Multi-base station coordinated beam resource allocation method and system for dynamic user access

Abstract

The application provides a multi-base station cooperative beam resource allocation method and system facing dynamic user access, and belongs to the technical field of wireless communication resource allocation, and comprises the following steps: acquiring state and constraint information at a current allocation moment, including system global state information, local observation information of each base station and communication resource constraint information; according to the state and constraint information, a hierarchical communication resource allocation framework is adopted, a high-dimensional combination action of user association + communication resource quantity is decomposed into two stages of cooperative service intention generation of a cooperative graph generation layer and specific beam resource allocation action generation of a constraint action generation layer; the specific beam resource allocation action is issued to a multi-base station cooperative communication system for execution; the state and constraint information at a next allocation moment are updated according to the execution result, and a rolling closed-loop resource allocation process is formed. The application can efficiently perform multi-base station cooperative beam resource allocation.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication resource allocation technology, and specifically refers to a multi-base station cooperative beam resource allocation method and system for dynamic user access. Background Technology

[0002] In 5G / 6G cellular networks, ultra-dense networks, and integrated air-space-ground access networks, the number of user terminals continues to increase, and terminal locations, service demands, channel states, and interference environments change dynamically over time. Multiple base stations need to coordinate services for dynamically accessing user terminals under constraints of limited physical resource blocks, beam count, transmit power, and base station load, in order to improve user transmission rates, link reliability, and network resource utilization efficiency.

[0003] Existing communication resource allocation methods typically require the simultaneous determination of multiple discrete or continuous variables, such as user associations, beam resources, physical resource blocks, and transmit power. As the number of candidate user terminals, base stations, and resource levels increases, the joint action space expands rapidly, leading to high allocation complexity and low policy search efficiency, making it difficult to meet the real-time allocation requirements in dynamic user access scenarios.

[0004] On the other hand, while multi-base station cooperative transmission or joint beamforming can improve the quality of service for edge users, it also introduces more complex resource coordination problems. If multiple base stations repeatedly allocate too many resources to the same user terminal, it can easily lead to a waste of physical resource blocks, beam resources, and power resources; if the cooperative service is insufficient, the signal-to-interference-plus-noise ratio or quality of service of the user terminal may fail to meet the requirements. Therefore, a resource allocation method is needed that can reduce the spatial complexity of multi-base station cooperative beam resource allocation and explicitly satisfy the hard constraints of communication resources. Summary of the Invention

[0005] To address the technical problems existing in the prior art, the present invention provides a multi-base station cooperative beam resource allocation method and system for dynamic user access, the technical solution of which is as follows: On the one hand, a multi-base station cooperative beam resource allocation method for dynamic user access is provided, the method comprising: S1. Obtain the status and constraint information at the current allocation time, wherein the status and constraint information includes system global status information, local observation information of each base station, and communication resource constraint information; S2. Based on the state and constraint information, the high-dimensional combination action of user association and communication resource quantity is decomposed into two stages: collaborative intent generation layer for collaborative service intent generation and constraint action generation layer for communication resource constraint solving. The collaborative intent generation layer transforms the high-dimensional, multi-discrete action space in the original beam resource allocation action, which simultaneously includes user association selection and the determination of the number of communication resources, into a low-dimensional collaborative service intent space, and generates collaborative service intents for each base station for candidate user terminals. The constraint action generation layer generates specific beam resource allocation actions that satisfy hard constraint conditions based on the collaborative service intent. S3. Send the specific beam resource allocation action to the multi-base station cooperative communication system for execution; S4. Update the status and constraint information for the next allocation time based on the execution results to form a rolling closed-loop resource allocation process.

[0006] Optionally, the system global state information is used to characterize the overall state of multiple base stations, candidate user terminals, and communication links; the local observation information is used to characterize the user terminal state, channel state, base station load, and remaining communication resource state within the observable range of a single base station; and the constraint state information is used to characterize at least one of the following: number of physical resource blocks, number of beams, transmit power budget, signal-to-interference-plus-noise ratio threshold, base station load, and user quality of service requirements.

