Multi-user selection method, apparatus, device, storage medium and program product

Abstract

The application discloses a multi-user selection method and device, equipment, storage medium and program product, constructs a weighted undirected graph according to acquired user information, and splits the weighted undirected graph according to a preset node upper limit value to obtain at least two subgraphs; then a quantum algorithm is used to calculate each subgraph to obtain at least one candidate solution corresponding to each subgraph; finally, a global undirected graph is constructed according to each subgraph and the corresponding candidate solution, and a global solution of multi-user selection is output according to the global undirected graph. By adopting the embodiment of the application, the solving accuracy in the multi-user selection process can be improved, so that resources can be reasonably allocated for users, and the communication quality is improved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication, and more particularly to a multi-user selection method, apparatus, device, storage medium, and program product. Background Technology

[0002] In wireless communication networks, multi-user selection refers to the process of effectively allocating network resources (such as spectrum and power) among multiple users according to certain rules and algorithms to achieve specific system performance requirements (such as ensuring channel quality, following user priorities, and suppressing interference). Traditional multi-user selection algorithms include greedy algorithms and genetic algorithms. These traditional algorithms perform reasonably well in small- to medium-sized networks, but their computational complexity and solution time increase dramatically when solving larger-scale user selection problems. For example, while the greedy algorithm is simple and fast, its results are often affected by the initial selection and are prone to getting stuck in suboptimal solutions. Genetic algorithms find the optimal solution by simulating natural selection and genetic processes, but in large-scale problems, their convergence speed and solution accuracy are affected by the population size and genetic operations, making it difficult to achieve ideal results. Therefore, traditional algorithms have low solution accuracy in multi-user selection processes, leading to unreasonable resource allocation and affecting communication quality. Summary of the Invention

[0003] The purpose of this invention is to provide a multi-user selection method, apparatus, device, storage medium, and program product that can improve the solution accuracy in the multi-user selection process, thereby enabling reasonable allocation of resources to users and improving communication quality.

[0004] To achieve the above objectives, embodiments of the present invention provide a multi-user selection method, including:

[0005] A weighted undirected graph is constructed based on the acquired user information; wherein each node in the weighted undirected graph represents a user, the weight value of the node represents the comprehensive weight of each user, and the edges between nodes represent the interference relationship between users;

[0006] The weighted undirected graph is divided according to a preset node upper limit value to obtain at least two subgraphs;

[0007] A quantum algorithm is used to compute each subgraph to obtain at least one candidate solution corresponding to each subgraph.

[0008] Construct a global undirected graph based on each subgraph and its corresponding candidate solutions, and output a global solution selected by multiple users based on the global undirected graph.

[0009] As an improvement to the above scheme, the user information includes downlink channel state information and user priority.

[0010] As an improvement to the above scheme, the step of constructing a weighted undirected graph based on the acquired user information includes:

[0011] The channel gain for each user and the interference coefficient between users are determined based on the downlink channel state information.

[0012] Calculate the independent weight for each user based on the channel gain and the user priority;

[0013] With the goal of maximizing the overall weight value of all users, an objective function is constructed based on the independent weights and the interference coefficients.

[0014] Construct a weighted undirected graph based on the objective function.

[0015] As an improvement to the above scheme, the weighted undirected graph is segmented according to a preset node upper limit value to obtain at least two subgraphs, including:

[0016] The initial community structure in the weighted undirected graph is identified using a community detection algorithm;

[0017] When the number of nodes in any initial community structure exceeds the node upper limit, the initial community structure is divided to obtain the corresponding sub-community structure;

[0018] By aggregating nodes across all community structures, a grid structure is obtained.

[0019] The mesh structure is divided according to the node upper limit and the preset optimization target to obtain at least two sub-graphs.

[0020] As an improvement to the above scheme, the node upper limit is determined based on the maximum subgraph size that the quantum computing system can process.

[0021] As an improvement to the above scheme, the step of using a quantum algorithm to calculate each subgraph to obtain at least one candidate solution corresponding to each subgraph includes:

[0022] Obtain the subgraph information corresponding to each subgraph; wherein, the subgraph information includes the comprehensive weight value of the subgraph and a two-dimensional array of inter-user interference signals;

[0023] The quantum algorithm is used to repeatedly calculate the information of each subgraph according to the preset number of repetitions to obtain the number of initial solutions corresponding to the number of repetitions.

[0024] From all initial solutions, obtain candidate solutions that satisfy the candidate conditions, and obtain at least one candidate solution corresponding to each subgraph.

[0025] As an improvement to the above scheme, candidate solutions that satisfy the candidate conditions are obtained from all initial solutions, including:

[0026] From all initial solutions, select the n1 target initial solutions with the highest probability of occurrence;

[0027] The current subgraph is weighted using the target initial solution to obtain the initial weight value corresponding to each target initial solution, where n1≥2;

[0028] Obtain the top n2 target initial weight values ​​from all initial weight values, and use the target initial solutions corresponding to the target initial weight values ​​as candidate solutions, where n2≥1.

[0029] As an improvement to the above scheme, the step of constructing a global undirected graph based on each subgraph and its corresponding candidate solutions includes:

[0030] An initial undirected graph is constructed based on each subgraph and its corresponding candidate solutions, and the initial comprehensive weight value solution of the initial undirected graph is calculated.

[0031] The target nodes that meet the node flipping conditions are flipped sequentially, and the initial undirected graph is updated based on the flipped nodes.

[0032] Calculate the reference composite weight solution for the updated initial undirected graph;

[0033] If the reference comprehensive weight value solution is greater than the initial comprehensive weight value solution, the next target node is flipped until all target nodes have been polled, and then the updated global undirected graph is output.

[0034] As an improvement to the above scheme, the step of sequentially flipping the target nodes that meet the node flipping conditions includes:

[0035] Calculate the importance value of each node in the initial undirected graph;

[0036] Sort each node according to its importance value from highest to lowest, and then flip the target nodes that meet the node flipping conditions in the order of sorting.

[0037] As an improvement to the above scheme, the node flipping condition is: the new solution after each node is flipped is among the candidate solutions of the subgraph corresponding to that node.

[0038] As an improvement to the above solution, the method further includes:

[0039] If the reference comprehensive weight value solution is less than or equal to the initial comprehensive weight value solution, the flipping of the current target node is abandoned, and the next target node is flipped.

[0040] As an improvement to the above scheme, the step of outputting a global solution for multi-user selection based on the global undirected graph includes:

[0041] The global undirected graph is iterated until a preset convergence condition is met, at which point a global solution selected by multiple users is output.

