Method and electronic device performed by a network node

Abstract

Embodiments of the present disclosure provide a method and an electronic device executed by a network node, relating to the field of artificial intelligence. The method comprises: determining a beamforming (BF) weight based on a first user equipment relationship and a channel estimation result of a reference signal, and processing a first signal based on the BF weight; wherein the first user equipment relationship is constructed based on a resource reuse relationship among multiple user equipments and a channel feature index of each user equipment. Optionally, part of the steps in the above method can be executed using an artificial intelligence model. The scheme can better achieve interference management by processing the first signal based on the BF weight determined based on the first user equipment relationship and the channel estimation result of the reference signal.

Description

Technical Field

[0001] This disclosure relates to the field of wireless communication technology, and more specifically, to a method and electronic device performed by a network node. Background Technology

[0002] To meet the coverage and emerging service demands of future 5G, 5GA, and 6G networks, the number of gNBs and UEs in the network will increase dramatically. Network traffic will grow exponentially. To ensure the transmission needs of a large number of devices, the network will increasingly adopt MIMO technology or emerging network architectures (such as cell-free networks). However, to guarantee the needs of a large number of services, the resource reuse rate within a single cell or between multiple cells will continuously increase, causing severe interference and degrading user experience and QoS. Therefore, effective interference management is crucial for network performance.

[0003] Currently, interference management schemes in communication systems fall into two main categories: traditional and AI / ML-based. However, both types of schemes mainly focus on beamforming (BF), equalization, successive interference cancellation (SIC), and resource coordination.

[0004] Because current interference management solutions still suffer from a large number of unresolved interferences and difficulties in assessing these unresolved interferences, it is necessary to provide a new interference management method. Summary of the Invention

[0005] The purpose of this disclosure is to optimize interference management schemes to improve the effectiveness of interference detection and elimination. To achieve this objective, the technical solutions provided by this disclosure are as follows: According to a first aspect of the present disclosure, a method performed by a network node in a wireless communication network is provided, the method comprising: Based on the channel estimation results of the first user equipment relationship and the reference signal, the beamforming (BF) weights are determined, wherein the first user equipment relationship is constructed based on the resource reuse relationship between multiple user equipments and the channel feature index of each user equipment; The first signal is processed based on the BF weight, wherein the first signal is a signal received from the first user equipment, or a signal sent to the first user equipment.

[0006] In conjunction with some embodiments of the first aspect, in an optional embodiment of this disclosure, the first user equipment relationship includes: reusing the performance of each user equipment and the magnitude of interference when performing resource reuse.

[0007] In conjunction with some embodiments of the first aspect, in an optional embodiment of this disclosure, the method further includes: Send first information including a first resource to the first user equipment, wherein the first resource is a resource allocated by the network device for transmission to the first user equipment; Receive the first signal sent by the first user equipment on the first resource; or, The first signal is sent to the first user equipment on the first resource.

[0008] In conjunction with some embodiments of the first aspect, in an optional embodiment of this disclosure, the method further includes: Based on the first user equipment relationship, determine the first resource allocated to the first user equipment; or... Based on the obtained interference magnitude, estimated interference cancellation parameters, and performance indicators of the first user equipment, the first resource allocated to the first user equipment is determined.

[0009] In conjunction with some embodiments of the first aspect, in an optional embodiment of this disclosure, determining the first resource allocated to the first user equipment based on the first user equipment relationship includes: Based on the resource reuse information in the first user equipment relationship, obtain the performance of the first user equipment when it reuses resources with other user equipment; Based on the performance indicators satisfying the first constraint, the resources allocated for transmission to each user equipment are determined.

[0010] In conjunction with some embodiments of the first aspect, in an optional embodiment of this disclosure, determining the first resource allocated to the first user equipment based on the acquired interference magnitude, estimated interference cancellation parameters, and performance indicators of the first user equipment includes: Based on the magnitude of the interference received by the first user equipment and the interference cancellation parameters, the magnitude of the interference that was not eliminated is determined; Based on the performance indicators of the first user equipment and the fact that the amount of uneliminated interference meets the conditions, the first resource is determined to be allocated to the first user equipment.

[0011] In conjunction with some embodiments of the first aspect, in an optional embodiment of this disclosure, determining the BF weights based on the channel estimation results of the first user equipment relationship and the reference signal includes: The BF weights are determined based on the first BF weights and the magnitude of interference related to the first user equipment in the first user equipment relationship; wherein the first BF weights are determined based on the feature matrix obtained by channel estimation of the reference signal; or, The BF weight is determined based on the directionality of the second BF weight and the magnitude of interference related to the first user equipment in the relationship with the first user equipment; wherein, the second BF weight is a combining weight obtained by channel estimation of the reference signal.

[0012] According to a second aspect of the present disclosure, a method performed by a network node in a wireless communication network is provided, the method comprising: The first resource allocated for transmission to the first user equipment is determined; Send first information to the first user equipment, the first information including the first resource; Receive a first signal sent by the first user equipment on the first resource, or send a first signal to the first user equipment on the first resource; Wherein, the first resource is determined based on the first user equipment relationship, and the first user equipment relationship is constructed based on the resource reuse relationship between multiple user equipments and the channel feature index of each user equipment; or, The first resource is determined based on the magnitude of interference experienced by the first user equipment, the estimated interference cancellation parameters, and the performance indicators of the first user equipment.

[0013] In conjunction with some embodiments of the second aspect, in one optional embodiment of this disclosure, the first user equipment relationship includes: reusing the performance of each user equipment and the magnitude of interference when performing resource reuse.

[0014] In conjunction with some embodiments of the second aspect, in an optional embodiment of this disclosure, the method further includes: Based on the channel estimation results of the first user equipment relationship and the reference signal, the beamforming (BF) weights are determined. The first signal is processed based on the BF weights.

[0015] In conjunction with some embodiments of the second aspect, in an optional embodiment of this disclosure, determining the BF weights based on the channel estimation results of the first user equipment relationship and the reference signal includes: The BF weights are determined based on the first BF weights and the magnitude of interference related to the first user equipment in the first user equipment relationship; wherein the first BF weights are determined based on the feature matrix obtained by channel estimation of the reference signal; or, The BF weight is determined based on the directionality of the second BF weight and the magnitude of interference related to the first user equipment in the relationship with the first user equipment; wherein, the second BF weight is a combining weight obtained by channel estimation of the reference signal.

[0016] In conjunction with some embodiments of the second aspect, in an optional embodiment of this disclosure, determining the first resource allocated to the first user equipment based on the first user equipment relationship includes: Based on the resource reuse information in the first user equipment relationship, obtain the performance of the first user equipment when it reuses resources with other user equipment; Based on the performance indicators satisfying the first constraint, the resources allocated for transmission to each user equipment are determined.

[0017] In conjunction with some embodiments of the second aspect, in an optional embodiment of this disclosure, the first resource allocated to the first user equipment is determined based on the obtained interference magnitude experienced by the first user equipment, the estimated interference cancellation parameters, and the performance indicators of the first user equipment, including: Based on the magnitude of the interference received by the first user equipment and the interference cancellation parameters, the magnitude of the interference that was not eliminated is determined; Based on the performance indicators of the first user equipment and the fact that the amount of uneliminated interference meets the conditions, the first resource is determined to be allocated to the first user equipment.

[0018] According to a third aspect of the present disclosure, a method performed by a network node in a wireless communication network is provided, the method comprising: Based on the first information of the first multiplexed MU user equipment (UE) and the second MU UE, interference-related information corresponding to the first MU UE is determined, and the second MU UE and the first MU UE perform MU MIMO multiplexing transmission. The first information includes at least one of the following: MU pairing information, signal strength-related information, and channel-related information. Based on the interference-related information corresponding to the first MU UE, the first artificial intelligence (AI) network is used to perform interference cancellation processing on the signal sent by the first MU UE.

[0019] In conjunction with some embodiments of the third aspect, in an optional embodiment of this disclosure, the interference-related information corresponding to the first MU UE includes: Interference weight information, indicating the interference strength of the second MU UE to the first MU UE on each frequency domain resource; and / or, Interference channel information indicates the interference channels of the second MU UE to the first MU UE on each frequency domain resource.

[0020] In conjunction with some embodiments of the third aspect, in an optional embodiment of this disclosure, determining the interference-related information corresponding to the first MU UE includes: Based on the MU pairing information and the signal strength related information, determine the interference weight information corresponding to the first MU UE; and / or, Based on the MU pairing information and the channel-related information, the interference channel information corresponding to the first MU UE is determined.

[0021] In conjunction with some embodiments of the third aspect, in one optional embodiment of this disclosure, the first AI network is a diffusion network, and the interference cancellation processing of the signal transmitted by the first MU UE using the first AI network includes: The number of processing steps for interference cancellation in the diffusion network is determined. Using the interference-related information corresponding to the first MU UE as a constraint condition of the diffusion network, and based on the determined number of processing steps, the first AI network is used to perform interference cancellation processing on the signal sent by the first MU UE.

[0022] In conjunction with some embodiments of the third aspect, in one optional embodiment of this disclosure, determining the number of processing steps for interference cancellation processing of the diffusion network includes: Based on the environmental information of the first MU UE, the second AI network is used to determine the maximum number of steps for the diffusion network to perform interference cancellation processing; Based on the maximum number of steps and the signal strength information of the first MU UE, the number of processing steps for interference cancellation by the diffusion network is determined.

[0023] In conjunction with some embodiments of the third aspect, in an optional embodiment of this disclosure, the interference cancellation processing of the signal transmitted by the first MU UE using the first AI network includes: The interference-related information corresponding to the first MU UE is used as the constraint condition of the diffusion network, and the interference cancellation process is performed iteratively T times, where T is the determined number of processing steps; The t-th interference cancellation process in the T-th interference cancellation process includes: Determine the desired outcome of the historical interference removal process; The results of historical interference elimination are weighted, and the weighted result is used as the result of the (t-1)th interference elimination process. Based on the result of the (t-1)th interference cancellation process and the interference-related information corresponding to the first MU UE, the interference of the first MU UE is predicted; Based on the predicted interference, the interference cancellation process is performed for the t-th time.

[0024] In conjunction with some embodiments of the third aspect, in one optional embodiment of this disclosure, the signal strength-related information includes: modulation and coding scheme (MCS) and / or signal-to-interference-plus-noise ratio (SINR); The channel-related information includes the UE's channel estimation results.

[0025] This disclosure also provides an electronic device including at least one processor configured to perform the methods provided in any of the foregoing embodiments of this disclosure.

[0026] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the methods provided in any of the above embodiments of this disclosure.

