Wireless communication resource allocation method and device based on distributed computing, equipment and storage medium

Abstract

The invention discloses a wireless communication resource allocation method based on distributed computing, and the method comprises the steps: S1, carrying out the modeling of an edge end agent distributed decision problem, and obtaining a wireless communication resource allocation decision model; s2, on the basis of an attention mechanism local observation feature fusion coding model, local observation related to a global communication state is extracted, global communication state features are constructed, and indexes for global target optimization are obtained; s3, training the wireless communication resource allocation decision model through centralized training; and S4, running the trained wireless communication resource allocation decision model on all communication terminal agents in a distributed manner, and jointly deciding allocation of wireless communication network resources. The invention also discloses a wireless communication resource allocation device based on distributed computing, corresponding equipment and a storage medium. According to the method, the wireless spectrum utilization rate is improved, and the reliability and effectiveness of wireless transmission are both considered.

Description

Technical Field

[0001] The present application relates to the technical field of wireless resource allocation, and more specifically, to a wireless communication resource allocation method, apparatus, device and storage medium based on distributed computing. Background Art

[0002] The purpose of radio resource control is to improve the effective use of resources and expand the capacity of the communication system on the one hand, and to improve system reliability and ensure communication QoS performance on the other hand. However, reliability and effectiveness are generally not taken into account at the same time. It is difficult to ensure the effectiveness of transmission if high reliability is required, and vice versa. Various technologies such as radio resource management are designed to maximize the utilization of wireless spectrum and achieve a balance between reliability and effectiveness when meeting the different QoS requirements of various services.

[0003] Reinforcement learning is a machine learning method that can learn the optimal strategy through the interaction between the agent and the environment. In reinforcement learning, the agent learns how to take the best action in different states by observing the state of the environment, taking actions and obtaining rewards. Reinforcement learning has been widely used in many fields, such as games, robot control, natural language processing, etc., and can be used to solve the above-mentioned problem of balancing reliability and effectiveness. However, many practical applications need to deal with the interaction between multiple agents. How to use reinforcement learning to solve the allocation of wireless communication resources in the case of multiple agents is a problem that needs to be solved urgently. Summary of the invention

[0004] In response to at least one defect or improvement need in the prior art, the present invention provides a wireless communication resource allocation method, apparatus, device and storage medium based on distributed computing, which will solve at least one of the problems existing in the above-mentioned background technology.

[0005] To achieve the above object, according to a first aspect of the present invention, a method for allocating wireless communication resources based on distributed computing is provided, the method comprising:

[0006] S1 models the distributed decision-making problem of edge agents and obtains the wireless communication resource allocation decision model;

[0007] S2 is based on the attention mechanism local observation feature fusion encoding model, extracts local observations related to the global communication state, constructs global communication state features, and obtains indicators for global goal optimization;

[0008] S3 trains the wireless communication resource allocation decision model through centralized training;

[0009] S4 distributes and runs the trained wireless communication resource allocation decision model on all communication terminal agents to jointly decide on the allocation of wireless communication network resources.

[0010] Furthermore, in the above-mentioned wireless communication resource allocation method based on distributed computing, the modeling of the distributed decision-making problem of the edge agent specifically includes:

[0011] The unmanned autonomous terminals are divided into two groups. The terminal group that currently occupies communication resources forms a set A, which is denoted as A = {a1, a2, …, a N}, the terminal group that is not currently occupying communication resources forms a set B, denoted as B = {b1, b2, …, b M};

[0012] Each terminal in set B is regarded as an intelligent agent. By establishing a transmission rate constraint, it is ensured that a payload of size ξ is successfully transmitted within a given time. The problem of multi-terminal competition for wireless communication resources is transformed into an optimization problem under constraints.