[0042] To achieve the above objectives, embodiments of the present invention also provide a multi-user selection device, comprising:

[0043] The weighted undirected graph construction module is used to construct a weighted undirected graph based on the acquired user information; wherein, each node of the weighted undirected graph represents a user, the weight value of the node represents the comprehensive weight of each user, and the edges between nodes represent the interference relationship between users;

[0044] An undirected graph segmentation module is used to segment the weighted undirected graph according to a preset node upper limit value to obtain at least two subgraphs;

[0045] The candidate solution output module is used to perform quantum algorithm calculations on each subgraph to obtain at least one candidate solution corresponding to each subgraph.

[0046] The global solution output module is used to construct a global undirected graph based on each subgraph and its corresponding candidate solutions, and output a global solution selected by multiple users based on the global undirected graph.

[0047] To achieve the above objectives, embodiments of the present invention also provide a multi-user selection device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the multi-user selection method as described in any of the above embodiments.

[0048] To achieve the above objectives, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the multi-user selection method as described in any of the above embodiments.

[0049] To achieve the above objectives, embodiments of the present invention also provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the multi-user selection method as described in any of the above embodiments.

[0050] Compared to existing technologies, this invention discloses a multi-user selection method, apparatus, device, storage medium, and program product. Through the parallel processing capabilities of quantum computing, it significantly improves the solution speed for larger-scale user selection tasks in future wireless networks, achieving near real-time resource allocation. Furthermore, quantum computing provides solutions closer to the global optimum, avoiding the problem of classical algorithms easily getting trapped in local optima, thus improving the solution accuracy in the multi-user selection process. This allows for more rational resource allocation to users and improves communication quality. Additionally, graph segmentation technology automatically adapts to changes in user distribution and overall weights, ensuring the real-time performance and effectiveness of the user selection scheme. Finally, integrating and optimizing the quantum computing results using classical computing optimization algorithms effectively handles errors in the quantum computing process, improving the robustness of the multi-user selection scheme. Attached Figure Description

[0051] Figure 1 This is a flowchart of a multi-user selection method provided in an embodiment of the present invention;

[0052] Figure 2 This is a flowchart of constructing a weighted undirected graph provided in an embodiment of the present invention;

[0053] Figure 3 This is an example of a weighted undirected graph provided in the embodiments of the present invention;

[0054] Figure 4 This is a flowchart of segmenting a weighted undirected graph provided in an embodiment of the present invention;

[0055] Figure 5 This is an example of segmenting a weighted undirected graph provided in an embodiment of the present invention;

[0056] Figure 6 This is a flowchart of obtaining candidate solutions using a quantum algorithm, provided in an embodiment of the present invention;

[0057] Figure 7 This is a circuit diagram of the quantum approximation optimization algorithm provided in the embodiments of the present invention;

[0058] Figure 8 This is a flowchart of the process of outputting a global solution selected by multiple users, provided in an embodiment of the present invention;

[0059] Figure 9 This is a structural block diagram of the multi-user selection device provided in an embodiment of the present invention;

[0060] Figure 10 This is a structural block diagram of a multi-user selection device provided in an embodiment of the present invention. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] The multi-user selection method described in this invention is applied to network devices, which may be evolved Node Bs (eNBs or eNodeBs) in Long Term Evolution (LTE) systems, Next Generation Radio Access Network (NG RAN) devices, base stations (gNBs) in NR systems, radio controllers in Cloud Radio Access Networks (CRANs), or relay stations, access points, vehicle-mounted devices, wearable devices, hubs, switches, bridges, routers, or network devices in future Public Land Mobile Networks (PLMNs), etc.

[0063] See Figure 1 , Figure 1 This is a flowchart of a multi-user selection method provided in an embodiment of the present invention. The multi-user selection method includes:

[0064] S1. Construct a weighted undirected graph based on the obtained user information;

[0065] S2. The weighted undirected graph is divided according to the preset node upper limit value to obtain at least two subgraphs;

[0066] S3. Use a quantum algorithm to calculate each subgraph to obtain at least one candidate solution corresponding to each subgraph;

[0067] S4. Construct a global undirected graph based on each subgraph and its corresponding candidate solutions, and output a global solution selected by multiple users based on the global undirected graph.

[0068] For example, each node in the weighted undirected graph represents a user, the weight of the node represents the overall weight of each user, and the edges between nodes represent the interference relationships between users. The user information includes downlink channel state information and user priority. To address the "scale wall" problem of classical algorithms, academia and industry have recently turned their attention to quantum computing. Quantum computing possesses parallel computing capabilities, enabling parallel computation of multi-user selection, thereby accelerating problem-solving speed and providing higher solution accuracy. However, current quantum computers are currently in the NISQ (Noisy Intermediate-Scale Quantum) stage, with a limited number of computing units (qubits), and cannot directly compute large-scale optimization problems. Therefore, it is necessary to decompose the entire large-scale problem to meet the current computing power of quantum computers. Based on this, this invention models the multi-user selection problem as a weighted undirected graph, decomposes the entire large-scale problem into several smaller-scale problems through subgraph partitioning, and efficiently solves candidate solutions for each group of smaller-scale problems using interactive quantum computing devices. Then, the candidate solutions of each group of smaller-scale problems are jointly optimized to solve the large-scale user selection problem.

[0069] See Figure 2 , Figure 2 This is a flowchart of constructing a weighted undirected graph provided in an embodiment of the present invention. Step S1 specifically includes:

[0070] S11. Determine the channel gain for each user and the interference coefficient between users based on the downlink channel state information;

[0071] S12. Calculate the independent weight of each user based on the channel gain and the user priority;

[0072] S13. With the goal of maximizing the overall weight value of all users, construct an objective function based on the independent weights and the interference coefficients;

[0073] S14. Construct a weighted undirected graph based on the objective function.

[0074] Specifically, in step S11, the base station collects downlink channel state information for each user and the resource priority for each user. For example, the base station here can be a generalized transceiver, composed of multiple transceiver nodes, establishing connections with all users within the service area. Based on the collected downlink channel state information for each user, the base station calculates the independent channel gain q for each user. i And the interference coefficient ρ between user channels i,j The calculation of channel gain and interference coefficient can be referenced from existing technologies, and will not be elaborated upon here. Simultaneously, the base station also collects the resource priority p of each user. i .

[0075] For example, the downlink channel state information may be channel matrix information composed of all downlink channels from all transmitting points to the user. The user priority can be determined based on at least one of the user's service type, channel conditions, and system state. For instance, different service types have different requirements and sensitivities to communication resources; real-time services usually have higher priority. Users with poor channel quality may be given higher priority because if resource allocation for these users is not prioritized, they may have difficulty obtaining good communication quality or even being unable to communicate normally. When the network load is light, resource allocation may focus more on the user's service needs and priority level, while when the network load is heavy, to avoid network congestion and system crashes, priority may be given to ensuring the communication needs of some critical users or users whose communication needs have a significant impact on system performance.