[0027] The solution provided in this disclosure, compared to existing related technologies, can more accurately detect and eliminate interference, thereby better meeting the actual needs of interference management. Optional implementation methods and their corresponding effects will be described later in conjunction with the embodiments. Attached Figure Description

[0028] Figure 1a A schematic diagram of an interference management system provided for related solutions; Figure 1b This disclosure provides a schematic diagram of a 16QAM constellation diagram; Figure 1c A schematic diagram illustrating the offset of received interference noise signal provided in an embodiment of this disclosure; Figure 1d A schematic diagram of a forward diffusion process provided in an embodiment of this disclosure; Figure 1e This is a schematic diagram of an overall IGDM architecture provided in an embodiment of the present disclosure; Figure 2a This disclosure provides a flowchart illustrating a method executed by a network node. Figure 2b This disclosure provides a flowchart illustrating a method executed by a network node. Figure 2c A schematic diagram of an interference management system based on an interference channel diffusion model provided in this disclosure embodiment; Figure 2d This disclosure provides a flowchart illustrating a method executed by a network node. Figure 3 A schematic diagram of an interference-multiplexing user relationship diagram provided in an embodiment of this disclosure; Figure 4 A schematic diagram of another interference-multiplexing user relationship diagram provided in an embodiment of this disclosure; Figure 5 A schematic diagram of a mutual interference impact map based on a matched channel feature index provided in an embodiment of this disclosure; Figure 6 A schematic diagram of an extended mutual interference effect spectrum provided in an embodiment of this disclosure; Figure 7 A schematic diagram of an interference management system provided in an embodiment of this disclosure; Figure 8 This is a schematic diagram of diffusion model training and noise reduction / interference removal processing according to an embodiment of the present disclosure; Figure 9 A schematic diagram illustrating a deployment method of a solution provided in an embodiment of this disclosure; Figure 10 A schematic diagram illustrating another deployment method provided in this embodiment of the disclosure; Figure 11 A schematic diagram illustrating another deployment method provided in this embodiment of the disclosure; Figure 12 A schematic diagram illustrating another deployment method provided in this embodiment of the disclosure; Figure 13 A schematic diagram illustrating another deployment method provided in this embodiment of the disclosure; Figure 14 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure; Figure 15 This is a schematic diagram of the model structure of the training diffusion model according to an embodiment of the present disclosure; Figure 16a A schematic diagram illustrating a resource coordination process provided in an embodiment of this disclosure; Figure 16b A schematic diagram of a BF method based on interference matrix reconstruction provided in an embodiment of this disclosure; Figure 16c This is a schematic diagram of the overall structure of an IGDM provided in an embodiment of the present disclosure; Figure 17 A schematic diagram of an environment-aware model size decision module provided in an embodiment of this disclosure; Figure 18 This is a schematic diagram of a model size decision structure provided in an embodiment of the present disclosure; Figure 19 This is a schematic diagram illustrating a user scenario with different frequency bands reused, as provided in an embodiment of this disclosure. Figure 20 A schematic diagram illustrating a reused user pattern provided in an embodiment of this disclosure; Figure 21A schematic diagram of an interference channel pattern provided in an embodiment of this disclosure; Figure 22 A schematic diagram of an interference intensity weighting pattern provided in an embodiment of this disclosure; Figure 23 A schematic diagram illustrating channel data and received data preprocessing and symbol representation after processing, provided in an embodiment of this disclosure; Figure 24 This is a schematic diagram of constellation point value preprocessing provided in an embodiment of the present disclosure; Figure 25 A schematic diagram of DM processing based on multi-state weighting provided in an embodiment of this disclosure; Figure 26 This is a schematic diagram of weighted state value calculation provided in an embodiment of the present disclosure; Figure 27 A schematic diagram illustrating interference noise prediction and t-th step state value calculation provided in an embodiment of this disclosure; Figure 28a A schematic diagram of the structure of an IGDM-based communication system provided in an embodiment of this disclosure; Figure 28b A schematic diagram illustrating interference characteristic condition extraction under variable user pairing provided in an embodiment of this disclosure; Figure 29 A schematic diagram illustrating the smoothing effect of multi-state DM processing provided in an embodiment of this disclosure; Figure 30 A schematic diagram of an environment-aware model size decision module provided in an embodiment of this disclosure; Figure 31 A schematic diagram of a decision region for a different modulation format provided in an embodiment of this disclosure; Figure 32 A schematic diagram illustrating interference cancellation with different number of steps based on MCS, provided for an embodiment of this disclosure; Figure 33 A schematic diagram illustrating the input and output of an adaptive model size decision module provided in an embodiment of this disclosure; Figure 34 This is a schematic diagram of a multi-state processing method with q=2 provided in an embodiment of the present disclosure; Figure 35a A schematic diagram of the input and output of a DM-based denoising model provided in an embodiment of this disclosure; Figure 35b A schematic diagram of a noise and interference prediction (DM) process guided by interference channel pattern and interference weight pattern provided in an embodiment of this disclosure; Figure 36 A schematic diagram illustrating a method for reducing the number of diffusion model steps based on MCS, as provided in this embodiment of the disclosure; Figure 37 A schematic diagram of a distillation model provided in an embodiment of this disclosure; Figure 38 A schematic diagram of an MCS and constellation diagram provided for an embodiment of this disclosure; Figure 39 A schematic diagram illustrating a DM processing procedure based on a multi-state register, provided for an embodiment of this disclosure; Figure 40a A schematic diagram illustrating the deployment location of an IGDM on the network side, provided as an embodiment of this disclosure; Figure 40b A schematic diagram illustrating the deployment location of an IGDM on the network side, provided as an embodiment of this disclosure; Figure 41 This is a schematic diagram of an IGDM interference cancellation communication system provided in an embodiment of the present disclosure; Figure 42 A schematic diagram illustrating constellation point distances of different modulation orders provided in an embodiment of this disclosure; Figure 43 This is a schematic diagram of a multi-state diffusion process provided in an embodiment of the present disclosure. Detailed Implementation

[0029] The following description, with reference to the accompanying drawings, is provided to aid in a thorough understanding of the various embodiments of this disclosure as defined by the claims and their equivalents. This description includes various specific details to aid understanding but should be considered exemplary only. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the various embodiments described herein without departing from the scope and spirit of this disclosure. Furthermore, for clarity and brevity, descriptions of well-known functions and structures may be omitted.

[0030] The terms and wording used in the following description and claims are not limited to their dictionary meanings, but are merely used by the inventors to enable a clear and consistent understanding of this disclosure. Therefore, it will be apparent to those skilled in the art that the following description of various embodiments of this disclosure is for illustrative purposes only and not for limiting the purpose of this disclosure as defined in the appended claims and their equivalents.

[0031] It should be understood that the singular forms of “a,” “an,” and “the” can also include plural references unless the context clearly indicates otherwise. Thus, for example, the reference to “component surface” includes referring to one or more such surfaces. When we say that an element is “connected” or “coupled” to another element, the element can be directly connected or coupled to the other element, or it can mean that the element and the other element are connected through an intermediate element. Furthermore, the use of “connected” or “coupled” herein can include wireless connections or wireless couplings.

[0032] The terms “comprising” or “may include” refer to the presence of a corresponding disclosed function, operation, or component that may be used in the various embodiments of this disclosure, rather than limiting the presence of one or more additional functions, operations, or features. Furthermore, the terms “comprising” or “having” may be interpreted as indicating certain characteristics, numbers, steps, operations, constituent elements, components, or combinations thereof, but should not be construed as excluding the possibility of the presence of one or more other characteristics, numbers, steps, operations, constituent elements, components, or combinations thereof.

[0033] The term "or" as used in the various embodiments of this disclosure includes any of the listed terms and all combinations thereof. For example, "A or B" may include A, may include B, or may include both A and B. When describing multiple (two or more) items, if the relationship between the multiple items is not explicitly defined, the multiple items may refer to one, more, or all of the multiple items. For example, the description "parameter A includes A1, A2, A3" can be implemented as parameter A includes A1 or A2 or A3, or it can be implemented as parameter A includes at least two of the three items A1, A2, and A3.

[0034] Unless otherwise defined, all terms used in this disclosure (including technical or scientific terms) have the same meaning as understood by one of those skilled in the art as described herein. Common terms as defined in dictionaries are to be interpreted as having a meaning consistent with the context in the relevant technical field and should not be interpreted ideally or overly formally unless expressly defined in this disclosure.

[0035] At least some of the functions of the device or electronic device provided in this disclosure embodiment can be implemented by an AI model, such as implementing at least one module of a plurality of modules of the device or electronic device by an AI model. AI-related functions can be executed by non-volatile memory, volatile memory, and a processor.

[0036] The processor may include one or more processors. In this case, the one or more processors may be general-purpose processors, such as central processing unit (CPU), application processor (AP), etc., or pure graphics processing unit, such as graphics processing unit (GPU), visual processing unit (VPU), and / or AI-specific processors, such as neural processing unit (NPU).

[0037] The one or more processors control the processing of input data based on predefined operating rules or artificial intelligence (AI) models stored in non-volatile and volatile memory. These predefined operating rules or AI models are provided through training or learning.

[0038] Here, "providing through learning" refers to obtaining predefined operating rules or an AI model with desired characteristics by applying a learning algorithm to multiple learning datasets. This learning can be performed within the device or electronic device itself, in which the AI ​​is executed according to the embodiment, and / or can be implemented via a separate server / system.

[0039] AI models can contain multiple neural network layers. Each layer has multiple weight values, and each layer performs neural network computations by calculating the input data of that layer (such as the computation results of the previous layer and / or the input data of the AI ​​model) and the multiple weight values ​​of the current layer. Examples of neural networks include, but are not limited to, convolutional neural networks (CNNs), deep neural networks (DNNs), recurrent neural networks (RNNs), restricted Boltzmann machines (RBMs), deep belief networks (DBNs), bidirectional recurrent deep neural networks (BRDNNs), generative adversarial networks (GANs), and deep Q-networks.

[0040] A learning algorithm is a method of training a predetermined target device (e.g., a robot) using multiple learning data sets to enable, allow, or control the target device to make determinations or predictions. Examples of such learning algorithms include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.

[0041] According to this disclosure, at least one step in a method performed in an electronic device, such as interference sensing and interference cancellation, can be implemented using an artificial intelligence model. The processor of the electronic device can perform preprocessing operations on the data to transform it into a form suitable for use as input to an artificial intelligence model. The artificial intelligence model can be obtained through training. Here, "obtained through training" means obtaining a predefined operating rule or artificial intelligence model configured to perform desired features (or objectives) by training a basic artificial intelligence model with multiple training data using a training algorithm.

[0042] In a communication system, a user's transmitted signal x, after passing through channel H and reaching the receiver, will be mixed with other user signals H. n x n Including the influence of noise n, the signal undergoes BF and MMSE processing at the receiving end. Let W represent the processing effect of BF and MMSE, then the received signal is y' = W. (Hx + H n x n + n). The transmitted signal x is a modulated complex signal, and the channel and noise are also complex signals. Taking 16QAM modulation as an example, each modulation symbol is a constellation point in a constellation diagram, such as... Figure 1b As shown, constellation points can be represented using vector magnitude and phase.

[0043] The transmission signal's experience with complex channels H and noise can be considered as an amplitude and phase shift in the constellation diagram, moving from position 1 to position 2. Adding interference from other user signals, the transmitted signal's position in the constellation diagram is considered to have shifted to position 3, as shown below. Figure 1c As shown. The offset includes the magnitude (W) (H) n x n The intensity of + n) and the phase θ (W) (H) n x n (phase of +n).

[0044] Therefore, the noise reduction process for the received communication signal is to restore the signal from position 3 to position 1.

[0045] Currently, interference management schemes in communication systems fall into two main categories: traditional and AI / ML-based. However, both types of schemes primarily focus on beamforming (BF), equalization, and resource coordination. Interference mainly originates from users within the same cell or different cells that use the same resources; these are also known as reused users.

[0046] like Figure 1a The communication system and interference management module shown herein, wherein, The specific method of BF is as follows: based on the reference signal y RS Channel estimation is performed to obtain the estimated channel. Based on the estimated channel, beamforming (BF) weights are designed to concentrate the UE signal onto a specific beam direction. This separates user signals that have been multiplexed using the same time-frequency space resources from the spatial distribution, obtaining low-interference user data. When receiving data with all antennas, the BF module becomes ineffective, meaning that user signals using the same time-frequency space resources interfere with each other.

[0047] The specific equalization method involves adjusting the equalizer weights to equalize the signal, thereby eliminating signal fading and interference. The design of the equalization weights depends on channel estimation and BF weights, meaning that the signal is compensated based on the obtained channel characteristics.