[0007] Optionally, the collaborative intent generation layer is implemented using a multi-agent reinforcement learning model based on centralized training and distributed execution, wherein the multi-agent reinforcement learning model employs a distributed policy network and a centralized evaluation network. The distributed policy network is deployed in the agents corresponding to each base station and is used in both the training and execution phases to output collaborative service intentions based on the local observation information of each base station. The centralized evaluation network is used only during the training phase to calculate the value function based on the system's global state information, the cooperative service intentions output by each base station, and communication quality feedback. The value function This is used to evaluate the expected long-term benefits of collaborative service intentions in the current global state, and to reward based on actual communication feedback and the value function. The advantage function is calculated as follows: The advantage function Used to measure the gain of the current collaborative service intent relative to the average level of the current policy, and used to update the distributed policy network parameters. This represents the cumulative reward received from the current moment.

[0008] Optionally, the processing procedure of the distributed policy network includes: For the i-th base station, at the current allocation time t, first select candidate service objects from all candidate user terminals according to the channel quality, service priority, mobility status, QoS requirements and base station load status of the candidate user terminals within its coverage area, and form a candidate service set. Subsequently, a base station-user feature vector is constructed for each candidate service object. The base station-user feature vector includes at least one of the following: channel state, signal-to-interference-plus-noise ratio, service requirements, service priority, minimum rate requirements, number of remaining physical resource blocks (PRBs) of the i-th base station, remaining transmit power, current load, and historical service feedback. Each layer of the policy network calculates the base station-user feature vector for each base station, outputs the service probability of the candidate user terminal corresponding to the i-th base station participating in the service, and forms a service probability vector that corresponds one-to-one with the set of candidate service objects. Subsequently, the service probability vector is binarized according to a preset decision threshold to obtain a collaborative service intent vector. Each element in the collaborative service intent vector corresponds to a candidate service object, used to indicate whether the i-th base station participates in serving the candidate service object, expressed by the formula:

[0009] in This represents the probability that the i-th base station participates in serving the j-th candidate service object. This represents the policy network corresponding to the i-th base station, and δ represents the preset decision threshold. This indicates that the i-th base station participates in serving the j-th candidate service object. =0 indicates that the service will not be provided.

[0010] Optionally, the processing of the constraint action generation layer includes: First, the collaborative service intent output by the collaborative intent generation layer is received, and the collaborative service intent is used as a resource allocation mask: for base station-user terminal combinations with the intent "do not participate in service", the constraint action generation layer does not allocate communication resources; for base station-user terminal combinations with the intent "participate in service", the constraint action generation layer adds them to the set of terminals to be allocated. Subsequently, the constraint action generation layer constructs an optimization problem based on the set of terminals to be allocated, the number of available physical resource blocks (PRBs), the number of beams, the transmit power budget, the base station load, the SINR threshold, and the user QoS requirements. Subsequently, the constraint action generation layer solves the optimization problem under the condition of satisfying the hard constraints of communication resources, obtains the specific resource allocation quantity of each base station for candidate user terminals, and outputs at least one of the following: user association result, PRB allocation result, beam resource allocation result, and transmit power allocation result.

[0011] Optionally, the constraint action generation layer aims to maximize the system's weighted throughput and generates specific beam resource allocation actions under constraints such as physical resource blocks, transmit power, number of service channels, signal-to-interference-plus-noise ratio, and user quality of service. This indicates whether base station i serves candidate user terminal j at the t-th allocation time. =1 indicates participation in the service. =0 indicates that you will not participate in the service; This represents the service priority weight of the j-th user terminal; Let represent the expected transmission rate when the i-th base station serves the j-th user terminal. Then, the constraint action generation layer constructs the following optimization problem: ; The optimization problem satisfies the following constraints: (1) The total number of physical resource blocks allocated to each base station shall not exceed the number of its available physical resource blocks: ,in This represents the number of physical resource blocks allocated by the i-th base station to the j-th candidate user terminal. This represents the total number of physical resource blocks available to the i-th base station at the t-th allocation time; (2) The total transmit power of each base station shall not exceed the power budget: , This represents the transmit power allocated by the i-th base station to the j-th candidate user terminal. This represents the transmit power budget for the i-th base station; (3) The number of candidate user terminals served by each base station at the same time shall not exceed the maximum number of service channels: , This indicates the maximum number of candidate user terminals that a single base station can serve simultaneously. (4) Only base station-user terminal combinations whose cooperative service intent is to participate in the service can be allocated physical resource blocks and transmit power: .