[0076] Specifically, in step S12, the base station calculates the weighted summation for the i-th user u. i Independent weights ω i It satisfies the following formula:

[0077] ω i =α*q i +β*p i (1);

[0078] Here, α and β are adjustment parameters that determine the proportion of influence of channel quality and priority in the overall weight.

[0079] Specifically, in step S13, the wireless network is associated with the multi-user set U = {u1, u2, ..., u} of the base station. n Select users from the list to maximize the total comprehensive weight value during communication. The total comprehensive weight value C satisfies the following formula:

[0080]

[0081] Where δ represents noise power, which is part of the denominator in the formula for calculating the overall weight, and is used to measure the impact of noise on signal reception; E(i,j) represents the user pair consisting of the i-th user and the j-th user.

[0082] For example, the optimization objective is to maximize the total comprehensive weight value. This requires simplifying the model and objective function, starting with simplifying the modeling of interference terms. Interference between users is usually related to their distance from each other, exhibiting a near-neighbor interaction phenomenon with significant attenuation due to distance. Therefore, a threshold can be set: interference exists above the threshold and not below it. Further simplification of the objective function involves comprehensively considering the weight ω of each node. i and interference edge weight ρ i,jThe optimization objective is to maximize the overall weight and minimize the interference by selecting a subset of users. The final optimization objective function is transformed into:

[0083]

[0084] Where S represents the selected subset of users, that is, a set of users chosen from all users; E s This represents the set of interfering edges between users in this subset, indicating that there is interference between these user pairs; λ is the penalty coefficient; x i The indicator variable is denoted as whether user i is selected. A value of 0 indicates that user i is not in the user subset S, while a value of 1 indicates that user i is in the user subset S. The goal of this formula is to optimize resource allocation by selecting an appropriate user subset S, maximizing the total sum of the selected user weights while minimizing the total interference among these users.

[0085] Specifically, in step S14, each user is regarded as a node in the graph. The weight of the node is the comprehensive weight value of each user. The edges between nodes represent the interference between users (if the interference coefficient is greater than a threshold value, there is an edge; if the interference coefficient is less than or equal to the threshold value, there is no edge), thus obtaining a weighted undirected graph.

[0086] For example, for any two users u in a subset S i and u j Compare their interference coefficients ρ i,j The relationship between ρ and a given threshold value, if ρ i,j If it exceeds the threshold value, then at node u i and u j Connect them with an edge, the weight of which is ρ. i,j This indicates that there is interference between the two. If ρ i,j If the value is less than or equal to the threshold, then node u i and u j There are no edges between them, and their interference is considered negligible. Through the above steps, the user set is transformed into a weighted undirected graph with comprehensive weights as node values ​​and interference relationships (based on threshold judgment) as edges, providing a model foundation for subsequent graph segmentation and multi-user selection algorithms. Figure 3 As shown, Figure 3 This is an example of a weighted undirected graph provided in this embodiment of the invention. There are 16 nodes, representing 16 users. Taking node 0 as an example, its independent weight is 1.67, and it interferes with nodes 3, 9, 10, and 13.

[0087] In this embodiment of the invention, by constructing a weighted undirected graph, the comprehensive weight of users (integrating channel quality, priority, etc.) and the relationship with interference are presented intuitively, providing a clear structured model for multi-user selection and making resource allocation optimization more in line with actual communication scenarios.

[0088] See Figure 4 , Figure 4 This is a flowchart of segmenting a weighted undirected graph provided in an embodiment of the present invention. In this case, step S2 specifically includes:

[0089] S21. Use a community detection algorithm to identify the initial community structure in the weighted undirected graph;

[0090] S22. When the number of nodes in any initial community structure is greater than the node upper limit, the initial community structure is divided to obtain the corresponding sub-community structure.

[0091] S23. Aggregate nodes of all community structures to obtain a grid structure;

[0092] S24. The mesh structure is divided according to the node upper limit value and the preset optimization target to obtain at least two sub-graphs.

[0093] For example, the node limit N is determined based on the maximum subgraph size that the quantum computing system can process. The base station obtains the upper limit N of the computing bit capacity of the configured quantum computing system, which can be input in the form of a one-dimensional array. Each element is the upper limit of the computing bits of the associated computer, meaning that the quantum computing system can only compute multi-user selection problems in subgraphs with no more than N nodes. Based on the configuration of the weighted undirected graph, graph partitioning technology is used to divide the entire weighted undirected graph into multiple subgraphs. This partitioning of a large-scale weighted undirected graph into multiple subgraphs satisfies two core conditions: 1) Subgraph size constraint: the number of nodes in each subgraph is close to and less than N, avoiding exceeding the quantum computing capacity; 2) Sparse connections between subgraphs: the number of connections between subgraphs is as small as possible, reducing cross-subgraph interference and computational complexity.

[0094] It should be noted that in a weighted undirected graph, the edge weights can represent the intensity of interference or communication needs between users. A smaller number of cut edges implies weaker interactions between subgraphs, making it suitable for parallel processing in quantum computing. For the multi-user selection problem in wireless networks, the model considers interference between users primarily to originate from their spatial locations. Interference is a nearest-neighbor interaction, therefore the interference matrix is ​​sparse and exhibits community properties.

[0095] Specifically, in step S21, for the entire weighted undirected graph, a community detection algorithm is used to partition the graph into communities, resulting in several initial community structures. The community detection algorithm can be one of the following: Louvain's algorithm, Local Search (LS) algorithm, Infomap algorithm, Girvan-Newman algorithm, label propagation algorithm, etc. In this embodiment of the invention, Louvain's algorithm will be used as an example for subsequent explanation.

[0096] Specifically, in step S22, when the number of nodes in any initial community structure is greater than the node upper limit, the initial community structure is divided to obtain the corresponding sub-community structure.

[0097] For example, the Louvain algorithm is applied to the entire weighted undirected graph for community partitioning. A greedy strategy is used to continuously move nodes to adjacent communities, maximizing modularity until a local optimum is reached. Utilizing community properties, nodes with dense interference (i.e., tightly connected nodes) are initially clustered into the same community, forming an initial partitioning framework and providing a "seed" structure for subsequent optimization. The Louvain algorithm is iterated, continuously moving nodes to adjacent communities to maximize modularity. Modularity is an indicator of the tightness of connections within a community and the sparsity of connections between communities. Optimizing modularity is not the final goal, but rather a means to generate the initial community structure. The Louvain algorithm can initially assign nodes with relatively dense interference to the same community, which is beneficial for subsequent subgraph boundary optimization. The initially partitioned community structure is adjusted so that the number of nodes in each community meets the upper limit N of a quantum computer. For communities exceeding the scale (i.e., the number of nodes is greater than N), a graph partitioning algorithm (such as the Kernighan-Lin algorithm or a heuristic method) is used for secondary partitioning, where the optimization objective is to minimize the sum of edge weights and the number of nodes between subgraphs, reducing computational complexity. Here, a balancing term can be introduced to control the size of the subgraph, so that the number of nodes in each subgraph is balanced and as close as possible to the computational limit N of a quantum computer.