[0048] Resource coordination works as follows: based on the user's reception performance, it determines whether users can reuse resources, and coordinates resources according to the resource reuse ratio. If there is significant interference between users, resulting in poor reception performance, the reuse ratio needs to be reduced, i.e., resources are sacrificed to meet performance requirements. Generally, adjustable resources include time-frequency space resources and power domain resources. Through resource coordination, the set of reused users and allocated resources can be obtained.

[0049] Typical Batch Filtering (BF) methods rely on the accuracy and real-time performance of channel estimation, and channel estimation is subject to delays due to channel changes, leading to estimation errors. These channel estimation errors can prevent BF from effectively isolating users, introducing interference. Furthermore, BF typically does not consider the impact of residual interference in the final user signal, and the magnitude of this interference is difficult to estimate. The signal after BF is used as the input for equalization at the receiver, where equalization helps to eliminate interference.

[0050] The effectiveness of equalization is related to the design of the equalization weights. Generally, equalization weights are designed based on channel estimation and BF (Browser-Filler) weights. Errors in channel estimation and BF weights lead to errors in the equalization weights. Therefore, the equalized signal still contains a significant amount of interference.

[0051] To further eliminate interference, a Successive Interference Cancellation (SIC) receiver can be used. SIC receivers perform interference cancellation serially, but cannot cancel interference from the highest priority users, and can only cancel partial interference from intermediate users. When the signal quality difference is small, the amount of uncancelled interference among users can be greater. Furthermore, SIC receivers suffer from error propagation problems. Therefore, the final signal still contains a large amount of interference, is inaccurate, and the magnitude of the interference is difficult to estimate. AI / ML-based methods often use single-process AI networks to directly cancel interference, such as CNN networks. However, these AI models have poor interpretability and cannot accurately capture the characteristics of interference superposition and removal processes, resulting in poor cancellation effects. In contrast, the diffusion model (DM) is a multi-step processing AI model based on state changes. It can capture the gradual changes in signal distribution, has strong interpretability, and has a better denoising effect. The diffusion model processes a multi-step iterative process. ,in ,and It is a hyperparameter. The process gradually adds noise to the original signal x0, such as... Figure 1dAs shown. Then, the reverse process is used to remove the noise. However, current DM-based interference removal methods only consider the characteristics of the target user. There is often correlation between multiplexed users, and as the correlation between multiplexed users increases and the number of multiplexed users increases, the features of multiplexed users become coupled together, making them difficult to distinguish. Simply relying on the characteristics of a single target user among the multiplexed users to eliminate interference will result in the eliminated interference being mixed with the useful signal, greatly reducing the interference cancellation effect.

[0052] Furthermore, based on the final interference-cancelled signal, performance metrics such as Signal-to-Interference-plus-Noise Ratio (SINR), Throughput (Tput), and Modulation Index (MI) can be obtained. Based on these performance metrics, the resource coordination module can physically isolate users with significant mutual interference or use a smaller multiplexing factor. During the multiplexing user selection process, the performance of users using the same resources is estimated, and the current user performance is used as the interference-free transmission performance. The performance of the multiplexed users is obtained based on the estimated interference level, but accurate interference information cannot be obtained.

[0053] If the obtained performance metrics are based on inaccurate signals, poor performance will result. To ensure performance, a smaller resource reuse factor is needed for resource coordination, sacrificing more resources. Furthermore, resource coordination depends on current performance metrics, which are related to the current relationships between multiplexed users. When network channels change, the obtained performance and the relationships between multiplexed users also change, affecting the resource coordination results.

[0054] In summary, the problems that existing solutions need to address include: 1) A large amount of residual interference was not eliminated, resulting in inaccurate final signal and poor performance.

[0055] 2) Unresolved interference is difficult to assess, and accurate interference information cannot be obtained.

[0056] 3) Channel estimation errors and channel variations lead to inaccurate BF design, and the BF design does not consider the impact of residual unresolved interference.

[0057] 4) Channel changes lead to changes in performance / interference and the correspondence with multiplexed users, and also cause changes in channel estimation. These changes result in inaccurate basis for BF and resource coordination design, BF cannot effectively isolate interference, resource coordination selects unsatisfactory multiplexed user pairs, and resource utilization is poor.

[0058] 5) Signals containing a large amount of unresolved interference have poor quality. To ensure transmission quality, resource coordination will sacrifice more resources.

[0059] As can be seen from the above issues, the accuracy of the output signal determines both user performance and the accuracy of BF (Block Flow) design and resource coordination. Furthermore, channel variations influence how to acquire and utilize effective interference, performance, and the correspondence between multiplexed users for BF and resource coordination. System interference originates from multiplexed users. Therefore, these three modules simultaneously determine the overall network performance.

[0060] This invention proposes an interference management method based on an interference pattern guided diffusion model (IGDM). This scheme utilizes MU pairing patterns (pairing information), estimated channels for paired MU users, and MCS (or modulation information or information representing signal strength) to perform interference cancellation on any one or more users among the paired users using a diffusion model. Furthermore, to meet the requirements of real-time interference cancellation, a multi-state diffusion processing method is designed to improve the robustness of the diffusion model, and an adaptive DM model size decision is made based on the communication environment and anti-interference requirements. The overall framework of the scheme is as follows: Figure 1e As shown.

[0061] First, interference-related information (i.e., interference pattern) for the current user is obtained based on the multiplexed user pairing information, channel information, and strength information. This interference-related information is then input into the diffusion model as a condition for the interference cancellation diffusion model. Simultaneously, interference diffusion weights are proposed, and an interference diffusion model is designed based on these weights. Under conditional guidance, this model is used for interference cancellation. This process can eliminate interference in the received signal, obtaining an accurate signal and determining the magnitude of the interference. Each step of the diffusion model is based on a weighted sum of cosine similarity values ​​from multiple historical states, considering the effects of multiple historical processing steps to reduce jitter in the diffusion process and improve robustness. Furthermore, the model size can be determined based on environmental information and the user's specific anti-interference requirements. Different environments and requirements necessitate the use of appropriate model sizes to eliminate interference and improve the model's effectiveness.

[0062] Secondly, based on the accurate information and interference magnitude after interference cancellation, a mutual interference impact map (which can also be described as a mutual interference map, which can correspond to the first user equipment relationship in the embodiments of this disclosure) based on the matching channel feature index is designed to obtain the interference, performance and multiplexing user relationship matching the current channel state, and solve the invalid information caused by channel changes.

[0063] Finally, based on the matched mutual interference map, the user with the best network performance is selected for reuse. At the same time, based on the interference magnitude and performance in the map, an interference-aware BF is designed for the selected reuse user.

[0064] The present invention can achieve the following beneficial effects: 1) Eliminate interference in the received signals of the current communication system to improve signal quality and system performance.

[0065] 2) By combining the input and output of the interference cancellation process and the diffusion model, the interference magnitude can be obtained (the interference magnitude can be perceived), providing interference values ​​for other interference-based module processing.

[0066] 3) Determine the model size based on different environments and user needs, simplify processing in simple environments and low-demand scenarios, and improve the applicability and effectiveness of the model.

[0067] 4) Consider the impact of historical multiple states based on the signal cosine similarity characteristics to reduce the fluctuations of multi-step processing in the diffusion model and improve robustness.

[0068] 5) By using interference cancellation and matching maps, accurate information on user performance, interference, and relationships with multiplexed users that match channel changes can be obtained. At the same time, channel changes can be recorded using channel feature indexes, reducing the memory usage problem of historical information caused by channel changes.

[0069] 6) Based on the matching interference impact map, a resource coordination strategy for multiplexed users that optimizes performance can be designed to improve system performance, taking into account the different performance and interference of multiplexed users.

[0070] 7) Combining perceived interference with BF weights can further eliminate interference and improve system performance.

[0071] The following description of several optional embodiments illustrates the technical solutions of this disclosure and the technical effects produced by these solutions. It should be noted that the following embodiments can be referenced, learned from, or combined with each other. Identical terms, similar features, and similar implementation steps in different embodiments will not be repeated.

[0072] Figure 2a A flowchart illustrating a method performed by a network node according to an embodiment of this disclosure is shown. Figure 2a As shown, the method may include: S101. Based on the channel estimation results of the first user equipment relationship and the reference signal, determine the beamforming (BF) weights.

[0073] Optionally, the first user equipment relationship is constructed based on the resource reuse relationship between multiple user equipments and the channel feature index of each user equipment.

[0074] Optionally, the first user equipment relationship includes: the performance of each user equipment being reused, and the magnitude of interference when reusing resources.

[0075] In some embodiments, network nodes can construct a first user equipment relationship, which can be in the form of a graph (e.g., described as a mutual interference graph or a mutual interference impact graph). It should be noted that the mutual interference graph and mutual interference impact graph mentioned below can both be understood as the aforementioned first user equipment relationship.

[0076] In some embodiments, the process of constructing a mutual interference map may include: Based on the user reuse relationships, construct a relationship diagram between interference and reused users; Each node in the relationship graph is diffused into a multi-performance node to obtain the mutual interference graph; Each performance node is connected to one performance node of other users, representing the performance of each user when the two users reuse each other. The weight of the connection between the two performance nodes is the amount of interference during reuse.

[0077] In some embodiments, the network device can acquire historical and current user multiplexing information, as well as performance indicators such as channel characteristics, interference, and signal reception quality of users during pairing, to construct an interference-multiplexing user relationship graph, such as... Figure 3 As shown. Then, considering that the same user shares resources with multiple different users in the scheduling process, and that the user's performance varies under different sharing relationships, the above interference-multiplication relationship graph is extended into a multi-node graph, namely a multi-node performance, interference-multiplication relationship graph (i.e., a mutual interference impact graph), as shown. Figure 4 As shown.

[0078] In some embodiments, after diffusing each node in the relationship graph into multi-performance nodes, the method further includes: The relationship graph is optimized based on the user's channel feature index; Based on the relationships between different sets of reused users in the optimized relationship graph, the optimized relationship graph is expanded to obtain the mutual interference graph.

[0079] In some embodiments, considering that multiplexing information, performance, and interference are historical information, when a user's channel undergoes significant changes, some historical multiplexing information and corresponding performance and interference become outdated and can no longer be used. Therefore, it is necessary to optimize the multi-node relationship graph based on the user's channel feature index. For example, nodes whose current channel index does not match the corresponding channel index in the graph are deleted. Figure 5 As shown.

[0080] In some embodiments, the optimized multi-node relationship graph can be further extended, for example, based on the relationship between different multiplexing user sets, to obtain a mutual interference graph, such as... Figure 6 As shown.

[0081] In the above embodiments, the specific process of constructing the mutual interference map can be found in the following description. Figure 2c The description includes the relevant content on "constructing a mutual interference impact map based on the matching channel feature index".

[0082] In some embodiments, step S101 may specifically include: The BF weights are determined based on the first BF weights and the magnitude of interference related to the first user equipment in the first user equipment relationship; wherein the first BF weights are determined based on the feature matrix obtained by channel estimation of the reference signal; or, The BF weight is determined based on the directionality of the second BF weight and the magnitude of interference related to the first user equipment in the relationship with the first user equipment; wherein, the second BF weight is a combining weight obtained by channel estimation of the reference signal.

[0083] In this embodiment, the specific implementation process of determining the BF weight based on the first user equipment relationship can be found in the following description. Figure 2c The description includes "Method 1 and Method 2 for BF design based on mutual interference influence matching graph".

[0084] S102. Process the first signal based on the BF weights.

[0085] Optionally, the first signal is a signal received from the first user equipment, or a signal sent to the first user equipment.

[0086] In some embodiments, a network node may perform BF processing on the received first signal using the BF weights determined above.

[0087] In some embodiments, a network node may perform BF processing on the first signal to be transmitted using the BF weights determined above.