[0013]

[0014] Among them, τ k [n] is a Boolean spectrum selection variable, indicating whether the kth link in set B transmits on the nth subband in set A. If yes, it is 1, otherwise it is 0; In the first time scale Δ τ The transmission capacity of whether the kth link in the next set B is transmitted on the nth subband in the set A. Pr represents the probability that the terminals in the set B compete for channel resources and obtain the maximum channel capacity under the premise that the set A occupies the channel resources.

[0015] Furthermore, in the above-mentioned wireless communication resource allocation method based on distributed computing, the optimization problem under the constraints is a partially observable Markov decision process, in which multiple intelligent agents interact with the wireless communication resource allocation decision model of the base station center in a distributed manner, make decisions on the observation results based on probability, collect samples into an experience sample library, and use a centralized training optimization strategy function based on the local observation feature fusion coding model based on the attention mechanism.

[0016] Furthermore, in the above-mentioned wireless communication resource allocation method based on distributed computing, the action space of the Markov decision process is spectrum wireless resource selection, channel time slot selection and power control.

[0017] Furthermore, in the wireless communication resource allocation method based on distributed computing, the local observation feature fusion coding model based on the attention mechanism extracts local observations related to the global communication state, specifically including the local observation information (σ1, σ2, …, σ k ), based on the attention mechanism, the state feature encoding of the intelligent agent is extracted to construct the local observation information of the unmanned platform intelligent terminal, the dependency relationship between the local observation information of each intelligent agent is obtained, and the observation information that meets the preset conditions is extracted to construct the global state.

[0018] Furthermore, the above-mentioned wireless communication resource allocation method based on distributed computing, wherein the wireless communication resource allocation decision model is trained through centralized training, specifically including randomly taking multiple selections for any local observation, designing multiple groups of global states from multiple perspectives and encoding them, and using the same set of value functions for training and learning.

[0019] According to a second aspect of the present invention, there is also provided a wireless communication resource allocation device based on distributed computing, comprising:

[0020] The first acquisition module is used to model the distributed decision-making problem of the edge agent and obtain a wireless communication resource allocation decision model;

[0021] The second acquisition module is used to extract local observations related to the global communication state based on the attention mechanism local observation feature fusion encoding model, construct the global communication state feature, and obtain the indicator for optimizing the global goal;

[0022] A training module, used for training a wireless communication resource allocation decision model through centralized training;

[0023] The deployment module is used to distribute the trained wireless communication resource allocation decision model on all communication terminal intelligent agents to jointly decide on the allocation of wireless communication network resources.

[0024] According to the third aspect of the present invention, there is also provided a wireless communication resource allocation device based on distributed computing, which includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit executes the steps of any one of the above methods.

[0025] According to the fourth aspect of the present invention, a storage medium is also provided, which stores a computer program that can be executed by a wireless communication resource allocation device based on distributed computing. When the computer program runs on the wireless communication resource allocation device based on distributed computing, the wireless communication resource allocation device based on distributed computing executes the steps of any one of the above-mentioned methods.

[0026] According to a fifth aspect of the present invention, there is also provided a computer program product, comprising a computer program, characterized in that when the computer program is executed by a processor, the steps of any one of the above-mentioned methods for allocating wireless communication resources based on distributed computing are implemented.

[0027] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0028] The wireless communication resource allocation method based on distributed computing provided by the present invention has its own strategy and value function in multi-agent reinforcement learning, and learns the optimal strategy by observing the state of the environment and the actions of other agents. It can use reinforcement learning to solve the allocation of wireless communication resources in the case of multiple agents, improve the utilization rate of wireless spectrum, and take into account the reliability and effectiveness of wireless transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 A flowchart of a wireless communication resource allocation method based on distributed computing is provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0032] The terms "first", "second", "third", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0033] Figure 1is a flow chart of a wireless communication resource allocation method based on distributed computing provided in an embodiment of the present application, such as Figure 1 As shown, a wireless communication resource allocation method based on distributed computing provided in an embodiment of the present application includes:

[0034] S1 models the distributed decision-making problem of edge agents and obtains the wireless communication resource allocation decision model;

[0035] S2 is based on the attention mechanism local observation feature fusion encoding model, extracts local observations related to the global communication state, constructs global communication state features, and obtains indicators for global goal optimization;

[0036] S3 trains the wireless communication resource allocation decision model through centralized training;

[0037] S4 distributes and runs the trained wireless communication resource allocation decision model on all communication terminal agents to jointly decide on the allocation of wireless communication network resources.