[0098] Specifically, in step S23, to simplify the graph size, nodes are aggregated through graph coarsening to form a simpler network structure. The coarsening process merges multiple densely connected nodes into a "super node" and recalculates the edge weights, making the segmentation of the entire graph more efficient.

[0099] Specifically, in step S24, the network structure after coarsening is further divided, reducing the complexity of the original graph and optimizing the overall segmentation efficiency. It's important to note that the coarsened graph is only used for segmentation optimization; the final subgraphs are restored to their original nodes for quantum computing processing. The number of cutting edges between each subgraph is considered during the target optimization process. This is achieved by combining graph coarsening assistance with GNN graph feature extraction, Kernighan-Lin algorithms, heuristic partitioning algorithms, etc., in conjunction with community detection algorithms to optimize the target and generate the final set of subgraphs. Figure 5 As shown, Figure 5 This is an example of segmenting a weighted undirected graph according to an embodiment of the present invention. The graph is divided into four subgraphs. During the segmentation of the weighted undirected graph, it is necessary to ensure that the number of nodes in each subgraph does not exceed N, and at the same time, the number of cut edges is minimized.

[0100] In this embodiment of the invention, the weighted undirected graph is divided according to the upper limit value of the nodes to ensure that the size of the subgraph matches the processing capacity of the quantum computing system and avoid computational overload. At the same time, the sparsity of the connections between subgraphs reduces cross-subgraph interference, improving the efficiency and feasibility of quantum computing.

[0101] See Figure 6 , Figure 6 This is a flowchart of obtaining candidate solutions using a quantum algorithm provided in an embodiment of the present invention. In this case, step S3 specifically includes:

[0102] S31. Obtain the subgraph information corresponding to each subgraph; wherein, the subgraph information includes the comprehensive weight value of the subgraph and a two-dimensional array of inter-user interference signals;

[0103] S32. The quantum algorithm is used to repeatedly calculate the information of each subgraph according to the preset number of repetitions to obtain the number of initial solutions corresponding to the number of repetitions.

[0104] S33. Obtain candidate solutions that satisfy the candidate conditions from all initial solutions, and obtain at least one candidate solution corresponding to each subgraph.

[0105] It should be noted that traditional algorithms (such as greedy algorithms and genetic algorithms) are prone to getting trapped in local optima, while quantum algorithms (such as quantum approximation optimization algorithms) utilize quantum parallelism and superposition states to explore multiple candidate solutions in the global solution space simultaneously, increasing the probability of finding the global optimum. This is especially suitable for NP-hard combinatorial optimization problems, such as the Weighted Maximum Independent Set (WMIS).

[0106] Specifically, in step S31, the modeling of the segmented subgraph is also a weighted maximum independent set (WMIS) problem, solved using a quantum algorithm. The quantum algorithm can quickly find the user combination with the maximum weight and explore multiple candidate solutions globally, avoiding the local optimum trap of traditional algorithms. In graph theory, an independent set refers to a group of non-adjacent nodes. WMIS requires maximizing the sum of the weights of the nodes in the independent set. In the context of the multi-user selection problem, the comprehensive weight value processed by the classical computing unit of the base station and the inter-user interference signal are imported as parameters into the associated quantum computing system. The comprehensive weight value is a two-dimensional array labeled with user units, where each comprehensive weight value is a numerical value. The user interference signal is also a two-dimensional matrix, with rows and columns equal to the number of users. Each element in the two-dimensional matrix corresponds to the interference between user i and user j.

[0107] Specifically, in step S32, the quantum algorithm can be the Quantum Approximate Optimization Algorithm (QAOA), where the number of bits at the quantum circuit input is equal to the number of nodes in each segmented subgraph. The classical computing unit groups the segmented subgraphs, and each group contains: 1. a two-dimensional array of comprehensive weight values; 2. a two-dimensional matrix of inter-user interference signals. The data is sent to the associated quantum computer for serial or parallel computation.

[0108] For example, see Figure 7 , Figure 7 This is a circuit diagram of the quantum approximation optimization algorithm provided in this embodiment of the invention. In QAOA, p is the circuit depth, representing the number of alternations between the problem Hamiltonian and the driving Hamiltonian in the hybrid layer. The larger p is, the more complex the circuit, and theoretically, it can capture a more complex solution space structure. The more nodes in the subgraph, the higher the problem complexity, requiring a larger p. The calculation process of the quantum approximation optimization algorithm is as follows:

[0109] 1) Mixed state The initial state indicates that the circuit starts with n qubits initialized to the |0> state. n corresponds to the number of nodes in the subgraph after partitioning (each qubit represents a user). To represent the generation of mixed states, a Hadamard gate is applied to n qubits. By preparing all qubits into a superposition state, forming a "mixed state," the foundation for quantum parallelism is laid.

[0110] 2) Ansatz (p-layer circuit), consisting of p layers of repeated quantum gate operations (corresponding to circuit depth p), each layer contains two core operations: the first parameter Related to the problem Hamiltonian (encoding the objective function of multi-user selection, such as maximizing user comprehensive weights or suppressing interference), the parameter β is similar to the time evolution parameter and is used to control the evolution of the quantum state under the problem Hamiltonian; the second parameter Related to the driving Hamiltonian (used for searching in the quantum state space, introducing quantum randomness and mixing to avoid local optima), the parameter γ is similar to the time evolution parameter, used to control the evolution of the quantum state under the problem Hamiltonian. By alternating these two types of operations, the optimal solution is explored in the quantum state space. The larger p is, the more refined the circuit characterizes the solution space (this needs to be combined with the subgraph size to balance complexity and decoherence risk).

[0111] 3) Measurement: The quantum bit is measured, and the quantum state is collapsed into a classical binary solution. Each solution corresponds to a user-selectable combination.

[0112] This roadmap fully presents the process of QAOA from initialization, quantum state preparation, parameterization optimization to measurement output. By adjusting parameters such as p, β, and γ, it efficiently searches for the optimal user combination by leveraging quantum parallelism, avoiding the local optimum trap of traditional algorithms.