[0088] In some embodiments, a network node receives a first signal sent by a first user equipment and inputs the first signal into a first artificial intelligence (AI) model for interference perception and interference cancellation.

[0089] Optionally, the first AI model is an interference channel diffusion model trained based on the interference removal and noise diffusion model.

[0090] In some embodiments, network devices can perform interference sensing and interference cancellation on received signals based on interference channel diffusion models trained by interference removal and noise diffusion models, thereby achieving better interference management.

[0091] In some embodiments, the training process of the interference channel diffusion model may include: Obtain the noise reduction and noise diffusion model and training set, wherein the training set includes: a set of sample signals, a set of noise addition steps, and model parameters; The training operation is repeatedly performed on the interference removal and noise diffusion model based on the training set until the training termination condition is met, thereby obtaining the trained first AI model; wherein, the training operation includes: A first sample signal is randomly selected from the set of sample signals, and a first step number is randomly selected from the set of noise addition steps; the channel of the first sample signal is obtained, and the channel matrix, interference weight, and noise weight of the corresponding real signal are calculated; the distributions corresponding to weighted interference and / or weighted noise are randomly sampled, and the predicted interference distribution and / or predicted noise distribution corresponding to the first sample signal are obtained using the interference removal and noise diffusion model; The training loss is determined based on the actual interference distribution and the predicted interference distribution corresponding to the first sample signal, and / or the actual noise distribution and the predicted noise distribution. The training termination condition includes: the training loss is less than a preset value.

[0092] In this embodiment, the training process for the interference channel diffusion model can be found in the following description. Figure 2c The description includes the "design of an interference channel diffusion model" and the relevant content of the "modeling process".

[0093] In some embodiments, the above method may further include: Send first information including a first resource to the first user equipment, wherein the first resource is a resource allocated by the network device for transmission to the first user equipment; The network node can receive the first signal sent by the first user equipment on the first resource; or, send the first signal to the first user equipment on the first resource.

[0094] In some embodiments, the above method may further include: Based on the first user relationship, the first resource allocated to the first user equipment is determined.

[0095] In this embodiment, the network device may notify the user equipment of the first resource allocated to the user equipment after resource coordination, so that the user equipment can transmit and / or receive signals on the first resource, thereby avoiding signal interference.

[0096] In some embodiments, the resource coordination process can be implemented using existing solutions, which will not be elaborated here.

[0097] It should be noted that the resource coordination process in this embodiment can be understood as follows: based on the goal of resource coordination, and under the condition that the performance indicators of the signal meet the constraints, resources for transmission are allocated to each user equipment with resource reuse relationship, and the set of reuse users to which each user equipment belongs can also be determined.

[0098] In some embodiments, resource coordination can be based on a mutual interference map. Optionally, network devices can perform resource coordination based on information such as the performance of each user in the mutual interference map (e.g., performance metrics such as SINR, TPUT, SE, etc.) and the magnitude of interference, so as to achieve the optimal result of resource coordination.

[0099] In some embodiments, the network device can determine whether the information in the mutual interference map is available based on the current channel information of the terminal; for available information, resource coordination is performed based on the mutual interference map; for unavailable information, resource coordination is performed based on existing methods.

[0100] In some embodiments, the process of resource coordination based on mutual interference maps may include: Based on the resource reuse information in the first user relationship, obtain the performance of the first user equipment when it reuses resources with other user equipment. Based on the performance indicators satisfying the first constraint, the resource coordination result is determined, which includes allocating resources for transmission to each user equipment.

[0101] Optionally, the resource coordination results may also include: the set of multiplexed users to which each user equipment belongs.

[0102] Optionally, the first constraint can be a condition that optimizes the performance of each user equipment when transmitting signals on the resources allocated to it.

[0103] In this embodiment, the specific implementation process of resource coordination based on the mutual interference map can be found in the following description. Figure 2c The description covers the relevant content of "resource coordination based on mutual interference influence matching graph".

[0104] In some embodiments, resource coordination can also be based on the effectiveness of interference cancellation. Optionally, network devices can perform resource coordination based on interference cancellation parameters (e.g., described as interference cancellation parameters) used to evaluate the effectiveness of interference cancellation. Optionally, the interference cancellation parameter can be the ratio between the signal after interference cancellation and the signal without interference cancellation. Optionally, the interference cancellation parameter can be used to calculate the magnitude of interference that was not cancelled.

[0105] In some embodiments, the first resource allocated to the first user equipment is determined based on the obtained interference magnitude experienced by the first user equipment, the estimated interference cancellation parameters, and the performance indicators of the first user equipment.

[0106] In some embodiments, the interference cancellation parameters can be determined based on a first error and a second error. Optionally, the first error can be the error between the output signal of the first AI model and a second signal, and the second error can be the error between the input signal of the first AI model and the second signal; wherein, the second signal can be understood as the ideal data signal of the signal input to the first AI model.

[0107] Optionally, the interference cancellation parameter can be the ratio of the first error to the second error.

[0108] In this embodiment, the specific implementation process of resource coordination based on the effectiveness of interference cancellation can be found in the following description. Figure 2c The description of "another method of resource coordination" is related to this.

[0109] In some embodiments, step S102 may include: preprocessing the first signal and then inputting it into the first AI model.

[0110] Optionally, the preprocessing includes beamforming (BF) processing, or the preprocessing includes BF processing and equalization processing, or the preprocessing includes BF processing, equalization processing, and serial interference cancellation (SIC) processing.

[0111] In some embodiments, after receiving a signal sent by a terminal, the network device may preprocess the signal before inputting the preprocessed signal into the first AI model.

[0112] Optionally, after receiving the signal sent by the terminal, the network device can input the signal into the BF module for processing, and then input the first AI model.

[0113] Optionally, after receiving the signal sent by the terminal, the network device can input the signal into the BF module for processing, then input it into the equalizer (or equalization module) for processing, and finally input it into the first AI model.

[0114] Optionally, after receiving the signal sent by the terminal, the network device can input the signal into the BF module for processing, then into the equalizer (or equalization module) for processing, then into the SIC receiver for processing, and finally into the first AI model.

[0115] In some embodiments, the processing of the BF mode can be implemented using existing solutions, which will not be elaborated here.

[0116] In some embodiments, the BF weights used in the BF processing are determined based on the cross-interference graph. Optionally, the network device can obtain new BF weights based on information such as the performance of each user in the cross-interference graph (e.g., performance metrics such as SINR, TPUT, SE, etc.) and the magnitude of interference, as well as the current BF weights, and use the new BF weights for BF processing when a signal is received.

[0117] In some embodiments, the method for determining the new BF weights includes any of the following: Method 1: Determine new BF weights based on the first BF weights and the magnitude of interference associated with the terminal in the mutual interference spectrum. The first BF weights are determined based on the feature matrix obtained by channel estimation of the first signal. The first BF can be understood as the current BF weights.

[0118] In this embodiment, the specific implementation process of determining the BF weights based on the mutual interference spectrum can be found in the following description. Figure 2c The description includes the relevant content of "Method 1 for BF design based on mutual interference influence matching graph".

[0119] Method 2: Determine new BF weights based on the directionality of the second BF weights and the magnitude of interference associated with the terminal in the mutual interference spectrum; wherein, the second BF weights are the combining weights obtained based on channel estimation of the first signal. The second BF can be understood as the current BF weights, used to characterize the beam direction.

[0120] In this embodiment, the specific implementation process of determining the BF weights based on the mutual interference spectrum can be found in the following description. Figure 2c The description includes the relevant content of "Method 2 for BF design based on mutual interference influence matching graph".

[0121] Figure 2b A flowchart illustrating a method performed by a network node according to an embodiment of this disclosure is shown. Figure 2b As shown, the method may include: S111. Determine the first resource allocated for transmission to the first user equipment.

[0122] Optionally, the first resource is determined based on the first user equipment relationship, which is constructed based on the resource reuse relationship between multiple user equipments and the channel feature index of each user equipment.

[0123] In some embodiments, the first user equipment relationship includes: reusing the performance of each user equipment and the magnitude of interference when reusing resources.

[0124] In some embodiments, step S111 may include: determining the first resource allocated to the first user equipment based on the first user equipment relationship.

[0125] Optionally, based on the resource reuse information in the first user equipment relationship, the performance of the first user equipment when sharing resources with other user equipment can be obtained; Based on the performance indicators satisfying the first constraint, the resources allocated for transmission to each user equipment are determined.

[0126] In this embodiment, the specific implementation process of determining the first resource allocated to the first user equipment based on the first user equipment relationship can be found in the relevant description above, or in the following description in conjunction with... Figure 2c The description covers the relevant content of "resource coordination based on mutual interference influence matching graph".

[0127] Optionally, the first resource is determined based on the magnitude of interference experienced by the first user equipment, the estimated interference cancellation parameters, and the performance indicators of the first user equipment.

[0128] In some embodiments, step S111 may include: determining the first resource allocated to the first user equipment based on the obtained interference magnitude, estimated interference cancellation parameters, and performance indicators of the first user equipment.

[0129] Optionally, the magnitude of the interference that was not eliminated can be determined based on the magnitude of the interference experienced by the first user equipment and the interference cancellation parameters. Based on the performance indicators of the first user equipment and the fact that the amount of uneliminated interference meets the conditions, the first resource is determined to be allocated to the first user equipment.

[0130] In this embodiment, the specific process of determining the first resource based on interference cancellation parameters through resource coordination can be found in the relevant description above, or in the following description. Figure 2c The description of "another method of resource coordination" is related to this.

[0131] S112. Send first information to the first user equipment, the first information including the first resource.

[0132] In some embodiments, after determining a first resource allocated to a first user equipment, a network node may notify the first user equipment so that the first user equipment may receive and / or transmit signals on the first resource.

[0133] S113. Receive a first signal sent by the first user equipment on the first resource, or send a first signal to the first user equipment on the first resource.

[0134] In some embodiments, the above method may further include: Based on the channel estimation results of the first user equipment relationship and the reference signal, the beamforming (BF) weights are determined. The first signal is processed based on the BF weights.

[0135] In some embodiments, network nodes may also determine the BF weights used when performing BF processing on signals based on the first user equipment relationship.

[0136] In this embodiment, the specific process of determining the BF weight can be found in the relevant description of step S101 above.

[0137] In the above embodiments, the network device can establish a first user equipment relationship. It should be noted that the process of establishing the first user equipment relationship can be found in the following description. Figure 3-6 ,as well as Figure 2c The description includes the relevant content on "constructing a mutual interference impact map based on the matching channel feature index".

[0138] In some embodiments, determining the BF weights based on the channel estimation results of the first user equipment relationship and the reference signal includes: The BF weights are determined based on the first BF weights and the magnitude of interference related to the first user equipment in the first user equipment relationship; wherein the first BF weights are determined based on the feature matrix obtained by channel estimation of the reference signal; or, The BF weight is determined based on the directionality of the second BF weight and the magnitude of interference related to the first user equipment in the relationship with the first user equipment; wherein, the second BF weight is a combining weight obtained by channel estimation of the reference signal.

[0139] In this embodiment, the specific implementation process of determining the BF weights based on the mutual interference spectrum can be found in the following description. Figure 2c The description includes the relevant content of "Method 2 for BF design based on mutual interference influence matching graph".

[0140] In some embodiments, a network node may determine a first AI model and send the first AI model to a terminal, wherein the first AI model is used to perform interference sensing and interference cancellation on the input signal.

[0141] Optionally, the first AI model is an interference channel diffusion model trained based on the interference removal and noise diffusion model. In some embodiments, the process by which the network node determines the first AI model can be described in conjunction with the following description. Figure 2c The description includes the "design of an interference channel diffusion model" and the relevant content of the "modeling process".