[0038] Specifically, the distributed decision-making problem of edge agents without wireless communication is modeled, and the wireless communication scenario of multiple unmanned platform autonomous terminals in a base station center environment is considered. The unmanned platform autonomous terminal devices perform their respective tasks in the central base station and communicate with each other. The positions of the terminals change dynamically. Considering that the terminal uploads the communication status to the center, which affects the channel communication payload, and in a complex communication environment, the center may not be able to obtain complete information of all unmanned autonomous terminals due to channel noise or interference, it is necessary to consider autonomous distributed negotiation of unmanned terminals and jointly decide on the allocation of communication resources.

[0039] The unmanned autonomous terminals are divided into two groups according to their communication status. The terminal group that currently occupies a little resource forms a set A, which is denoted as A = {a1, a2, …, a N}, the terminal group that needs to be allocated communication resources forms a set B, denoted as B = {b1, b2, ..., b M}.

[0040] When communication resources are tight, in order to improve channel utilization efficiency, terminals in set A can reuse channel resources in set B. Therefore, the goal of communication resource allocation is to design a reasonable resource allocation solution for the terminal set, allocate channel resources, and achieve the required performance requirements with minimal channel overhead.

[0041] Considering each terminal in set B as an intelligent agent, under the premise of maximizing the communication channel capacity, by establishing a transmission rate constraint, it is ensured that the payload of size ξ is successfully transmitted within a given time. The problem of multi-terminal competition for wireless communication resources is transformed into an optimization problem under constraints as follows:

[0042]

[0043] Among them, τ k [n] is a Boolean spectrum selection variable, indicating whether the kth link in set B is transmitted on the nth subband in set A. If yes, it is 1, otherwise it is 0; Indicates that at a certain time scale Δ τ The transmission capacity of whether the kth link in set B is transmitted on the nth subband in set A. Pr represents the probability that the terminals in set B compete for channel resources to obtain the maximum channel capacity under the premise that set A occupies channel resources. The optimization object is τ k [n], that is, the choice of whether the k-th link in set B is transmitted on the n-th subband in set A.

[0044] The above conditional optimization problem is modeled as a partially observable Markov decision process, in which multiple intelligent terminals interact with the wireless communication resource allocation model of the base station center in a distributed manner, try to make decisions on the observation results with a certain probability, collect samples into the experience sample library, and use centralized training to optimize the strategy function based on the attention mechanism model.

[0045] In the above multi-agent reinforcement learning resource allocation problem, each terminal to which resources are to be allocated is regarded as an agent. At time step t, the global state S t All channel states and agent actions need to be considered. However, for each agent, this information is hidden, and agent i can only understand the current environment through the observation function. Given the current environment state S t , the observation function of the environment is O, and the action is a t , the agent receives the next reward R t+1 The environment status is updated to S t+1 , and the new observation space Z t+1 .

[0046] The action space of the Markov decision process can be designed as spectrum wireless resource selection, channel time slot selection and power control. Since the above optimization function is continuous, it means that the action space is infinite. In order to simplify the algorithm process, the action space can be designed as a finite discrete value according to the actual application scenario.

[0047] Optionally, the power can be set to multiple levels based on experience: [25, 15, 5, -5, -15, -50] dBm, and the size of the action space is 6*M.

