 5G base stations. The information of target users includes the number of users and location distribution of users. S200: constructing graph structure data, where the graph structure data includes nodes and edges connecting the nodes, and one node represents a target user or a 5G base station; There is an edge between the node representing a 5G base station and the node of an adjacent 5G base station. An adjacent 5G base station indicates a 5G base station that can communicate with each other through a short-distance communication link. There is an edge between a target user and an adjacent target user. An adjacent target user indicates a user who needs to send demand information through the target user. There is an edge between the node representing a 5G base station and all nodes representing target users. There is an edge between the nodes corresponding to the target users of the base station link connected to the same 5G base station. The node characteristics of the node representing a 5G base station are equal to the 5G base station characteristics, and the node characteristics of the node representing the target user are equal to the target user characteristics. The observation information of each node is encoded into sequence data, which includes n sequence units. The nth sequence unit represents the observation information at the current moment, and the nsth sequence unit represents the observation information at the sth moment before the current moment. S310, each 5G base station is configured with a total model, the total model includes a linear layer, a first hidden layer, a second hidden layer and an output layer, the sequence data are respectively input into the first hidden layer of the total model, and the first hidden state of each node is respectively output; The graph structure data and the first hidden state of the node are input into the second hidden layer of the overall model, and the second hidden layer outputs the second hidden state; the output layer of the overall model inputs the second hidden state and outputs the initial local strategy; S320, receiving an initial local strategy output by a model of an adjacent 5G base station through a short-range communication link; S330, concatenating the sequence data with the initial local strategy received at the corresponding moment and inputting them into the linear layer of the overall model respectively, the linear layer of the overall model outputs sequence features, the sequence features are respectively input into the first hidden layer of the overall model, and the first hidden state of each node is respectively output; Inputting the graph structure data and the first hidden state of the node into the second hidden layer of the model, and the second hidden layer outputs the second hidden state; The output layer of the overall model inputs the second hidden state and outputs the executed local policy; S340, parse the local strategy executed by the overall model output to obtain a 5G base station.

2. A 5G SMS production method according to claim 1, characterized in that: The calculation formula for the first hidden layer is as follows: u (t) =σ(W u X (t) +W u H (t-1) +b u ) r (t) =σ(W r X (t) +W r H (t-1) +b r ) C (t) =tanh(W c X (t) +W c r (t) ⊙H (t-1) +b c ) H (t) =(1-u (t) )⊙C (t) +u (t) ⊙H (t-1) Among them, W u , W r , W c represents the weight parameter, b u 、b r 、b c represents the bias parameter, ⊙ represents the dot product, where X (t) represents the tth sequence unit of the sequence data, H (t) and H (t-1) They represent the t-th and t-1-th hidden states respectively. The n-th sequence hidden state is used as the node feature of the node. n ≥ t ≥ 1, where n represents the total number of sequence units of the sequence data. When t = 1, H (t) =X (t) , tanh is the hyperbolic tangent function, and σ represents the S-type function.

3. A 5G SMS making method according to claim 1, characterized in that: The calculation formula for the second hidden layer is as follows: where h i represents the hidden state of the node of the i-th node, represents the aggregation coefficient of the i-th node, K i represents the set of nodes that have edges with the i-th node, σ represents the activation function, and W h represents the state weight coefficient, x j Represents the node features of the jth node; Z i =W z x i Z j =W z x j where x i and x j Represent the node features of the i-th and j-th nodes respectively, W1 represents the aggregation weight coefficient, represents the splicing weight coefficient, T represents transpose, exp represents the exponential function with natural constant as the base, LeakyRelu represents the modified linear unit function, K i Represents the set of nodes that have edges with the i-th node.

4. A 5G SMS making method according to claim 1, characterized in that: The calculation formula of the output layer is as follows: Where y represents the control vector of the base station represented by the vth node. A component of the control vector represents the probability value of an action. The set of all actions is called the action space, which contains all possible actions that may be selected. v represents the hidden state of the vth node, h v represents the hidden state of the vth node, ∑ v∈N h v represents the concatenation of the hidden states of all nodes, N represents the set of nodes related to the recognition results, and W y is the weight parameter, b y is the bias parameter and σ represents the S-type function.

5. A 5G SMS making method according to claim 1, characterized in that: The local strategy is represented as a matrix. The value of the element in the i-th row and j-th column of the matrix represents the time slot resource allocated by the i-th base station to the j-th user. If the value is 0, it means that the i-th base station has no connection with the j-th user.

6. A 5G SMS making method according to claim 1, characterized in that: Step 101, constructing a target network, the target network replicates the overall model; Step 102, randomly obtain an experience from the experience pool, the experience includes the state S at time t t , action A executed at time t t , the state S at time t+1 t+1 , perform action A t Rewards received R t+1 ; The states at time t and t+1 represent the sequence data and graph structure data at the tth and t+1th time of the total model input respectively; Then calculate the state S t The output of the target network corresponds to A t The probability value Indicates S t+1 The maximum probability value of the target network output; Step 103, according to and Calculate the loss value loss and update the overall model; Indicates S t The following total model output corresponds to A t The probability value of γ is the discount factor, a value between 0 and 1; Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current overall model; Step 105, until the number of collected experiences reaches a set threshold or the overall model converges, the step is terminated.

7. A 5G SMS making method according to claim 1, characterized in that: The actions in the action space are obtained by random combinations of all possible values ​​of the control parameters; For time slot resources with continuous values, the mean is discretized into point values, and then combined with the optional values ​​of other time slot resources to obtain the action.

8. A 5G SMS making method according to claim 1, characterized in that: Reward R t+1 The calculation formula is as follows: Where, ΔT t represents the number of bits transmitted from time t to time t+1, Δt represents the time interval between time t and time t+1, It is a throughput indicator, measured in bits per second (bit / s), representing the data rate successfully transmitted between the base station and the user per unit time. t Represents the transmission delay index, S (T，L) is the satisfaction value of the target user, w1, w2, w3 are all weight systems; Satisfaction value S (T，L) The value range of is between [0, 1], and the calculation formula is as follows: Among them, σ is an S-type function that can compress the input to the interval [0, 1]; T ef and L ef are the expected throughput and latency reference values ​​respectively.

9. A 5G SMS production system, executing the steps in a 5G SMS production method as described in any one of claims 1 to 8, characterized in that: include: Observation information collection module: responsible for collecting 5G base station status information and target user information; Graph structure building module: builds a graph structure based on observation information, where nodes represent base stations or users, and edges represent the connection relationship between base stations / users; Feature encoding module: encodes the observation information of each node into time series data; Graph neural network model: includes linear layer, hidden layer and output layer, with sequence data and graph structure as input and initial local resource allocation strategy as output; Local strategy integration module: receives the initial strategy of the neighboring base station through the short-distance link, combines its own observations with the initial strategy, and outputs the executed local strategy; Execution strategy analysis module: Analyze the final output local strategy to obtain the specific resource allocation plan for each base station to each user.

10. A storage medium, characterized in that: Non-temporary computer-readable instructions are stored for executing one or more steps in a 5G SMS production method as described in any one of claims 1-8.

Images