[0142] In some embodiments, after a network node trains an interference channel propagation model, it can send the model to a first user equipment so that the first user equipment can use the model to perform interference sensing and interference cancellation on the received signals.

[0143] In some embodiments, the above method may further include: Receive second information sent by the first user equipment, the second information including: the performance and interference level obtained by the first user equipment running the first AI model; Network nodes can construct a mutual interference map based on the second information.

[0144] In some embodiments, after receiving the interference channel propagation model sent by the network node, the first user equipment runs the model to perform interference sensing and interference cancellation on the received signal, and obtains information such as performance and interference magnitude before feeding it back to the network node. The network node can construct a mutual interference map based on the information fed back by the first user equipment.

[0145] In some embodiments, the process by which network nodes construct mutual interference maps can be described in conjunction with the following description. Figure 2c The description includes the relevant content on "constructing a mutual interference impact map based on the matching channel feature index".

[0146] In some embodiments, a network node may send a first signal to a first user equipment on a first resource allocated to the first user equipment based on a mutual interference map.

[0147] In some embodiments, sending a first signal to a first user equipment on a first resource includes: After performing BF processing on the first signal, it is sent on the first resource; The BF weights used in the BF processing are determined based on the mutual interference spectrum.

[0148] In some embodiments, network nodes may perform BF processing on signals to be sent to user equipment before sending them on resources determined based on resource coordination.

[0149] In this embodiment, the specific implementation process of determining the BF weights based on the mutual interference spectrum can be found in the following description. Figure 2c The description includes "Method 1 and Method 2 for BF design based on mutual interference influence matching graph".

[0150] In some embodiments, the first user equipment may also perform the following methods: The first AI model sent by the network node is received, wherein the first AI model is an interference channel diffusion model trained based on the interference removal and noise diffusion model. The first signal from the network node is input into the first AI model for interference perception and interference cancellation.

[0151] In some embodiments, the first user equipment may receive an interference channel propagation model sent by a network node in order to run the model to perform interference sensing and interference cancellation on the received signals.

[0152] In some embodiments, the above method may further include: Obtain the performance and interference level of the first user equipment; Send a second piece of information to the network device, the second piece of information including: the performance and interference level of the first user equipment, the second piece of information being used to construct a mutual interference map.

[0153] In some embodiments, after receiving the interference channel propagation model sent by the network device, the first user equipment runs the model to perform interference sensing and interference cancellation on the received signal, and obtains information such as performance and interference magnitude before feeding it back to the network node, so that the network node can construct a mutual interference map based on the information fed back by the first user equipment.

[0154] In some embodiments, the above method may further include: The network node receives the first signal sent on the first resource, which is determined by resource coordination based on the mutual interference map.

[0155] In some embodiments, the first user equipment may receive signals sent by network nodes on a first resource determined based on resource coordination.

[0156] In some embodiments, network nodes may perform resource coordination based on a cross-interference map to determine resources for transmitting signals to the first user equipment. Optionally, after determining the first resource, the network node may notify the terminal so that the terminal can receive signals transmitted by the network node on the first resource.

[0157] In some embodiments, the process by which network devices coordinate resources based on mutual interference maps can be referred to in conjunction with the following description. Figure 2c The description covers the relevant content of "resource coordination based on mutual interference influence matching graph".

[0158] In some embodiments, the above method may further include: the first signal is a signal after BF processing, wherein the BF weights used in the BF processing are determined based on the cross-interference spectrum.

[0159] In some embodiments, network nodes may perform BF (Browser-Free) processing on the signal before transmitting it. Optionally, the BF weights used by the network node when performing BF processing on the signal may be determined based on the cross-interference graph.

[0160] In this embodiment, the specific implementation process of determining the BF weights based on the mutual interference spectrum can be found in the following description. Figure 2c The description includes "Method 1 and Method 2 for BF design based on mutual interference influence matching graph".

[0161] It should be noted that, in the above embodiments, the interference management method provided by this disclosure can be applied to one network node, or to multiple network nodes simultaneously, or to both network nodes and user equipment simultaneously.

[0162] Figure 2d A flowchart illustrating a method performed by a network node according to an embodiment of this disclosure is shown. Figure 2d As shown, the method may include: S121. Based on the first information of the first multiplexed MU user equipment (UE) and the second MU UE, determine the interference-related information corresponding to the first MU UE. Optionally, the second MU UE and the first MU UE perform MU MIMO multiplexing transmission.

[0163] Optionally, the first information includes at least one of the following: MU pairing information, signal strength related information, and channel related information.

[0164] S122. Based on the interference-related information corresponding to the first MU UE, the first artificial intelligence (AI) network is used to perform interference cancellation processing on the signal sent by the first MU UE.

[0165] In some embodiments, the interference-related information corresponding to the first MU UE includes: interference weight information, and / or, interference channel information.

[0166] Optionally, the interference weight information indicates the interference intensity of the second MU UE to the first MU UE on each frequency domain resource; Optionally, the interference channel information indicates the interference channel of the second MU UE to the first MU UE on each frequency domain resource.

[0167] In some embodiments, step S121 may include: Based on the MU pairing information and the signal strength related information, determine the interference weight information corresponding to the first MU UE; and / or, Based on the MU pairing information and the channel-related information, the interference channel information corresponding to the first MU UE is determined.

[0168] In some embodiments, step S122 may include: The number of processing steps for interference cancellation in the diffusion network is determined. Using the interference-related information corresponding to the first MU UE as a constraint condition of the diffusion network, and based on the determined number of processing steps, the first AI network is used to perform interference cancellation processing on the signal sent by the first MU UE.

[0169] In some embodiments, determining the number of processing steps for interference cancellation processing of the diffusion network includes: Based on the environmental information of the first MU UE, the second AI network is used to determine the maximum number of steps for the diffusion network to perform interference cancellation processing; Based on the maximum number of steps and the signal strength information of the first MU UE, the number of processing steps for interference cancellation by the diffusion network is determined.

[0170] In some embodiments, the interference cancellation processing of the signal transmitted by the first MU UE using the first AI network includes: The interference-related information corresponding to the first MU UE is used as the constraint condition of the diffusion network, and the interference cancellation process is performed iteratively T times, where T is the determined number of processing steps; The t-th interference cancellation process in the T-th interference cancellation process includes: Determine the desired outcome of the historical interference removal process; The results of historical interference elimination are weighted, and the weighted result is used as the result of the (t-1)th interference elimination process. Based on the result of the (t-1)th interference cancellation process and the interference-related information corresponding to the first MU UE, the interference of the first MU UE is predicted; Based on the predicted interference, the interference cancellation process is performed for the t-th time.

[0171] In the above embodiments, the signal strength related information includes: modulation and coding scheme (MCS) and / or signal-to-interference-plus-noise ratio (SINR); the channel related information includes the UE's channel estimation results.

[0172] To address the interference problem, this disclosure proposes an interference management system (or device) and method based on an interference pattern-guided diffusion model. For example... Figure 2c An interference management system based on an interference channel diffusion model provided in an embodiment of this disclosure is illustrated.

[0173] The interference management method provided in this disclosure is applicable to interference management for multiple users sharing the same cellular coverage area and for user resource sharing between different cellular cells.

[0174] The scheme of this disclosure includes four parts: an IGDM module for interference cancellation and sensing, generation of a mutual interference effect matching graph based on the matching channel feature index, resource coordination based on the graph or based on the effectiveness of interference cancellation, and BF design based on the graph.

[0175] The network side constructs an interference impact map based on the performance of each UE (SINR, TPUT, spectral efficiency, SE, mutual information, MI, etc.), interference and historical pairing information, as well as the channel characteristics of the user at the time of pairing and the currently obtained user channel characteristics. The pairing relationship of users under the current channel characteristics, and the performance and interference under pairing, are represented by the map, i.e., the mutual interference impact map.

[0176] Based on the mutual interference impact map, the network side selects the user with the best target to reuse the same resources. These reused users constitute one or more sets of reused users, and each set contains one or more reused users.

[0177] The network side will distribute the resources used by the user to the UE side, and the UE side will transmit data to the network side on the received resources.

[0178] The network side designs the BF weights based on the channel estimation results and interference information obtained from the interference map. The network side processes the received data signal through the BF module, and the output signal is used as the input to the equalizer.

[0179] The signal output from the equalizer or SiC is ultimately input to the IGDM module for final residual interference cancellation and interference detection. The IGDM module can eliminate noise and interference through diffusion processing, and at the same time obtain the magnitude of the interference.

[0180] The IGDM module can be placed after the existing processing modules in the current system as shown in the figure, or the existing BF, equalization, and SIC modules can be replaced or partially replaced by IGDM (replacing only SIC, or only equalization, or equalization + SIC, etc.).

[0181] The two modules, Graph-based Resource Coordination and Graph-based Baseline Design, are optional. Either module can be omitted, used simultaneously, or only one module can be used. If omitted, the corresponding method for that module will be the traditional method.

[0182] The interference management system provided in the embodiments of this disclosure includes several parts, which will be described in detail below.

[0183] 1. Constructing a mutual interference impact map based on the matching channel feature index (i.e., generating a mutual interference impact map based on the matching channel feature index).

[0184] Acquire historical and current user multiplexing information, as well as performance indicators such as channel characteristics, interference, and signal reception quality of users during pairing (Tput, MI, etc. are used to determine user performance indicators after the next resource multiplexing).

[0185] First, construct an interference-reusable user relationship graph based on the user reuse relationship, such as... Figure 3 As shown.

[0186] The interference-reuse relationship graph is represented by a weighted directed graph, where nodes represent users or pairs of users reusing resources, and edges between nodes indicate that users in the two nodes have reused resources. The weight of an edge represents the magnitude of interference, and the direction of the edge represents the interference from the starting user to the ending user.

[0187] Considering that the same user may share resources with multiple different users during scheduling, and that user performance varies under different sharing relationships, the aforementioned interference-multiplexing relationship graph is extended into a multi-node graph, namely a multi-node performance and interference-multiplexing relationship graph (mutual interference impact graph). The multi-node performance and interference-multiplexing relationship graph (e.g.) Figure 4 The graph shown represents the performance and interference levels obtained by the user under different reuse relationships and different interference conditions.

[0188] In the mutual interference influence graph, each user node is diffused into multiple performance nodes, and each performance node is connected to one performance node of other users, representing the performance of each user when the two users reuse each other. The weight of the edge is the magnitude of the interference when reused.

[0189] Since the reuse information, performance, and interference are historical information, when a user's channel undergoes significant changes, some historical reuse information and corresponding performance and interference become outdated and can no longer be used. In order to obtain effective information, the channel characteristics of each user during each reuse are recorded.

[0190] Channel features are characterized using a channel feature index, reducing the storage space required to store channel features. Channel features can be obtained by analyzing the channel's spectral characteristics, including spectral amplitude, CIR, phase, Doppler spread, and spectral periodicity.

[0191] Each characteristic corresponds to an index. Optionally, an index can be created for each spectral characteristic size (e.g., a table mapping characteristic values ​​to indices). Then, each time, the spectral characteristic value is compared with the spectral size in the table and mapped to the index. One index corresponds to one spectral value. The interval for the change of the spectral value corresponding to each index can be set to d. Spectral values ​​can be, for example, spectral amplitude (each spectral characteristic exists independently in permutations and combinations). After amplitude normalization, for example, amplitude value 0 corresponds to index 0, amplitude value 0.1 corresponds to index 1, and so on.

[0192] The different index values ​​are determined by the magnitude of the change in the end-user channel characteristics. If the magnitude of the change exceeds a certain threshold, the index will change to another value.