[0113] Specifically, in step S33, candidate solutions that meet the candidate conditions are obtained from all initial solutions, including: obtaining the top n1 target initial solutions in terms of occurrence probability from all initial solutions; using the target initial solutions to perform weighted calculations on the current subgraph to obtain the initial weight value corresponding to each target initial solution, n1≥2; obtaining the top n2 target initial weight values ​​from all initial weight values, and using the target initial solutions corresponding to the target initial weight values ​​as candidate solutions, n2≥1.

[0114] For example, the depth parameter p of the quantum circuit is set according to the size of the segmented subgraph, and the initial parameters of the circuit are preset based on parameters similar to those of previous data structures. The number of repetitions M is set, and a series of feasible solutions are obtained through quantum circuit computation. For the top n1 target initial solutions with the highest probability of occurrence, the current subgraph is weighted using these target initial solutions to obtain initial weight values ​​corresponding to each target initial solution. The top n2 candidate solutions are obtained by comparing the initial weight values. The k different segmented subgraphs are computed in parallel or serially using multiple quantum computing systems to obtain a two-dimensional array (each of the k subgraphs has n2 candidate solutions).

[0115] In this embodiment of the invention, the parallelism and global search advantages of quantum algorithms (such as quantum approximation optimization algorithms) are utilized to quickly solve for the optimal user combination of a subgraph, avoiding the local optimum trap of traditional algorithms, generating high-quality candidate solutions for each subgraph, and enhancing the local optimization effect.

[0116] See Figure 8 , Figure 8 This is a flowchart of the output of a global solution selected by multiple users provided in an embodiment of the present invention. In this case, step S4 specifically includes:

[0117] S41. Construct an initial undirected graph based on each subgraph and its corresponding candidate solutions, and calculate the initial comprehensive weight value solution of the initial undirected graph;

[0118] S42. Flip the target nodes that meet the node flipping conditions in sequence, and update the initial undirected graph according to the flipped nodes;

[0119] S43. Calculate the updated reference composite weight value solution for the initial undirected graph;

[0120] S44. If the reference comprehensive weight value solution is greater than the initial comprehensive weight value solution, flip the next target node until all target nodes have been polled, and output the updated global undirected graph.

[0121] Specifically, in step S41, the k candidate solutions for the subgraphs are combined and optimized to obtain a global solution. First, a global overall weight value solution is constructed based on the optimal solution of each subgraph, and the objective function for optimization is the overall weight value. The aim is to construct a global solution through the initial combination of optimal solutions of the subgraphs, and to measure the initial effect of global resource allocation by maximizing the overall weight value as the optimization direction.

[0122] Specifically, in step S42, the step of sequentially flipping the target nodes that satisfy the node flipping condition includes: calculating the importance value of each node in the initial undirected graph; sorting the nodes according to their importance values ​​from high to low; and sequentially flipping the target nodes that satisfy the node flipping condition according to the sorting order. The node flipping condition is that the new solution after flipping each node is among the candidate solutions in the subgraph corresponding to that node.

[0123] It's important to note that the core function of node flipping is to optimize the quality of the global solution by adjusting the affiliation or state of nodes in the candidate solution space of the subgraph. In multi-user selection problems, the connections between subgraphs are sparse and limited, and some nodes have a significant impact on global communication quality and resource allocation (such as nodes with a large number of connections). By flipping the candidate solution space of the subgraphs to which these important nodes belong, we can change the resource allocation or connection relationships of nodes, explore better global resource configuration methods, and thus improve the overall global weight value, making network resource allocation more reasonable.

[0124] For example, the edges between subgraphs in the global graph formed by the optimal solutions of the subgraphs are sparse and finite. For the generated intersecting edges, the importance of each node is determined based on the number of edges it connects to, and the nodes are ranked according to the importance parameter. The number of edges connected to each node is counted; the more edges a node has, the more critical it is in the global connectivity, and the greater its impact on interference distribution and resource allocation. Based on this, the importance of each node is determined and ranked. For example, if node A connects to 5 edges between subgraphs and node B connects to 2, then node A is more important than node B. Following the order of node importance from high to low, a "node flip" is performed on the candidate solution space of the subgraph to which the node belongs. Flipping essentially involves adjusting the resource allocation or connection relationships of nodes (such as changing the subgraph a node connects to, adjusting its occupied spectrum / power resources, etc.), trying different configuration methods to optimize the quality of the global solution. For example, for node A, which has high importance, a different resource allocation mode is switched in the candidate solutions of its subgraph, and the impact on the global solution is observed.

[0125] For example, not all nodes need to be flipped. Instead, nodes are judged sequentially according to their importance to determine if they meet the flipping conditions. If they do, they are flipped; otherwise, the flipping is abandoned. Before performing a node flip, it is first determined whether the new solution generated after the flip exists in the set of n² candidate solutions of the subgraph. If it does not exist, it means that the solution generated by this flip operation does not meet the computation rules or resource constraints of the subgraph, so no flipping is performed, and the operation is directly applied to the next node. If it exists, a global solution reconstruction is performed, that is, the flipped node solution is integrated into the initial global graph.

[0126] Specifically, in step S43, after a target node is flipped, since the initial global graph has been updated, the reference comprehensive weight value solution of the updated initial undirected graph can be recalculated to evaluate the effect of the flipping operation.

[0127] Specifically, in step S44, if the reference comprehensive weight value solution is greater than the initial comprehensive weight value solution, the next target node is flipped, and this process continues until all target nodes have been polled, at which point the updated global undirected graph is output. Further, the method also includes: if the reference comprehensive weight value solution is less than or equal to the initial comprehensive weight value solution, the flipping of the current target node is abandoned, and the next target node is flipped.

[0128] For example, if the overall global weight value increases after the operation, it indicates that the node flip optimization is effective. The change is retained, and the same optimization operation is performed on the next node until all nodes have been processed, forming a new global undirected graph space. If the weight value does not increase, the flip is abandoned, and the next node is processed. In this way, the quality of the global solution is gradually improved.

[0129] Further, in step S4, the step of outputting a global solution selected by multiple users based on the global undirected graph includes: iterating the global undirected graph until a preset convergence condition is met, and then outputting a global solution selected by multiple users.

[0130] For example, steps S41 to S44 are repeated continuously, namely, continuously sorting the importance of nodes, flipping node solutions, judging the validity of solutions, reconstructing the global solution, and calculating weight values. When the preset convergence conditions are met (such as the number of iterations reaching a set value, or the average performance of modularity reaching a set value), the iteration stops, and it is considered that the global solution has reached relative optimality, completing the global optimization for multi-user selection.

[0131] In this embodiment of the invention, a global undirected graph is constructed based on the subgraph and its candidate solutions. Local optimization results are integrated to further explore the global optimal solution. This hierarchical optimization strategy (from local to global) improves the overall efficiency of wireless network resource allocation, effectively suppresses interference, increases system throughput and communication quality, and enhances user experience.