[0012] Optionally, the constraint action generation layer is solved using an iterative greedy allocation method: In each round, the base station-user combination with the highest unit resource revenue and meeting the constraints is selected, and the corresponding physical resource blocks, transmit power and beam resources are allocated, and the remaining resources are updated. If a combination results in the failure to meet resource budget, signal-to-interference-plus-noise ratio threshold, or minimum rate requirements, then skip that combination or reduce its resource allocation level. Repeat the above process until no feasible combination exists or base station resources are exhausted, and finally obtain the specific resource allocation action that satisfies the hard constraints.

[0013] On the other hand, a multi-base station cooperative beam resource allocation system for dynamic user access is provided, the system comprising: The acquisition module is used to acquire the status and constraint information at the current allocation time. The status and constraint information includes global system status information, local observation information of each base station, and communication resource constraint information. The decomposition module is used to decompose the high-dimensional combination of user association and communication resource quantity into two stages: collaborative service intent generation at the collaborative intent generation layer and specific beam resource allocation action generation at the constraint action generation layer, based on the state and constraint information and using a hierarchical communication resource allocation framework. The collaborative intent generation layer transforms the high-dimensional, multi-discrete action space in the original beam resource allocation action, which simultaneously includes user association selection and the determination of the number of communication resources, into a low-dimensional collaborative service intent space, and generates collaborative service intents for each base station for candidate user terminals. The constraint action generation layer generates specific beam resource allocation actions that satisfy hard constraint conditions based on the collaborative service intent. The execution module is used to send the specific beam resource allocation action to the multi-base station cooperative communication system for execution; The update module is used to update the status and constraint information at the next allocation time based on the execution results, forming a rolling closed-loop resource allocation process.

[0014] On the other hand, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described multi-base station cooperative beam resource allocation method for dynamic user access.

[0015] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement the above-described multi-base station cooperative beam resource allocation method for dynamic user access.

[0016] The beneficial effects of the technical solution provided by this invention include at least the following: First, the present invention outputs the intention of each base station to participate in the collaborative service through the collaborative intention generation layer, transforming the high-dimensional multi-discrete communication resource allocation action into a low-dimensional collaborative service intention, thereby reducing the strategy search dimension and improving allocation efficiency.

[0017] Second, this invention improves the executability of output beam resource allocation actions by explicitly processing hard constraints such as the number of physical resource blocks, the number of beams, the transmit power budget, the signal-to-interference-plus-noise ratio threshold, and the user quality of service requirements through the constraint action generation layer.

[0018] Third, the present invention divides the work of the collaborative intent generation layer and the constraint action generation layer, so that the collaborative intent generation layer is responsible for judging the user service participation relationship, and the constraint action generation layer is responsible for the specific allocation of communication resources, thereby reducing the duplication of services by multiple base stations and the consumption of ineffective resources.

[0019] Fourth, by performing feedback updates of state and constraint information, this invention forms a rolling closed-loop allocation process oriented towards dynamic user access and channel changes, which is applicable to scenarios such as 5G / 6G multi-base station collaborative transmission, ultra-dense networks, and integrated air-space-ground access networks. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a multi-base station cooperative beam resource allocation method for dynamic user access provided by an embodiment of the present invention; Figure 2 This is a general block diagram of a multi-base station cooperative beam resource allocation method for dynamic user access provided in an embodiment of the present invention; Figure 3 This is a block diagram of a multi-base station cooperative beam resource allocation system for dynamic user access provided by an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0022] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0023] A multi-base station cooperative communication system includes multiple base stations (or distributed access nodes) and multiple dynamically accessing candidate user terminals. The base stations act as decision nodes, and the candidate user terminals are the objects of resource allocation. Communication resources include physical resource blocks, beam resources, transmit power, time-frequency resources, or combinations thereof.

[0024] In the original resource allocation method, the original joint action space (high-dimensional multi-discrete action space) needs to simultaneously determine: 1. Which user terminal (UE) is being served; 2. How many Physical Resource Blocks (PRBs) are allocated; 3. How many beam resources are allocated; 4. How many power levels are allocated. The action of the i-th base station at time t needs to simultaneously determine the number of communication resources allocated to multiple candidate user terminals, which can be expressed as: ,in, Indicates the number of candidate user terminals. This indicates the maximum level of communication resources that can be allocated to a single candidate user terminal. This represents the number of communication resources allocated by the i-th base station to the j-th candidate user terminal at time t. The number of communication resources can be the number of physical resource blocks, the number of beam resources, power levels, or time-frequency resource blocks. Therefore, when the number of candidate user terminals Ne is large, the original action space increases rapidly, leading to increased strategy search complexity.