[0193] Optionally, the current spectral characteristic can only use the amplitude. For example, if the amplitude value is 0.3, it can be determined whether the change between the current amplitude of 0.3 and the previous amplitude value (e.g., 0.2) exceeds the change interval d according to the change interval d of the spectral value corresponding to each index in the correspondence table between index and characteristic value. If it exceeds the change interval d, it is considered that the channel has undergone a large change, and the index changes from the index corresponding to the amplitude value of 0.2 to the index corresponding to the amplitude value of 0.3.

[0194] Spectral feature analysis is performed on the estimated channel of the last received user, and the results are mapped to an index. This is then matched against the corresponding index in the mutual interference impact graph. If the index values ​​do not match, the edge corresponding to the multiplexing relationship and the post-multiplexing performance node are deleted from the graph. If a user had only one performance node before deletion, that node is retained. The resulting graph is the mutual interference impact graph based on the matched channel feature indexes (e.g., ...). Figure 5 (As shown).

[0195] Considering that the mutual interference impact matching graph only includes the performance of reused users, the performance of unreused users can only be estimated after reuse when allocating reused resources (e.g., if a user is reused, an offset is subtracted from the current performance). To obtain more effective performance interference-reused user relationship information, the mutual interference impact matching graph is expanded. Based on the relationship between different reused user sets in the graph, if a reused user set S1 is a subset of another reused user set S2, and S2 = S1 ↔ S3, then the interference and performance magnitude when users in S3 are reused with users in S1 can be obtained through algebraic operations. This yields the final mutual interference impact matching graph containing more effective information (e.g., ...). Figure 6 (As shown).

[0196] Second: Resource coordination based on mutual interference influence matching graph.

[0197] The resource coordination based on the mutual interference influence matching graph involves obtaining the performance metrics (such as SINR, TPUT, SE, etc.) when a user and one or more users are simultaneously reused, based on the reuse situation in the constructed graph. If there is no pairing (e.g., ... Figure 6 Given users D and E and their corresponding performance, it is assumed that the performance after pairing will deteriorate by N compared to the current performance.

[0198] It should be noted that N can be understood as the impact of the interference. The size of N can be an estimated value, such as 2 dB, or the size of the interference predicted by the existing scheme.

[0199] The goal of resource coordination is:

[0200] Resource coordination is structured as an optimization problem, aiming to achieve the optimal performance metric (e.g., TPUT) under certain constraints (which may correspond to the first constraint mentioned above). The optimization objective P is designed as a communication availability metric such as SINR, Tput, SE, or MI. The constraints are any one or more existing constraints in the network, such as resource utilization, QoS reaching the corresponding level, other evaluation metrics for users, correlations, or networks, or constraints on the calculation method that a certain metric should satisfy. The resource coordination process mainly involves obtaining interference performance and channel-related information based on the interference map, and selecting the pairing that maximizes the objective as the final resource coordination result, such as... Figure 16a As shown.

[0201] Optionally, correlation characterizes the degree of correlation or similarity between user channels at different locations; higher similarity indicates greater interference. However, correlation can only characterize linear similarity, while channels typically exhibit nonlinear relationships. Therefore, Jensen-Shannon divergence (JSD) is introduced to evaluate the channel similarity between users at different spatial locations in communication. The JSD calculation formulas for user m's channel and user n's channel are as follows:

[0202] JSD can be used as a constraint for resource coordination; a smaller JSD indicates a higher similarity between two channels. The resource coordination optimization problem is solved using existing optimization methods (such as greedy algorithms) to obtain the resource coordination result. The output result is the set of reused users and the allocated resources.

[0203] In JSD calculations, the original channel matrices of users m and n can also be transformed into amplitude matrices. , and phase matrix , Then, a two-dimensional kernel function estimation (2D-KDE) is performed on the amplitude and phase matrices to estimate the joint probability distribution of the original amplitude and phase, as shown in the formula:

[0204] in, The kernel function used is the Gaussian kernel formula, which is:

[0205] in, , To estimate bandwidth, the Silverman rule is typically used. ， Or, Scott's rule get.

[0206] Where N is the total amount of data, which is the number of all elements in the matrix. Let Variance be the variance.

[0207] in, Amplitude matrix or The i-th element in Phase matrix or The i-th element in.

[0208] Obtain the joint probability distribution Then, the original amplitude and phase can be substituted to obtain the probability matrix. and The JSD can then be calculated using the following formula:

[0209] In addition, when transforming the probability distribution, low-complexity calculation methods can be used, such as normalizing the magnitude matrix and phase matrix separately:

[0210] The resulting matrix has a sum of 1 for all elements, and can also be represented as a probability matrix. Then, the JSD is calculated for the two magnitude probability matrices and the two phase probability matrices respectively. The final JSD is:

[0211] For example, the resource coordination output includes the resources for each user. For instance, to achieve the above objective, the best outcome is when users A and C use different resources, while users B and A use the same resources. Therefore, the output resource coordination result is the resources allocated to each user. Additionally, users can be grouped into different sets based on whether they use the same resources; users {AB} are a reuse set, and user {C} is another reuse set.

[0212] It should be understood that resource coordination is essentially about selecting the resources and reuse scenarios for the next user's actual usage. The graph includes historical and current performance interference during reuse. This performance interference data can then further assist in the next resource coordination process.

[0213] Optionally, the result of resource coordination each time is that each user can only be in one reuse set. For example: Figure 5 The example shows two multiplexing scenarios for user A: user C and multiplexed users BC. This is based on historical data: once user A was multiplexed with user C, and another time user A was multiplexed with multiplexed users BC. However, in a single transmission, user A can only be in one set of multiplexed users.

[0214] When resource coordination is not based on mutual interference influence matching graphs, traditional resource coordination methods are used.

[0215] After interference cancellation and perception through IGDM, the signal quality is improved, and the overall performance is enhanced. This means the model can eliminate channel interference and its tolerance to interference is increased. Since multiplexing with more users introduces more interference, IGDM allows for the reuse of more users while maintaining the same performance. Based on the better signal quality and perceived interference obtained after IGDM, a resource coordination scheme based on the effectiveness of interference cancellation is designed. Optionally, the signal quality after multiplexing can be obtained based on signal quality (SINR) and residual interference degradation, i.e., more accurate overall performance can be obtained through resource coordination based on more accurate signal quality and interference. Interference effectiveness is used to obtain the interference value that still exists after multiplexing, i.e., the SINR degradation value caused by interference (…). ).

[0216] Another resource coordination method is resource coordination based on the effectiveness of interference elimination, which can further improve resource utilization and performance.

[0217] Define the effectiveness of interference cancellation ( ): The decoded data (ack) is represented as the original ideal data 'a' (e.g., in QPSK, the data can be mapped to the corresponding constellation point, such as 1+i). The IGDM input and output data signal sizes are b and c, respectively, and b and c are processed to have the same dimension according to the ideal data dimension. The Error Vector Magnitude (EVM) is used to evaluate the EVM values ​​of a and b, and a and c, respectively, and is denoted as EVM. ab EVM ac Define interference elimination effectiveness as EVM. ac / EVM ab。

[0218] The goal of resource coordination is:

[0219] Constraint 3) above can be any one or more existing constraints in the network, such as resource utilization, QoS reaching the corresponding level, user correlation, JSD, or other evaluation indicators of the network, or a calculation method constraint that a certain indicator should meet. Optionally, constraint 3) can be: the correlation between users is less than a threshold, but it is not limited to this.

[0220] Wherein, Inteference offset is the final evaluated interference size, that is, the remaining interference size that has not been eliminated, calculated based on the effectiveness of interference elimination and perceived interference.

[0221] It should be noted that, Alternatively, it can be represented by the letter E, then constraint 2) above can be: .

[0222] A relatively accurate magnitude of interference can be obtained based on the perceived magnitude of the interference and the estimated effectiveness of interference elimination. When coordinating resources, an accurate sinr (i.e., sinr after reuse) caused by the interference can be obtained.

[0223] like Figure 7 The diagram illustrates an interference management system based on resource coordination for interference cancellation effectiveness. This system comprises two main interference management processes: 1) an integrated IS and IM module based on IGDM; and 2) resource coordination based on IM effectiveness (which can also be described as a paired UE selection algorithm), which is based on an evaluation method for IM effectiveness. Optionally, 2) processes the IS and IM results from 1). Therefore, IS and IM are performed simultaneously by IGDM.

[0224] Figure 7The IS and IM based on IGDM in the text can be understood as the IGDM model design part in the following text (including: weight design, derivation of formula (8) (i.e.: Figure 7 IGDM model design and interference noise perception cancellation part (in the middle) Figure 7 In the context of IS and IM by IGDM, IS stands for Interference Sensing and IM stands for Interference Cancellation.

[0225] Third: Allocate resources to users.

[0226] After resource coordination, the resources allocated to users are distributed to them, and users send signals to the receiving end on the allocated resources.

[0227] 4. Design of BF based on mutual interference influence matching spectrum.

[0228] The BF design based on the mutual interference impact matching map involves, after receiving the signal transmitted by the user, determining the optimal number of antennas or adjusting the merging weights based on the interference magnitude obtained from the mutual interference impact matching map, according to the information of the multiplexed user set after resource coordination and the interference magnitude obtained from the mutual interference impact matching map, while ensuring that the adjusted performance meets the requirements. If no valid interference is obtained in the map, the interference is considered to be zero.

[0229] Optionally, methods for BF design based on mutual interference influence matching maps include: Method 1: Perform SVD decomposition on the estimated channel matrix, and take the eigenvectors with the largest eigenvalues ​​as the BF weights. Based on the interference levels to multiplexed users in the spectrum, take the eigenvectors with the largest eigenvalues ​​as the final BF weights. The value of K is related to the interference level; the greater the interference, the larger the K value. To avoid performance loss in the received signal, MK is limited to >= L. Here, L is the number of eigenvectors corresponding to the channel matrix, and is an integer greater than 1.

[0230] Method 2: For example Figure 16b As shown, the interference matrix is ​​reconstructed based on the interference magnitude obtained from the spectrum. After obtaining the channel estimation matrix, the interference matrix is ​​subtracted from the channel estimation matrix to obtain the new channel matrix. Based on the new de-interferenced channel matrix, traditional methods such as SVD / precoding matrix / maximum ratio combining are used for BF design.

[0231] Method 3: Calculate the channel combining weight based on the estimated channel, and combine the data from all antennas into a weight with fewer antennas. The combining weight can be adjusted according to the magnitude of interference between multiplexed users and the directionality of the combining weight, concentrating a user's signal in a direction with better signal strength.

[0232] The process of adjusting the merging weights may include: obtaining the merging weights for multiple users; adjusting the merging weights (i.e., their direction) based on the interference levels of any two multiplexed users, directing them towards a position with less interference; calculating the user signal quality after adjusting the merging weights; and using the adjusted merging weights if the signal quality improves.

[0233] For example, after resource coordination, if the user reuse set includes two sets {ABC} and {D}, then UE A is only affected by interference from UE B and UE C. The weights can then be adjusted based on the magnitude of interference from UE B to UE A and from UE C to UE A obtained from the graph, as well as performance-related factors. If UE B significantly interferes with UE A, the weight direction of UE A can be adjusted to be further away from UE B. Here, the BF merging weight is a spatially directional weight.

[0234] Optionally, the signal quality can be estimated based on the current adjusted merge weights (i.e., the merged signal quality can be re-estimated using the adjusted weights), and the weights can be left unadjusted if the signal quality is found to be poor.

[0235] It should be noted that when not using graph-based BF weights, the traditional BF weight design method is adopted.

[0236] 5. Balance and SIC elimination.

[0237] The signal after BF is equalized using equalization modules such as MMSE / ZF. The equalized signal is then subjected to SiC cancellation. The SiC module is an optional interference cancellation module, and its use is selected based on factors such as hardware complexity, performance, and cost.