[0132] Furthermore, the present invention provides a specific embodiment, including the following steps:

[0133] Step 1: The base station collects downlink channel status information (channel matrix information composed of downlink channels from all transmitting points to the user) and resource priorities of nearby managed users. The number of users is 200.

[0134] Step 2: Calculate the channel gain q for each user based on the built-in algorithm of the base station server. i Interference intensity ρ between users i,j And each user's priority for resources p i When the adjustment parameters for the overall weight are set to α = 0.5 and β = 0.5, the independent weights for each user satisfy the following:

[0135] ω i =0.5q i +0.5p i (4);

[0136] Step 3: In the multi-user selection, the objective function of the base station is to maximize the total comprehensive weight value C, which satisfies the above formula (2).

[0137] In the actual calculation process, the simplified model and objective function are transformed into a weighted maximum independent set problem. To address the difficulty of fractional solutions, it is necessary to simplify the interference terms. The solution involves setting a threshold: channel allocation interference above the threshold is prohibited, represented by the presence of an edge in an undirected graph. Channel allocation below the threshold is allowed, represented by the absence of an edge between two points in an undirected graph. For example, the interference term is first normalized to the interval [0,1], with a threshold of 0.6. Above 0.6, an edge exists; below 0.6, no edge exists.

[0138] Step 4: Based on the simplified maximum independent set model in Step 3, the objective function is transformed into solving for the Hamiltonian. The maximum value of satisfies the following formula:

[0139]

[0140] Where, x i A value of 0 or 1 corresponds to whether user i or x is selected. j A value of 0 or 1 corresponds to whether user j is selected.

[0141] Step 5: The variable in the quantum computing system is z i The value can be either -1 or 1. Therefore, x in the objective function... i Replace with x j Replace with In the formula, I is the identity matrix, and z i and z j All are Pauli Z matrices. Solving for the maximum value of the Hamiltonian is also transformed into solving for its negative value. The minimum value. The objective Hamiltonian of the problem is transformed through calculation into:

[0142]

[0143] Step Six: Each user within the base station corresponds to a node in the weighted undirected graph, and each user has an independent weight ω. i It is the weight of the points in the weighted undirected graph. Then, based on the interference determination in step three, the edge matrix of the weighted undirected graph is constructed, and the weighted undirected graph G = (V, E) is automatically generated. V represents a two-dimensional array of the comprehensive weight value parameters of each user, and E represents the channel interference matrix between users.

[0144] Step 7: The base station obtains the upper limit of the computational qubits of the associated quantum computing system, N = 30.

[0145] Step 8: Call the graph partitioning algorithm pre-built into the base station computing cluster.

[0146] Step 9: Since the interference between users is a spatially related nearest neighbor interaction, the interference between users is a sparse matrix of 200*200 and has community properties.

[0147] Step 10: The algorithm first applies the Louvain algorithm to the undirected graph G = (V, E) for community detection. This algorithm can efficiently identify community structures in the graph.

[0148] Step 11: Iteratively run the Louvain algorithm, attempting to move nodes to neighboring communities to maximize modularity. Set the number of iterations to 50 or the average performance of modularity to 0.2 as the convergence result. Once convergence is achieved, the community structure is generated.

[0149] Step 12: This process reduces the complexity of the graph while maintaining its structural characteristics and the integrity of its original information. At this stage, the merging criteria consider the connectivity between vertices and the weights of edges, merging highly connected nodes as much as possible. This initially distributes nodes with denser interference into the same community, which is beneficial for subsequent subgraph boundary optimization.

[0150] Step 13: Determine the size of the subgraph and set the saturation parameter 30 as the upper limit N of the quantum computer's computational qubits.

[0151] Step Fourteen: If some subgraphs still do not meet this condition, further graph partitioning or merging is required to adjust the initial community structure so that the number of nodes in each subgraph meets the quantum computer's upper limit of 30. A secondary partitioning is performed by calling the base station computing cluster's built-in partitioning algorithm, such as the Kernighan-Lin algorithm or the heuristic k-way partitioning algorithm.

[0152] Step 15: Define the objective function to minimize the number of cutting edges between subgraphs to improve computational efficiency. During optimization, graph coarsening is used to aggregate nodes, forming a simpler network structure. The coarsening process merges multiple densely connected nodes into a "super node" and recalculates edge weights, making the segmentation of the entire graph more efficient.

[0153] Step 16: During the optimization process, consider the balance of subgraph sizes to avoid producing subgraphs that are too large or too small. This can be achieved by introducing a subgraph size adjustment term to dynamically adjust the division between subgraphs.

[0154] Step 17: Obtain the set of partitioned subgraphs S = [25, 24, 28, 23, 25, 26, 24, 25]. Each partitioned subgraph is a weighted maximum independent set problem to describe the multi-user selection task in a wireless network.

[0155] Step 18: Each subgraph satisfies the computational capability of the associated quantum computing system. Input these subgraphs into the quantum computing system for parallel or serial computation. The classical computing unit inputs 8 sets of parameters into the quantum computing unit, each set including: 1. A two-dimensional array of user comprehensive weight values; 2. A two-dimensional matrix of interference signals between users.

[0156] Step 19: Calculate the weighted maximum independent set of the subgraph using the quantum approximation optimization algorithm.

[0157] Step 20: The number of bits in the input quantum computing system is equal to the number of nodes in the subgraph [25, 24, 28, 23, 25, 26, 24, 25]. The input variable of the quantum approximation optimization algorithm is generated by the built-in learning algorithm analyzing the structure of the subgraph. Figure 7 [β1…β] in Ansatz 20 ,γ1…γ 20 By utilizing problems with similar graph structures in previous data, the initial parameters of the circuit can be preset to improve computational efficiency.

[0158] Step 21: The depth P and number of repetitions M of the quantum approximation optimization algorithm are determined by the size of the input subgraph and are automatically identified and imported by the quantum computing device to ensure that the algorithm can reach the optimal solution at a reasonable depth. In this example, the depth p is set to 20 and the number of repetitions to 2000.

[0159] Step 22: The quantum circuit calculation obtains a series of feasible solutions. The total comprehensive weight value is calculated for the top n1 feasible solutions with the highest probability of occurrence (determined by the size of the subgraph; in the example, the top 10 are selected).

[0160] Step 23: Compare the calculated total comprehensive weight values ​​and select the top n2 candidate solutions with the largest total comprehensive weight values ​​(determined by the subgraph size; in the example, the top 5 are selected) to facilitate final optimization. These candidate solutions represent channel allocation schemes with better total comprehensive weight values, ensuring the diversity and reliability of the selected schemes.