[0048] Considering the fairness of resource allocation and the overall performance of the system, we define agent a i The reward function is:

[0049]

[0050] That is, the goal is to increase the probability of successful transmission of effective load of the set B terminals within the time period t; where, is the instantaneous capacity of all links of terminals in set B, ρ i [m] represents the channel selection probability of the i-th terminal in set B; γ is a constant greater than or equal to 0, and its value is derived from the maximum measured value of the transmission rate when the power is less than 0 in a real network environment.

[0051] Resource allocation requires scheduling resources across multiple links while considering each agent’s local observations to make decisions. The decision model consists of two parts: a decision network for wireless communication resource allocation that takes local observations as input and outputs corresponding actions; and a policy evaluation network that outputs the corresponding value function based on the layout observations and corresponding actions.

[0052] Distributed agents simply splice together the local observations of multiple agents to make independent decisions, which cannot accurately describe the information that needs to be expressed in the global state, thereby weakening the learning ability of the model's value function. Therefore, a reinforcement learning method based on the attention mechanism is adopted to adaptively fuse multiple local observations using the attention mechanism method to construct global state features; and based on the global features from different perspectives, a corresponding value function update method is proposed to improve the model's global feature encoding ability and the strategy evaluation ability of the global action function.

[0053] The reinforcement learning method based on the attention mechanism specifically includes: given the local observation information (σ1,σ2,…,σ k ), use the multi-head attention mechanism to extract the state feature encoding of the terminal agent to construct the local observation information of the unmanned platform intelligent terminal, analyze the dependency between the local observations of each terminal agent, and extract the most important observation information to construct the global state. Randomly select multiple times for any local observation, design multiple groups of different global states from multiple different perspectives and perform state encoding, and use the same set of value function models for training and learning. Starting from multiple different perspectives, each agent can be regarded as a separate perspective, that is, through the action selection of each agent, observe and sample the feedback of the global state, in order to improve the association ability between multiple agents and the overall state goal,

[0054] The multi-agent reinforcement learning model based on the multi-head attention model adopts centralized training and distributed deployment to realize the wireless communication resource decision-making process of a single terminal agent and the entire network.

[0055] The wireless communication resource allocation method based on distributed computing provided in the embodiment of the present application is that in multi-agent reinforcement learning, each agent has its own strategy and value function, and learns the optimal strategy by observing the state of the environment and the actions of other agents. Reinforcement learning can be used to solve the allocation of wireless communication resources in the case of multiple agents, improve the utilization rate of wireless spectrum, and take into account the reliability and effectiveness of wireless transmission.

[0056] Optionally, the wireless communication resource allocation method based on distributed computing provided in an embodiment of the present application, wherein the modeling of the distributed decision-making problem of the edge agent specifically includes:

[0057] The unmanned autonomous terminals are divided into two groups. The terminal group that currently occupies communication resources forms a set A, which is denoted as A = {a1, a2, …, a N}, the terminal group that is not currently occupying communication resources forms a set B, denoted as B = {b1, b2, …, b M};

[0058] Each terminal in set B is regarded as an intelligent agent. By establishing a transmission rate constraint, it is ensured that a payload of size ξ is successfully transmitted within a given time. The problem of multi-terminal competition for wireless communication resources is transformed into an optimization problem under constraints.

[0059]

[0060] Among them, τ k [n] is a Boolean spectrum selection variable, indicating whether the kth link in set B transmits on the nth subband in set A. If yes, it is 1, otherwise it is 0; In the first time scale Δ τ The transmission capacity of whether the kth link in the next set B is transmitted on the nth subband in the set A. Pr represents the probability that the terminals in the set B compete for channel resources and obtain the maximum channel capacity under the premise that the set A occupies the channel resources.

[0061] Optionally, in the wireless communication resource allocation method based on distributed computing provided in an embodiment of the present application, the optimization problem under the constraints is a partially observable Markov decision process, wherein multiple intelligent agents interact with the wireless communication resource allocation decision model of the base station center in a distributed manner, make decisions on the observation results based on probability, collect samples into an empirical sample library, and use a centralized training optimization strategy function based on the local observation feature fusion coding model based on the attention mechanism.