[0025] To reduce the search complexity of the original high-dimensional multi-discrete action space, this invention provides a multi-base station cooperative beam resource allocation method for dynamic user access. This method can be implemented by an electronic device, which can be a terminal or a server. Figure 1 The flowchart of this method is shown below. Figure 2 The diagram shown is an overall block diagram of the method. The processing flow may include the following steps: S1. Obtain the status and constraint information at the current allocation time, wherein the status and constraint information includes system global status information, local observation information of each base station, and communication resource constraint information; Optionally, the system global state information is used to characterize the overall state of multiple base stations, candidate user terminals, and communication links; the local observation information is used to characterize the user terminal state, channel state, base station load, and remaining communication resource state within the observable range of a single base station; and the constraint state information is used to characterize at least one of the following: number of physical resource blocks, number of beams, transmit power budget, signal-to-interference-plus-noise ratio threshold, base station load, and user quality of service requirements.

[0026] S2. Based on the state and constraint information, the high-dimensional combination action of user association and communication resource quantity is decomposed into two stages: collaborative intent generation layer for collaborative service intent generation and constraint action generation layer for communication resource constraint solving. The collaborative intent generation layer transforms the high-dimensional, multi-discrete action space in the original beam resource allocation action, which simultaneously includes user association selection and the determination of the number of communication resources, into a low-dimensional collaborative service intent space, and generates collaborative service intents for each base station for candidate user terminals. The constraint action generation layer generates specific beam resource allocation actions that satisfy hard constraint conditions based on the collaborative service intent. Optionally, the collaborative intent generation layer is implemented using a multi-agent reinforcement learning model based on centralized training and distributed execution, wherein the multi-agent reinforcement learning model employs a distributed policy network and a centralized evaluation network. The distributed policy network is deployed in the agents corresponding to each base station and is used in both the training and execution phases to output collaborative service intentions based on the local observation information of each base station. The centralized evaluation network is used only during the training phase to calculate the value function based on the system's global state information, the collaborative service intentions output by each base station, and communication quality feedback (which includes at least one of user throughput, SINR compliance, QoS satisfaction, resource utilization, and load balancing). The value function This is used to evaluate the expected long-term benefits of collaborative service intentions in the current global state, and to reward based on actual communication feedback and the value function. The advantage function is calculated as follows: The advantage function Used to measure the gain of the current collaborative service intent relative to the average level of the current policy, and used to update the distributed policy network parameters. This represents the cumulative reward received from the current moment.

[0027] Optionally, the processing procedure of the distributed policy network includes: For the i-th base station, at the current allocation time t, first select candidate service objects from all candidate user terminals according to the channel quality, service priority, mobility status, QoS requirements and base station load status of the candidate user terminals within its coverage area, and form a candidate service set. Optionally, the filtering rules include: Coverage filtering: Remove user terminals that are not within the coverage area of ​​the i-th base station or whose channel quality with the i-th base station is lower than a preset threshold; Base station load filtering: If the current load of the i-th base station exceeds the preset load threshold, the priority of that base station in participating in new user services will be reduced. User demand screening: Based on the service priority and rate requirements of candidate user terminals, retain user terminals with high service urgency or high synergistic gain. Then, a sorting and truncation process is performed: candidate scores are calculated for the filtered user terminals, and the top k scores are selected. c Each user terminal forms a candidate service object set for the i-th base station.

[0028] Subsequently, a base station-user feature vector is constructed for each candidate service object. The base station-user feature vector includes at least one of the following: channel state, signal-to-interference-plus-noise ratio, service requirements, service priority, minimum rate requirements, number of remaining physical resource blocks (PRBs) of the i-th base station, remaining transmit power, current load, and historical service feedback. Each layer of the policy network calculates the base station-user feature vector for each base station, outputs the service probability of the candidate user terminal corresponding to the i-th base station participating in the service, and forms a service probability vector that corresponds one-to-one with the set of candidate service objects. Subsequently, the service probability vector is binarized according to a preset decision threshold to obtain a cooperative service intent vector. Each element in the cooperative service intent vector corresponds to a candidate service object, used to indicate whether the i-th base station participates in serving the candidate service object, expressed by the formula:

[0029] in This represents the probability that the i-th base station participates in serving the j-th candidate service object. This represents the policy network corresponding to the i-th base station, and δ represents the preset decision threshold. This indicates that the i-th base station participates in serving the j-th candidate service object. =0 indicates that the service will not be provided.