[0161] Step 24: Perform parallel computation on the 8 different subgraphs after segmentation through multiple quantum computing systems, or perform serial computation on a single system (the example obtains 8 dimensions, 5 candidate solutions for each dimension, forming a two-dimensional array).

[0162] Step 25: Perform global optimization on the classic computing units of the generated two-dimensional array transmitted back to the base station.

[0163] Step 26: Combine the optimal solutions of each subgraph into a global solution, and calculate the total comprehensive weight value as the initial global solution. The objective function of optimization is the total global comprehensive weight value, aiming to maximize system performance.

[0164] Step 27: Construct a global graph based on the optimal solutions of each subgraph. Selected nodes are edgeless within the subgraph (no interference). However, very sparse edges may be generated between subgraphs due to node selection (the number of edges between subgraphs is sparse during partitioning, and edges are only retained when two points are selected at the same time).

[0165] Step 28: The edges between subgraphs are sparse and finite. For the edges that are generated, the importance of each node is determined according to the number of edges generated by the node, and the nodes that form the edges are sorted by importance.

[0166] Step 29: Each subgraph has only some nodes connected to other subgraphs. Candidate solutions for a subgraph might include cases like XXX0011 and XXX1100. Nodes are ranked by importance, and nodes are flipped, for example, from feasible solution XXX0011 to XXX1100. If no candidate solution has a flipped node, the next important node is flipped.

[0167] Step 30: After flipping, reconstruct the new large graph, calculate the new total comprehensive weight value and the situation of interference edges. If the overall optimization is improved, retain it and continue to optimize the next node until all important nodes are optimized and a new global solution is formed.

[0168] Step 31: Repeat steps 28 to 30 to continuously iterate and optimize the global solution of the entire weighted undirected graph until the global total weight value reaches the preset convergence condition, and finally obtain the multi-user selection scheme of the optimized global undirected graph.

[0169] In this embodiment of the invention, the parallel processing capability of quantum computing can significantly improve the solution speed of large-scale user selection tasks in future wireless networks, achieving near real-time resource allocation. Furthermore, quantum computing provides solutions closer to the global optimum, avoiding the problem of classical algorithms easily getting trapped in local optima, thus improving the solution accuracy in multi-user selection processes. This allows for the rational allocation of resources to users and improves communication quality. Additionally, graph segmentation technology can automatically adapt to changes in user distribution and overall weights, ensuring the real-time performance and effectiveness of the user selection scheme. Finally, integrating and optimizing quantum computing results using classical computing optimization algorithms effectively handles errors in the quantum computing process, improving the robustness of multi-user selection schemes.

[0170] See Figure 9 , Figure 9 This is a structural block diagram of a multi-user selection device 100 provided in an embodiment of the present invention. The multi-user selection device 100 includes:

[0171] The weighted undirected graph construction module 11 is used to construct a weighted undirected graph based on the acquired user information; wherein, each node of the weighted undirected graph represents a user, the weight value of the node represents the comprehensive weight of each user, and the edges between nodes represent the interference relationship between users;

[0172] The undirected graph segmentation module 12 is used to segment the weighted undirected graph according to a preset node upper limit value to obtain at least two subgraphs;

[0173] The candidate solution output module 13 is used to perform quantum algorithm calculations on each subgraph to obtain at least one candidate solution corresponding to each subgraph.

[0174] The global solution output module 14 is used to construct a global undirected graph based on each subgraph and its corresponding candidate solutions, and output a global solution selected by multiple users based on the global undirected graph.

[0175] Specifically, the weighted undirected graph construction module 11 includes:

[0176] The parameter determination unit is used to determine the channel gain of each user and the interference coefficient between users based on the downlink channel state information.

[0177] An independent weight calculation unit is used to calculate the independent weight of each user based on the channel gain and the user priority.

[0178] The objective function construction unit is used to construct an objective function based on the independent weights and the interference coefficients, with the goal of maximizing the comprehensive weight value of all users.

[0179] A weighted undirected graph construction unit is used to construct a weighted undirected graph based on the objective function.

[0180] Specifically, the undirected graph segmentation module 12 includes:

[0181] An initial community structure building unit is used to identify the initial community structure in the weighted undirected graph using a community detection algorithm;

[0182] An initial community structure partitioning unit is used to partition the initial community structure to obtain the corresponding sub-community structure when the number of nodes in any initial community structure is greater than the upper limit value of the number of nodes.

[0183] The community structure aggregation unit is used to aggregate nodes of all community structures to obtain a grid structure.

[0184] An undirected graph segmentation unit is used to segment the mesh structure according to the upper limit value of the nodes and a preset optimization objective to obtain at least two subgraphs.

[0185] Specifically, the candidate solution output module 13 includes:

[0186] A subgraph information acquisition unit is used to acquire subgraph information corresponding to each subgraph; wherein, the subgraph information includes the comprehensive weight value of the subgraph and a two-dimensional array of inter-user interference signals;

[0187] An initial solution generation unit is used to repeatedly calculate the information of each subgraph using a quantum algorithm according to a preset number of repetitions, so as to obtain a number of initial solutions corresponding to the number of repetitions.

[0188] The candidate solution generation unit is used to obtain candidate solutions that satisfy the candidate conditions from all initial solutions, and to obtain at least one candidate solution corresponding to each subgraph.

[0189] Specifically, the candidate solution generation unit is used to: obtain the top n1 target initial solutions in terms of occurrence probability from all initial solutions; perform weighted calculations on the current subgraph using the target initial solutions to obtain an initial weight value corresponding to each target initial solution, where n1≥2; obtain the top n2 target initial weight values ​​from all initial weight values, and use the target initial solutions corresponding to the target initial weight values ​​as candidate solutions, where n2≥1.

[0190] Specifically, the global solution output module 14 includes:

[0191] An initial undirected graph construction unit is used to construct an initial undirected graph based on each subgraph and its corresponding candidate solutions, and to calculate the initial comprehensive weight value solution of the initial undirected graph.

[0192] A node flipping unit is used to flip target nodes that meet the node flipping conditions in sequence, and update the initial undirected graph based on the flipped nodes.

[0193] The weight value calculation unit is used to calculate the reference comprehensive weight value solution of the updated initial undirected graph;

[0194] The judgment unit is used to flip the next target node if the reference comprehensive weight value solution is greater than the initial comprehensive weight value solution, until all target nodes have been polled, and then output the updated global undirected graph.

[0195] Specifically, the judgment unit is further configured to: if the reference comprehensive weight value solution is less than or equal to the initial comprehensive weight value solution, abandon the flipping of the current target node and flip the next target node.

[0196] Specifically, the node flipping unit is used to: calculate the importance value of each node in the initial undirected graph; sort the nodes according to their importance values ​​from high to low; and flip the target nodes that meet the node flipping conditions in sequence according to the sorting order.