[0238] IGDM can perform interference cancellation on each user simultaneously in parallel. It can also be further enhanced by using a serial approach. This involves sorting multiplexed users according to the quality of their received signals, performing IGDM interference cancellation on the first user, subtracting the first user's interference-cancelled signal from the received signal for the second user, and then performing IGDM interference cancellation on subsequent users. This serial method can further optimize the input signal quality and improve the performance of the signal after interference cancellation.

[0239] 6. Interference cancellation and interference sensing based on IGDM.

[0240] The interference cancellation and interference sensing method based on IGDM requires the design of an IGDM model. This model, based on diffusion models (DM), designs interference and noise removal models according to interference-related information conditions; that is, an interference pattern-guided diffusion model, and models the interference. The location and structure of the IGDM in the communication system are as follows: Figure 28a As shown.

[0241] Specifically, the overall structure of the interference pattern-guided diffusion model IGDM is as follows: Figure 16c As shown, the model includes an environment-aware model size decision module and an interference-aware cancellation module. The interference-aware cancellation module includes a guidance condition acquisition module, a multi-state processing module, a DM processing step count based on anti-interference requirements, and a DM-based interference-aware cancellation module. The inputs to the IGDM include the estimated channel of the multiplexed user, the MCS of the multiplexed user, the pairing information of the multiplexed user in each frequency band, and the received signal after MMSE.

[0242] In some embodiments, the environmental perception model size decision module is used to determine the maximum model size of the interference perception and elimination module, such as Figure 33 As shown. Due to the complexity of the communication environment, different environments have different channel and interference conditions, and users' anti-interference needs also vary. Using a uniform model size to handle interference can lead to performance degradation due to too many or too few processing steps. To remove interference more efficiently and with lower complexity, an environment-based adaptive model size decision module is proposed. This module obtains the maximum number of diffusion processing steps and the register order used to store historical diffusion states based on environmental information, such as... Figure 33 As shown.

[0243] The model size includes the maximum number of DM steps T and the number of states q in the multi-state processing module, such as... Figure 17 As shown, determining different model sizes based on different environments can reduce processing complexity in certain environments.

[0244] The environment-aware model size decision module first determines whether the UE's received signal quality meets QoS (quality of service) requirements based on environmental parameters, and then classifies users into two categories: IGDM and non-IGDM users. Figure 30 As shown.

[0245] The environmental parameters include quality indicators related to QoS requirements, such as user received signal delay, SINR, and Tput. Non-IGDM users do not require IGDM interference cancellation. IGDM users require IGDM interference cancellation. For IGDM users, input their corresponding environmental information (including user density, traffic volume, average number of multiplexed users, channel conditions, and other factors related to the user's environment and affecting received signal quality (received power, etc.). Based on the environmental information, determine the level of interference experienced by the user and classify users into three levels: high interference, medium interference, and low interference. Taking the average number of multiplexed users and the performance of the interfered user as an example, the method for classifying the interference environment level is shown in Table 1. For interfered users whose performance is below a threshold (Th... low ) and the average number of reused users is below the threshold (Nth) low For users whose environment level is low interference, the performance of affected users is within a low threshold (Th). low ) and high threshold (Th high And the average number of reused users is below the threshold (Nth) low For users whose environmental level is medium interference, see Table 1 below for other level classifications.

[0246] Table 1

[0247] When using three or more environmental parameters to classify levels, the table can be expanded into a multi-dimensional parameter combination, classifying levels according to the range of each parameter. The adaptive size decision for different levels is as follows: Figure 17 As shown.

[0248] The adaptive model size decision module can learn the maximum model size using an AI model (such as a CNN), or it can obtain the maximum model size using traditional statistical or other classification methods. The module's input includes environmental information such as the channel, user performance (SINR / TPUT, etc.), and historical maximum model sizes. Figure 18 As shown. For example, using a CNN network with channel and interference environment, user performance, and historical model size as input, the maximum DM steps for a user in a low interference environment are learned to be 100, with 3 states and a 2dB improvement in SINR; in a high interference environment, the maximum DM steps are 800, with 5 states. The model size gradually decreases as the interference level changes from high to medium to low. Users in the same environment use the same model size. The reuse of users is evaluated in real time. When the user environment changes, the above process is repeated to determine the maximum model size for the user. If the user environment changes to a non-interference scenario, IGDM is no longer needed.

[0249] Once the maximum model size is determined, the IGDM corresponding to the maximum model size is used to perform interference detection and cancellation for the relevant users. IGDM processing includes: Because the bandwidth required for each user to transmit signals in each time slot is different, different users are displayed on different frequency bands, such as... Figure 19 As shown, the correlation between paired users leads to signal and interference coupling, and the coupling characteristics change with different frequency bands. To effectively eliminate interference under varying user pairing information and interference coupling characteristics, this paper proposes using interference characteristics as a guide for interference elimination. Interference characteristics include interference intensity and angle, and the pattern can adapt to different user pairing situations. Furthermore, the intensity and phase of the interference characteristics are used to decouple interference and signal. The intensity of the interference characteristics can be obtained by multiplying SINR, MCS, or power by the estimated channel, etc. The phase is the known estimated channel phase.

[0250] First, interference-related information is obtained based on the multiplexing pattern of the multiplexing user and / or the channel information of the interfering user and / or the interference strength (interference strength can be obtained from SINR, MCS, power, etc.). The interference-related information is represented in pattern form, and the processing procedure is as follows: Figure 28b As shown (using MCS as an example to obtain intensity information), interference-related information patterns are used to quantitatively define the characteristics of interfering users, thereby better distinguishing them from target users and improving the effectiveness of interference cancellation.

[0251] When considering user 1 as the target user and performing interference cancellation on user 1, it is first necessary to obtain the multiplexing user information for user 1 in different frequency bands, such as... Figure 19 As shown. To ensure data dimensionality consistency during denoising, missing user information is added to frequency bands with missing dimensions to form a reused user pattern, as shown. Figure 20 As shown.

[0252] Because of the correlation between multiplexed users, to better distinguish the received signal characteristics of target users and interfering users within the multiplexed user group, interference-related information is obtained using the estimated channel and received signal strength indication of the multiplexed users. Furthermore, the interference-related information is represented graphically in the multiplexed user pattern. The interference-related information pattern includes the channel pattern of the interfering user's transmitted signal to the receiver and the pattern representing the weight of the received signal strength after reception. The interference channel pattern involves filling the corresponding positions in the multiplexed user pattern with the estimated channels of the target user and the multiplexed user; empty users are represented by zero channels, such as... Figure 21 As shown. To obtain the interference strength weighting pattern, interference strength information is first obtained. Interference strength represents the proportion (weight) of the target user and multiplexing user's respective strength in the mixed received signal. This can be achieved by multiplying the received signal strength (power) by the estimated channel P. H), received signal SINR, reference signal SINR, or modulation and coding scheme (MCS) represent the intensity of the received signal for each multiplexed user. i Calculate the sum of the received signal strengths (Intensity) for all multiplexed users. sum Then the intensity ratio of each reused user is W. Int,i,j =Intensity i,j / Intensity sum Where i represents the user, and j represents the j-th frequency band of the target user. Taking MCS as an example, assuming there is one frequency band, then... = The intensity ratios of target users and interfering users are mapped to the reused user pattern, with empty users represented by a zero intensity ratio, resulting in an interference intensity weight pattern, as shown below. Figure 22 As shown.

[0253] Since the received communication signal and the channel matrix are both complex signals, each complex value is represented as a point on a two-dimensional coordinate axis (the horizontal and vertical axes represent the real and imaginary parts, respectively). For DM processing, preprocessing is required to convert the channels in the interference channel pattern and the equalized received signal into real numbers F = (A, ...). (, Re, Im), where A is the amplitude value, Let Re be the phase, Re be the real part, and Im be the imaginary part, as shown below. Figure 23 As shown, preprocessing is the process of obtaining the amplitude, phase, real part, and imaginary part values ​​and combining them into a set F.

[0254] Furthermore, the user transmits signals in the corresponding frequency band using a given MCS, and each MCS corresponds to a specific modulation method, such as... Figure 38 As shown, modulation schemes include QPSK, 4QAM, 16QAM, 32QAM, 64QAM, and 256QAM. Each modulation method corresponds to a different constellation diagram, meaning the modulated signal always corresponds to a specific point in the constellation diagram. Therefore, the characteristics of the MCS (Modulation Sequence) as a target user can be used as input to provide additional information for interference cancellation, such as... Figure 16c As shown. However, the points in the constellation diagram are complex signals, which also need to be preprocessed to be converted into real numbers F=(A, In the form of Re, Im, such as Figure 24 As shown.

[0255] The interference channel pattern, interference intensity weight pattern, equalized received signal, and target user's MCS are used as DM inputs. The interference channel pattern and interference intensity weight pattern are used as conditional inputs to the DM model to guide DM to predict and eliminate interference.

[0256] First, regarding the received signal y ’ = W (Hx + H n x n + n), where x, x n The unknown transmitted signal affects the effectiveness of interference removal. To improve accuracy, interference signal patterns and interference weight patterns are used as guides. The model input and output diagram is shown below. Figure 35a As shown.

[0257] The interference channel pattern H represents the channel experienced by the multiplexed user, which can affect the phase deviation of the signal in the constellation diagram. The interference intensity pattern represents the intensity proportion W of the aliased multiplexed user signals. Int,i,j This guides the impact on interference intensity. During each step of the DM processing, an unknown signal is transmitted from each interfering user. and noise As independent signals, they are superimposed to produce an effect. Each prediction step... and During the process, interference intensity and channel patterns are used as guidance, such as Figure 35b As shown.

[0258] Because the constellation diagrams corresponding to the target user's MCS are different, the spacing between constellation points is large for low modulation order constellations, such as... Figure 31 As shown in Table 2, the soft decision region is large during decoding. Therefore, it has a higher tolerance for interference at low MCS levels, with MCS representing the anti-interference requirement. For users with low anti-interference requirements, the number of interference removal steps can be reduced, eliminating the need for more precise interference cancellation and improving the efficiency of DM processing. Figure 32 As shown. Although multi-step DM processing requires a large processing time, it can be distilled into a 2-4 step model using distillation methods, as shown. Figure 37 As shown, this reduces processing latency and meets the latency requirements of communication systems. Furthermore, a fewer initial DM model improves the efficiency of the distillation process, and fewer steps are required after distillation, further reducing processing latency.

[0259] Table 2 Modulation levels and soft decision region sizes for different modulation formats

[0260] Based on the MCS level, the DM model with a maximum step count of T is further reduced in step count. The final DM step count is T1 = T / normalization (Modulation order). The processing is illustrated below. Figure 36 As shown, higher modulation levels require more DM steps to obtain more accurate interference prediction and cancellation results.

[0261] In each step of the DM operation, the interference magnitude I and Gaussian white noise N are predicted. After each step, the predicted interference value is subtracted from the equalized received signal obtained in the previous step to obtain the signal value for that step. This process is based on a Markov chain, with each step corresponding to a state S, such as... Figure 25 As shown. Obtaining the value of the next state based on the previous state leads to large fluctuations in state changes, reducing the robustness of the model. Therefore, this invention obtains weighted state values ​​based on q historical state information, and performs the next DM operation (such as...) based on these weighted state values. Figure 25 (As shown), reducing fluctuations between states makes the changes in each state smoother, improving the efficiency and robustness of DM processing, such as Figure 29 As shown.

[0262] To store q previous states, a q-order disturbance state register is used to record the state values. Based on the recorded multiple states, the weight of each state is calculated, and the weighted sum is used to obtain a smoother state value to reduce fluctuations and enhance the robustness of the model. The processing flowchart is as follows: Figure 39 As shown.