[0030] Optionally, the processing of the constraint action generation layer includes: First, the collaborative service intent output by the collaborative intent generation layer is received, and the collaborative service intent is used as a resource allocation mask: for base station-user terminal combinations with the intent "do not participate in service", the constraint action generation layer does not allocate communication resources; for base station-user terminal combinations with the intent "participate in service", the constraint action generation layer adds them to the set of terminals to be allocated. Subsequently, the constraint action generation layer constructs an optimization problem based on the set of terminals to be allocated, the number of available physical resource blocks (PRBs), the number of beams, the transmit power budget, the base station load, the SINR threshold, and the user QoS requirements. Subsequently, the constraint action generation layer solves the optimization problem under the condition of satisfying the hard constraints of communication resources, obtains the specific resource allocation quantity of each base station for candidate user terminals, and outputs at least one of the following: user association result, PRB allocation result, beam resource allocation result, and transmit power allocation result.

[0031] Optionally, the constraint action generation layer aims to maximize the system's weighted throughput and generates specific beam resource allocation actions under constraints such as physical resource blocks, transmit power, number of service channels, signal-to-interference-plus-noise ratio, and user quality of service. This indicates whether base station i serves candidate user terminal j at the t-th allocation time. =1 indicates participation in the service. =0 indicates that you will not participate in the service; This represents the service priority weight of the j-th user terminal; Let represent the expected transmission rate when the i-th base station serves the j-th user terminal. Then, the constraint action generation layer constructs the following optimization problem: ; The optimization problem satisfies the following constraints: (1) The total number of physical resource blocks allocated to each base station shall not exceed the number of its available physical resource blocks: ,in This represents the number of physical resource blocks allocated by the i-th base station to the j-th candidate user terminal. This represents the total number of physical resource blocks available to the i-th base station at the t-th allocation time; (2) The total transmit power of each base station shall not exceed the power budget: , This represents the transmit power allocated by the i-th base station to the j-th candidate user terminal. This represents the transmit power budget for the i-th base station; (3) The number of candidate user terminals served by each base station at the same time shall not exceed the maximum number of service channels: , This indicates the maximum number of candidate user terminals that a single base station can serve simultaneously. (4) Only base station-user terminal combinations whose cooperative service intent is to participate in the service can be allocated physical resource blocks and transmit power: .

[0032] Optionally, the constraint action generation layer is solved using an iterative greedy allocation method: In each round, the base station-user combination with the highest unit resource revenue and meeting the constraints is selected, and the corresponding physical resource blocks, transmit power and beam resources are allocated, and the remaining resources are updated. If a combination results in the failure to meet resource budget, signal-to-interference-plus-noise ratio threshold, or minimum rate requirements, then skip that combination or reduce its resource allocation level. Repeat the above process until no feasible combination exists or base station resources are exhausted, and finally obtain the specific resource allocation action that satisfies the hard constraints.

[0033] S3. Send the specific beam resource allocation action to the multi-base station cooperative communication system for execution; S4. Update the status and constraint information for the next allocation time based on the execution results to form a rolling closed-loop resource allocation process.

[0034] like Figure 3 As shown, this embodiment of the invention also provides a multi-base station cooperative beam resource allocation system for dynamic user access, the system comprising: The acquisition module 310 is used to acquire the status and constraint information at the current allocation time. The status and constraint information includes the system global status information, the local observation information of each base station, and the communication resource constraint information. The decomposition module 320 is used to decompose the high-dimensional combination action of user association + communication resource quantity into two stages: collaborative service intent generation in the collaborative intent generation layer and specific beam resource allocation action generation in the constraint action generation layer, based on the state and constraint information and using a hierarchical communication resource allocation framework. The collaborative intent generation layer transforms the high-dimensional, multi-discrete action space in the original beam resource allocation action, which simultaneously includes user association selection and the determination of the number of communication resources, into a low-dimensional collaborative service intent space, and generates collaborative service intents for each base station for candidate user terminals. The constraint action generation layer generates specific beam resource allocation actions that satisfy hard constraint conditions based on the collaborative service intent. Execution module 330 is used to send the specific beam resource allocation action to the multi-base station cooperative communication system for execution; The update module 340 is used to update the status and constraint information at the next allocation time based on the execution result, forming a rolling closed-loop resource allocation process.