[0062] Optionally, in the wireless communication resource allocation method based on distributed computing provided by an embodiment of the present application, the action space of the Markov decision process is spectrum wireless resource selection, channel time slot selection and power control.

[0063] Optionally, the wireless communication resource allocation method based on distributed computing provided in the embodiment of the present application, the local observation feature fusion coding model based on the attention mechanism, extracts local observations related to the global communication state, specifically including the local observation information (σ1, σ2, …, σ k ), based on the attention mechanism, the state feature encoding of the intelligent agent is extracted to construct the local observation information of the unmanned platform intelligent terminal, the dependency relationship between the local observation information of each intelligent agent is obtained, and the observation information that meets the preset conditions is extracted to construct the global state.

[0064] Optionally, the wireless communication resource allocation method based on distributed computing provided in an embodiment of the present application, wherein the wireless communication resource allocation decision model is trained through centralized training, specifically including randomly taking multiple selections for any local observation, designing multiple groups of global states from multiple perspectives and encoding them, and using the same set of value functions for training and learning.

[0065] Optionally, in one embodiment, the decision steps for selecting a centralized training distributed implementation deployment are as follows:

[0066] Initialize the parameters θ and experience sample library Γ of the joint network and each agent’s decision network;

[0067] Initialize the environment and obtain the initial state S of the environment 0 and observed V 0 , randomly claim to initialize the joint action A 0 ;

[0068] Update Status S 1 , obtain the observation information V of all agents at the next moment 1 and the initial reward R 1 ; The sample (S 0 ,τ 0 ,A 0 ,R 1 ,S 1 ,τ 1 ) is stored in the experience sample library Γ;

[0069] For each agent participating in the competition for communication resources, the local state observed by the agent is and the action at the previous moment Input into the decision network to obtain the local action value function corresponding to the execution action

[0070] Update Status S t+1 , obtain the observation information V of all agents at the next moment t+1 The joint reward R of the environment feedback t+1, will (S t ,τ t ,A t ,R t+1 ,S t+1 ,τ t+1 ) is stored in the experience sample library Γ;

[0071] Randomly sample data in small batches from the experience sample library to the evaluation network, use the multi-head attention mechanism to minimize the loss function and update the parameters θ.

[0072] The present application also provides a wireless communication resource allocation device based on distributed computing, comprising:

[0073] The first acquisition module is used to model the distributed decision-making problem of the edge agent and obtain a wireless communication resource allocation decision model;

[0074] The second acquisition module is used to extract local observations related to the global communication state based on the attention mechanism local observation feature fusion encoding model, construct the global communication state feature, and obtain the indicator for optimizing the global goal;

[0075] A training module, used for training a wireless communication resource allocation decision model through centralized training;

[0076] The deployment module is used to distribute the trained wireless communication resource allocation decision model on all communication terminal intelligent agents to jointly decide on the allocation of wireless communication network resources.

[0077] The present application also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented. The computer-readable storage medium may include, but is not limited to, any type of disk, including a floppy disk, an optical disk, a DVD, a CD-ROM, a micro drive, and a magneto-optical disk, a ROM, a RAM, an EPROM, an EEPROM, a DRAM, a VRAM, a flash memory device, a magnetic card or an optical card, a nanosystem (including a molecular memory IC), or any type of medium or device suitable for storing instructions and / or data.

[0078] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0079] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0080] In the several embodiments provided in the present application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0081] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0082] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0083] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, disk or optical disk and other media that can store program code.