[0197] It is worth noting that the working process of each module in the multi-user selection device 100 described in the embodiments of the present invention can refer to the working process of the multi-user selection method described in the above embodiments, and will not be repeated here.

[0198] See Figure 10 , Figure 10 This is a structural block diagram of a multi-user selection device 200 provided in an embodiment of the present invention. The multi-user selection device 200 includes a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, it implements the steps in the various multi-user selection method embodiments described above.

[0199] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the multi-user selection device 200.

[0200] The multi-user selection device 200 may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will understand that the schematic diagram is merely an example of the multi-user selection device 200 and does not constitute a limitation on the multi-user selection device 200. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the multi-user selection device 200 may also include input / output devices, network access devices, buses, etc.

[0201] The processor 21 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 21 is the control center of the multi-user selection device 200, connecting all parts of the multi-user selection device 200 via various interfaces and lines.

[0202] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements various functions of the multi-user selection device 200 by running or executing the computer programs and / or modules stored in the memory 22 and calling the data stored in the memory 22. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 22 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0203] Wherein, if the modules / units integrated in the multi-user selection device 200 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor 21, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0204] Furthermore, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the multi-user selection method as described in any of the above embodiments.

[0205] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A multi-user selection method, characterized in that, include: A weighted undirected graph is constructed based on the acquired user information; wherein each node in the weighted undirected graph represents a user, the weight value of the node represents the comprehensive weight of each user, and the edges between nodes represent the interference relationship between users; The weighted undirected graph is divided according to a preset node upper limit value to obtain at least two subgraphs; A quantum algorithm is used to compute each subgraph to obtain at least one candidate solution corresponding to each subgraph. Construct a global undirected graph based on each subgraph and its corresponding candidate solutions, and output a global solution selected by multiple users based on the global undirected graph.

2. The multi-user selection method as described in claim 1, characterized in that, The user information includes downlink channel state information and user priority.

3. The multi-user selection method as described in claim 2, characterized in that, The step of constructing a weighted undirected graph based on the acquired user information includes: The channel gain for each user and the interference coefficient between users are determined based on the downlink channel state information. Calculate the independent weight for each user based on the channel gain and the user priority; With the goal of maximizing the overall weight value of all users, an objective function is constructed based on the independent weights and the interference coefficients. Construct a weighted undirected graph based on the objective function.

4. The multi-user selection method as described in claim 1, characterized in that, The weighted undirected graph is segmented according to a preset node upper limit value to obtain at least two subgraphs, including: The initial community structure in the weighted undirected graph is identified using a community detection algorithm; When the number of nodes in any initial community structure exceeds the node upper limit, the initial community structure is divided to obtain the corresponding sub-community structure; By aggregating nodes across all community structures, a grid structure is obtained. The mesh structure is divided according to the node upper limit and the preset optimization target to obtain at least two sub-graphs.

5. The multi-user selection method as described in claim 1, characterized in that, The node upper limit is determined based on the maximum subgraph size that the quantum computing system can process.

6. The multi-user selection method as described in claim 1, characterized in that, The step of using a quantum algorithm to calculate each subgraph to obtain at least one candidate solution corresponding to each subgraph includes: Obtain the subgraph information corresponding to each subgraph; wherein, the subgraph information includes the comprehensive weight value of the subgraph and a two-dimensional array of inter-user interference signals; The quantum algorithm is used to repeatedly calculate the information of each subgraph according to the preset number of repetitions to obtain the number of initial solutions corresponding to the number of repetitions. From all initial solutions, obtain candidate solutions that satisfy the candidate conditions, and obtain at least one candidate solution corresponding to each subgraph.

7. The multi-user selection method as described in claim 6, characterized in that, From all initial solutions, obtain candidate solutions that satisfy the candidate conditions, including: From all initial solutions, select the n1 target initial solutions with the highest probability of occurrence; The current subgraph is weighted using the target initial solution to obtain the initial weight value corresponding to each target initial solution, where n1≥2; Obtain the top n2 target initial weight values ​​from all initial weight values, and use the target initial solutions corresponding to the target initial weight values ​​as candidate solutions, where n2≥1.

8. The multi-user selection method as described in claim 1, characterized in that, The step of constructing a global undirected graph based on each subgraph and its corresponding candidate solutions includes: An initial undirected graph is constructed based on each subgraph and its corresponding candidate solutions, and the initial comprehensive weight value solution of the initial undirected graph is calculated. The target nodes that meet the node flipping conditions are flipped sequentially, and the initial undirected graph is updated based on the flipped nodes. Calculate the reference composite weight solution for the updated initial undirected graph; If the reference comprehensive weight value solution is greater than the initial comprehensive weight value solution, the next target node is flipped until all target nodes have been polled, and then the updated global undirected graph is output.

9. The multi-user selection method as described in claim 8, characterized in that, The step of sequentially flipping the target nodes that meet the node flipping conditions includes: Calculate the importance value of each node in the initial undirected graph; Sort each node according to its importance value from highest to lowest, and then flip the target nodes that meet the node flipping conditions in the order of sorting.

10. The multi-user selection method as described in claim 8, characterized in that, The node flipping condition is: the new solution after each node flips is among the candidate solutions of the subgraph corresponding to that node.

11. The multi-user selection method as described in claim 8, characterized in that, The method further includes: If the reference comprehensive weight value solution is less than or equal to the initial comprehensive weight value solution, the flipping of the current target node is abandoned, and the next target node is flipped.

12. The multi-user selection method as described in claim 1, characterized in that, The step of outputting a global solution for multi-user selection based on the global undirected graph includes: The global undirected graph is iterated until a preset convergence condition is met, at which point a global solution selected by multiple users is output.

13. A multi-user selection device, characterized in that, include: The weighted undirected graph construction module is used to construct a weighted undirected graph based on the acquired user information; wherein, each node of the weighted undirected graph represents a user, the weight value of the node represents the comprehensive weight of each user, and the edges between nodes represent the interference relationship between users; An undirected graph segmentation module is used to segment the weighted undirected graph according to a preset node upper limit value to obtain at least two subgraphs; The candidate solution output module is used to perform quantum algorithm calculations on each subgraph to obtain at least one candidate solution corresponding to each subgraph. The global solution output module is used to construct a global undirected graph based on each subgraph and its corresponding candidate solutions, and output a global solution selected by multiple users based on the global undirected graph.

14. A multi-user selection device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the multi-user selection method as described in any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the multi-user selection method as described in any one of claims 1 to 12.

16. A computer program product, characterized in that, Includes a computer program / instruction that, when executed by a processor, implements the multi-user selection method as described in any one of claims 1 to 12.

Images