[0035] The multi-base station cooperative beam resource allocation system for dynamic user access provided in this embodiment of the invention has a functional structure that corresponds to the multi-base station cooperative beam resource allocation method for dynamic user access provided in this embodiment of the invention, and will not be described again here.

[0036] Figure 4 This is a schematic diagram of the structure of an electronic device 400 provided in an embodiment of the present invention. The electronic device 400 may vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 401 and one or more memories 402. The memory 402 stores at least one instruction, which is loaded and executed by the processor 401 to implement the steps of the multi-base station cooperative beam resource allocation method for dynamic user access described above.

[0037] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in a terminal to complete the aforementioned multi-base station cooperative beam resource allocation method for dynamic user access. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device.

[0038] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-base station cooperative beam resource allocation for dynamic user access, characterized in that, The method includes: S1. Obtain the status and constraint information at the current allocation time, wherein the status and constraint information includes system global status information, local observation information of each base station, and communication resource constraint information; S2. Based on the state and constraint information, the high-dimensional combination action of user association and communication resource quantity is decomposed into two stages: collaborative intent generation layer for collaborative service intent generation and constraint action generation layer for communication resource constraint solving. The collaborative intent generation layer transforms the high-dimensional, multi-discrete action space in the original beam resource allocation action, which simultaneously includes user association selection and the determination of the number of communication resources, into a low-dimensional collaborative service intent space, and generates collaborative service intents for each base station for candidate user terminals. The constraint action generation layer generates specific beam resource allocation actions that satisfy hard constraint conditions based on the collaborative service intent. S3. Send the specific beam resource allocation action to the multi-base station cooperative communication system for execution; S4. Update the status and constraint information for the next allocation time based on the execution results to form a rolling closed-loop resource allocation process.

2. The method according to claim 1, characterized in that, The system global state information is used to characterize the overall state of multiple base stations, candidate user terminals, and communication links. The local observation information is used to characterize the user terminal state, channel state, base station load, and remaining communication resource state within the observable range of a single base station. The constraint state information is used to characterize at least one of the following: number of physical resource blocks, number of beams, transmit power budget, signal-to-interference-plus-noise ratio threshold, base station load, and user quality of service requirements.

3. The method according to claim 1, characterized in that, The collaborative intent generation layer is implemented using a multi-agent reinforcement learning model based on centralized training and distributed execution. The multi-agent reinforcement learning model employs a distributed policy network and a centralized evaluation network. The distributed policy network is deployed in the agents corresponding to each base station and is used in both the training and execution phases to output collaborative service intentions based on the local observation information of each base station. The centralized evaluation network is used only during the training phase to calculate the value function based on the system's global state information, the cooperative service intentions output by each base station, and communication quality feedback. The value function This is used to evaluate the expected long-term benefits of collaborative service intentions in the current global state, and to reward based on actual communication feedback and the value function. The advantage function is calculated as follows: The advantage function Used to measure the gain of the current collaborative service intent relative to the average level of the current policy, and used to update the distributed policy network parameters. This represents the cumulative reward received from the current moment.

4. The method according to claim 3, characterized in that, The processing procedure of the distributed policy network includes: For the i-th base station, at the current allocation time t, first select candidate service objects from all candidate user terminals according to the channel quality, service priority, mobility status, QoS requirements and base station load status of the candidate user terminals within its coverage area, and form a candidate service set. Subsequently, a base station-user feature vector is constructed for each candidate service object. The base station-user feature vector includes at least one of the following: channel state, signal-to-interference-plus-noise ratio, service requirements, service priority, minimum rate requirements, number of remaining physical resource blocks (PRBs) of the i-th base station, remaining transmit power, current load, and historical service feedback. Each layer of the policy network calculates the base station-user feature vector for each base station, outputs the service probability of the candidate user terminal corresponding to the i-th base station participating in the service, and forms a service probability vector that corresponds one-to-one with the set of candidate service objects. Subsequently, the service probability vector is binarized according to a preset decision threshold to obtain a collaborative service intent vector. Each element in the collaborative service intent vector corresponds to a candidate service object, used to indicate whether the i-th base station participates in serving the candidate service object, expressed by the formula: in This represents the probability that the i-th base station participates in serving the j-th candidate service object. This represents the policy network corresponding to the i-th base station, and δ represents the preset decision threshold. This indicates that the i-th base station participates in serving the j-th candidate service object. =0 indicates that the service is not provided.