[0084] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, which can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0085] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

[0086] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] It will be easily understood by those skilled in the art that 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 and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A wireless communication resource allocation method based on distributed computing, characterized in that: include: S1 models the distributed decision-making problem of edge agents and obtains the wireless communication resource allocation decision model; S2 is based on the attention mechanism local observation feature fusion encoding model, extracts local observations related to the global communication state, constructs global communication state features, and obtains indicators for global goal optimization; S3 trains the wireless communication resource allocation decision model through centralized training; S4 distributes and runs the trained wireless communication resource allocation decision model on all communication terminal agents to jointly decide on the allocation of wireless communication network resources.

2. The wireless communication resource allocation method based on distributed computing according to claim 1, characterized in that: The modeling of the distributed decision-making problem of the edge agent specifically includes: The unmanned autonomous terminals are divided into two groups. The terminal group that currently occupies communication resources forms a set A, which is denoted as A = {a1, a2, …, a N }, the terminal group that is not currently occupying communication resources forms a set B, denoted as B = {b1, b2, ..., b M }; Each terminal in set B is regarded as an intelligent agent. By establishing a transmission rate constraint, it is ensured that a payload of size ξ is successfully transmitted within a given time. The problem of multi-terminal competition for wireless communication resources is transformed into an optimization problem under constraints. Among them, τ k [n] is a Boolean spectrum selection variable, indicating whether the kth link in set B transmits on the nth subband in set A. If yes, it is 1, otherwise it is 0; In the first time scale Δ τ The transmission capacity of whether the kth link in the next set B is transmitted on the nth subband in the set A. Pr represents the probability that the terminals in the set B compete for channel resources and obtain the maximum channel capacity under the premise that the set A occupies the channel resources.

3. The wireless communication resource allocation method based on distributed computing according to claim 2, characterized in that: The optimization problem under the constraints is a partially observable Markov decision process, in which multiple intelligent agents interact with the wireless communication resource allocation decision model of the base station center in a distributed manner, make decisions on the observation results based on probability, collect samples into an experience sample library, and use centralized training optimization strategy function based on the local observation feature fusion coding model based on the attention mechanism.

4. The wireless communication resource allocation method based on distributed computing according to claim 3, characterized in that: The action space of the Markov decision process is spectrum radio resource selection, channel time slot selection and power control.

5. The wireless communication resource allocation method based on distributed computing according to claim 4, characterized in that: The attention-based local observation feature fusion coding model extracts local observations related to the global communication state, specifically including the local observation information (σ1, σ2, …, σ k ), based on the attention mechanism, the state feature encoding of the intelligent agent is extracted to construct the local observation information of the unmanned platform intelligent terminal, the dependency relationship between the local observation information of each intelligent agent is obtained, and the observation information that meets the preset conditions is extracted to construct the global state.

6. The wireless communication resource allocation method based on distributed computing according to claim 5, characterized in that: The wireless communication resource allocation decision model is trained through centralized training, specifically including randomly taking multiple selections for any local observation, designing and encoding multiple groups of global states from multiple perspectives, and using the same set of value functions for training and learning.

7. A wireless communication resource allocation device based on distributed computing, characterized in that: include: The first acquisition module is used to model the distributed decision-making problem of the edge agent and obtain a wireless communication resource allocation decision model; The second acquisition module is used to extract local observations related to the global communication state based on the attention mechanism local observation feature fusion encoding model, construct the global communication state feature, and obtain the indicator for optimizing the global goal; A training module, used for training a wireless communication resource allocation decision model through centralized training; The deployment module is used to distribute the trained wireless communication resource allocation decision model on all communication terminal intelligent agents to jointly decide on the allocation of wireless communication network resources.

8. A wireless communication resource allocation device based on distributed computing, characterized in that: The method comprises at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit executes the steps of the method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: It stores a computer program that can be executed by a wireless communication resource allocation device based on distributed computing. When the computer program runs on the wireless communication resource allocation device based on distributed computing, the wireless communication resource allocation device based on distributed computing executes the steps of the method described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the wireless communication resource allocation method based on distributed computing as claimed in any one of claims 1 to 6 are implemented.

Images