5. The method according to claim 1, characterized in that, The processing procedure of the constraint action generation layer includes: First, the collaborative service intent output by the collaborative intent generation layer is received, and the collaborative service intent is used as a resource allocation mask: for base station-user terminal combinations with the intent "do not participate in service", the constraint action generation layer does not allocate communication resources; for base station-user terminal combinations with the intent "participate in service", the constraint action generation layer adds them to the set of terminals to be allocated. Subsequently, the constraint action generation layer constructs an optimization problem based on the set of terminals to be allocated, the number of available physical resource blocks (PRBs), the number of beams, the transmit power budget, the base station load, the SINR threshold, and the user QoS requirements. Subsequently, the constraint action generation layer solves the optimization problem under the condition of satisfying the hard constraints of communication resources, obtains the specific resource allocation quantity of each base station for candidate user terminals, and outputs at least one of the following: user association result, PRB allocation result, beam resource allocation result, and transmit power allocation result.

6. The method according to claim 5, characterized in that, The constraint action generation layer aims to maximize the system's weighted throughput and generates specific beam resource allocation actions under constraints such as physical resource blocks, transmit power, number of service channels, signal-to-interference-plus-noise ratio, and user quality of service. This indicates whether base station i serves candidate user terminal j at the t-th allocation time. =1 indicates participation in the service. =0 indicates that you will not participate in the service; This represents the service priority weight of the j-th user terminal; Let represent the expected transmission rate when the i-th base station serves the j-th user terminal. Then, the constraint action generation layer constructs the following optimization problem: ; The optimization problem satisfies the following constraints: (1) The total number of physical resource blocks allocated to each base station shall not exceed the number of its available physical resource blocks: ,in This represents the number of physical resource blocks allocated by the i-th base station to the j-th candidate user terminal. This represents the total number of physical resource blocks available to the i-th base station at the t-th allocation time; (2) The total transmit power of each base station shall not exceed the power budget: , This represents the transmit power allocated by the i-th base station to the j-th candidate user terminal. This represents the transmit power budget for the i-th base station; (3) The number of candidate user terminals served by each base station at the same time shall not exceed the maximum number of service channels: , This indicates the maximum number of candidate user terminals that a single base station can serve simultaneously. (4) Only base station-user terminal combinations whose cooperative service intent is to participate in the service can be allocated physical resource blocks and transmit power: .

7. The method according to claim 6, characterized in that, The constraint action generation layer is solved using an iterative greedy allocation method: In each round, the base station-user combination with the highest unit resource revenue and meeting the constraints is selected, and the corresponding physical resource blocks, transmit power and beam resources are allocated, and the remaining resources are updated. If a combination results in the failure to meet resource budget, signal-to-interference-plus-noise ratio threshold, or minimum rate requirements, then skip that combination or reduce its resource allocation level. Repeat the above process until no feasible combination exists or base station resources are exhausted, and finally obtain the specific resource allocation action that satisfies the hard constraints.

8. A multi-base station cooperative beam resource allocation system for dynamic user access, characterized in that, The system includes: The acquisition module is used to acquire the status and constraint information at the current allocation time. The status and constraint information includes global system status information, local observation information of each base station, and communication resource constraint information. The decomposition module is used to decompose the high-dimensional combination of user association and communication resource quantity into two stages: collaborative service intent generation at the collaborative intent generation layer and specific beam resource allocation action generation at the constraint action generation layer, based on the state and constraint information and using a hierarchical communication resource allocation framework. The collaborative intent generation layer transforms the high-dimensional, multi-discrete action space in the original beam resource allocation action, which simultaneously includes user association selection and the determination of the number of communication resources, into a low-dimensional collaborative service intent space, and generates collaborative service intents for each base station for candidate user terminals. The constraint action generation layer generates specific beam resource allocation actions that satisfy hard constraint conditions based on the collaborative service intent. The execution module is used to send the specific beam resource allocation action to the multi-base station cooperative communication system for execution; The update module is used to update the status and constraint information at the next allocation time based on the execution results, forming a rolling closed-loop resource allocation process.

9. An electronic device comprising a processor and a memory, wherein the memory stores at least one instruction, characterized in that, The processor loads and executes at least one instruction to implement the multi-base station cooperative beam resource allocation method for dynamic user access as described in any one of claims 1-7.

10. A computer-readable storage medium storing at least one instruction, characterized in that, The at least one instruction is loaded and executed by the processor to implement the multi-base station cooperative beam resource allocation method for dynamic user access as described in any one of claims 1-7.

Images