[0263] Each processing iteration involves q states, meaning the denoising process is performed on the next step using a sliding window approach, such as... Figure 34 As shown.

[0264] Among them, the weighted state values The acquisition of is related to the first q states, and the calculation method is as follows: Get the values ​​corresponding to q states , … ; The average of the values ​​of the last q-1 states obtained from the q states is obtained. ; Will point to The direction is taken as the equivalent direction, that is, the direction of data change processed by DM, such as Figure 26 As shown in Figure (a); Calculate the cosine similarity between the final q state directions and the equivalent directions. i=1,2…q, such as Figure 26 As shown in Figure (b), and will As weights for q states.

[0265] Calculate the weighted state value ,like Figure 26 As shown in Figure (c).

[0266] Sending signal x to target user i i , experiencing signal H i After that, the signal is Hi x i Meanwhile, due to interference from multiplexed users, the signal of user i is mixed with interference signals (such as H1x1, H2x2) transmitted by other multiplexed users and received by the receiver, while also being degraded by Gaussian white noise n. Therefore, for each step of DM processing, the prediction of interference and noise is based on the weighted state value of the previous step. The correlation spectrum with interference information is performed. The user-transmitted signal x and noise are unknown and are generated by DM prediction, such as... Figure 27 As shown in Figure (a), the channels the signal experiences and the received strength can be estimated by the channels H1 and H2 and the strength weight W, respectively. INF,1 W INF,2 Decision, such as Figure 27 As shown in Figure (b), the signal is quantitatively defined. This reduces the uncertainty of interference information and makes the prediction more accurate. Where W represents the processing (such as BF, equalization, etc.) that the user-transmitted signal undergoes at the receiving end. After predicting the interference and noise, the sum of all interference and noise is the total interference and noise, as shown in Figure (b). Figure 27 As shown in Figure (c). From Subtracting interference and noise from the signal gives the signal value at step t. = - ( U represents the number of users reused, such as Figure 27 As shown in Figure (d) of the document.

[0267] In this model, the signal to be de-interferenced input to the interference diffusion model is related to the processing of the received signal on the network side. The input received signal is designed with diffusion weights suitable for the diffusion model. Through the designed weights, the interference is modeled as a weighted term in the diffusion model. Specifically: 1) Design interference diffusion weights.

[0268] The received signal model for user i is:

[0269] Let be the complex-valued channel for user i. To interfere with the user channel (complex value). This is an independent and identically distributed Rayleigh fading channel. The real channel matrix for user i. The number of users who reuse user i is U. A complex signal sent by user i, with dimension Q. 1, To interfere with the complex signal sent by user k, the dimension is Q. 1, Transmission power meets (or after normalization) For complex Gaussian white noise and . The received signal is a real number with a dimension of 2Q. 1, It is real Gaussian noise.

[0270] Considering or The signal after passing through any module in BF / Equalization / SIC Perform diffusion weighting design. That is, the IGDM module can replace all or part of the current receiver signal processing (BF, equalization, SIC), or be placed after all current processing modules. Let... To The processing operation is the same as the meaning of W described above. Then... for:

[0271] If yes No processing was performed (IGDM replaced all current processing modules), then Equal to formula (1), if the processing operation is BF (IGDM replaces the remaining processing modules following BF), then , For BF weights, if the processing operation includes equalization (IGDM is located after all current processing modules), then After equilibrium and then through SIC Same as the model after equilibrium is only.

[0272] For real number received signals:

[0273] right After normalization, then:

[0274] The designed interference diffusion weights are respectively , , , where are the weights of useful data items, interfering data items, and noise items, respectively.

[0275]

[0276] 2) Design an interference channel propagation model IGDM; The interference channel weights of the DM were redesigned. Based on the redesigned weights, the interference is modeled as additional weighted noise in the DM, called IGDM, which is a model for denoising interference and noise diffusion.

[0277] According to Formula 7, we can In diffusion form:

[0278] Formula (8) is the IGDM model. The initial signal without added noise interference. , After t-step noise addition, the result is... . Follows conditional distribution

[0279] That is: through the diffusion model, Conditional distribution and The conditional distributions are consistent and can be evaluated using the Kullback-Leibler (KL) divergence: .

[0280] Training and learning are conducted through a forward diffusion process. The distribution of the loss function is then analyzed, and a reverse operation is used to remove noise and interference. In other words, the goal during forward diffusion training is to obtain the loss function. Less than the set threshold and This is the output value at step t of the diffusion model.

[0281] like Figure 8 The diagram shown is a schematic diagram of diffusion model training and noise reduction / interference removal processing according to an embodiment of this disclosure. Figure 8 Adding noise interference in step m After signal The corresponding distribution can be represented by a single distribution symbol, i.e. . Figure 8 In the middle, signal y represents the signal of any user, for example... This represents the user's signal after step m.

[0282] in, The training process (forward diffusion process) for the noise interference term is as follows: Method 1: Model the interference distribution and noise distribution separately. The training process may include: 1) Determine the IGDM input parameters: training data, parameter set {1,…,T}, and MCS, MU pairing information, MU paired user channel, user transmit power / SINR (or other information indicating user strength ratio, such as tuning format); 2) Obtain the MU pairing pattern, calculate the interference weight and channel pattern; 3) Determine whether the loss function has reached the threshold range; 4) If the threshold range is not reached, then: Randomly select from training data , randomly select parameter t from the parameter set; Get ,calculate , , ;

[0283] The prediction module predicts noise and interference; Update the gradient value of the prediction module according to the above formula (8) and loss function.

[0284] 5) If the threshold range is reached, stop training and output the result. and .

[0285] Method 2: Unify the interference distribution and noise distribution into a single distribution. Modeling is performed, in which, , For normalization The corresponding distribution. The modeling process may include: 1) Determine the IGDM input parameters: training data, parameter set {1,…,T}, and MCS, MU pairing information, MU paired user channel, user transmit power / SINR (or other information indicating user strength ratio, such as tuning format); 2) Obtain the MU pairing pattern, calculate the interference weight and channel pattern; 3) Determine whether the loss function has reached the threshold range. 4) If the threshold range is not reached, then: Randomly select from training data , randomly select parameter t from the parameter set;

[0286] The prediction module predicts noise and interference; Update the gradient value of the prediction module according to the above formula (8) and loss function.

[0287]

[0288] After the IGDM acquires noise and interference, it eliminates the noise and interference through a reverse process.

[0289]

[0290] in, This is the data signal after noise and interference have been removed. It more closely approximates the quality of the actual transmitted signal, eliminating the effects of interference and resulting in greater accuracy.

[0291] d represents the value obtained at each step during the training of the diffusion model. The training of the diffusion model involves gradually adding noise to a noise-free signal, such as... Figure 8 Each loop represents a step, and a small amount of noise interference is added at each step to obtain the final signal whose difference from the temporal noise distribution is less than that of the loss function. Then, in the reverse process, the noise interference in the received signal is eliminated step by step.

[0292] In the process of learning noise, the diffusion model can either learn the real and imaginary parts of the noise, or it can learn the phase and amplitude separately through two parallel models, and finally synthesize them into complex noise. The model structure is as follows: Figure 15 As shown.

[0293] Another method for detecting interference is: based on the output of the IGDM. and input Calculate the signal difference value Diff, then: To sense the magnitude of interference.

[0294] according to SINR can be calculated, and performance parameters such as Tput and SE can be obtained. If these performance parameters are better, the magnitude of the perceived interference can be used, or the magnitude of the remaining interference obtained based on the effectiveness of interference elimination can be used for the generation of interference maps and BF design.

[0295] If the perceived interference magnitude is used, the purpose of BF (Browser Filtering) is to optimize the data after BF and enhance the efficiency of subsequent interference elimination. If the residual interference magnitude is used, the purpose of BF is to further eliminate residual interference and enhance the processing efficiency of subsequent modules. For the resource coordination module, any interference represents the magnitude of mutual influence between users.

[0296] To effectively detect and eliminate interference, embodiments of this disclosure propose an IGDM system based on integrated IS and IM, which can be used to detect and eliminate interference that was not eliminated after MMSE equalization. IM processing is performed on the designed IGDM to improve signal quality and interference immunity. Then, the detected interference is determined based on the designed weights and the IM results.

[0297] When IGDM is deployed in a base station, such as Figure 40a As shown, it can be deployed in a data unit (DU), or as... Figure 40bThe diagram shows that it can be deployed on the MMU. The overall network-side operation process after deployment is as follows: After receiving the signal, the network acquires channel information and preprocessing weights (BF, MMSE). The AI ​​model performs interference cancellation, and the interference-cancelled signal is transmitted to the demodulation and decoding modules. Interference information is transmitted to the MAC for mutual interference impact map generation, BF selection, and multiplexing user selection.

[0298] The solutions of the embodiments of this disclosure will be described in detail below in conjunction with different application scenarios.

[0299] Example 1: Interference cancellation within the same cellular coverage area based on multi-user multiplexing The interference management method based on IGDM proposed in this disclosure can be deployed on the network side of a cell (e.g., Figure 9 (As shown). The matching map (which can be the mutual interference impact map constructed above), interference cancellation effectiveness calculation module, IGDM interference sensing and cancellation module, BF based on mutual interference impact map, and resource coordination module based on mutual interference impact map (or interference cancellation effectiveness) are all deployed on the network side. Among them, BF based on mutual interference impact map and resource coordination based on mutual interference impact map are optional modules, selected according to whether interference sensing BF and resource coordination are required. The IGDM module can be placed on the current network side to process the received signal or replace all or part of the current modules.

[0300] Interference in multi-user multiplexing systems within the same cellular coverage area arises from multiple users sharing the same resources. Multiple users simultaneously access the same access point, which can be the same AP in a cell-free system or a gNB in ​​a MIMO system.

[0301] In some alternative implementations, the IGDM module is considered to be placed after the processing module of the current communication system, and resource coordination is used based on the mutual interference spectrum (BF) and the mutual interference spectrum (or interference cancellation effectiveness).

[0302] In the above implementation scheme, the interference management process may include: Step 1: Construct a mutual interference impact map.

[0303] Obtain historical and current multiplexing user set information, along with corresponding interference, performance, and channel characteristic information. Based on the multiplexing user information and interference, first derive an interference-multiplexing user relationship graph. Construct multiple performance nodes from the user nodes in the graph. Match the currently obtained channel characteristic index with the channel characteristic index corresponding to each performance node in the graph, and delete mismatched performance nodes and their corresponding edges. If a performance node is a standalone node, retain it. Infer the interference and performance between unpaired users based on user pairing relationships to obtain more useful information.

[0304] Step 2: Perform resource coordination and send the coordinated resources to the UE side.

[0305] Based on the matched mutual interference impact map or the calculated interference elimination effectiveness, and the constructed optimization objective, the resource coordination result is obtained through optimization, i.e., the reused users and the resources used. The network side distributes the allocated resources to the users. The network side sends resources to the users, and the users send signals to the network side on those resources.

[0306] Step 3: The network side receives the signal sent by the user and performs BF processing on it.

[0307] Optionally, the BF weights can be adjusted based on the mutual interference impact map. The BF weights are first calculated based on the estimated channel, and then the direction of the BF weights is adjusted according to the inter-user interference impact in the mutual interference impact map.

[0308] Step 4: Equalize the signal processed by BF. The equalized signal is then input to the IGDM module for interference cancellation and interference detection.

[0309] Step 5: The IGDM module performs reverse interference cancellation on the input signal according to the diffusion process of training. The interference that is cancelled is the magnitude of the perceived interference.

[0310] In this embodiment, the processing model used during IGDM training is related to the model of the modules through which the received signal passes. The signal passes through the BF and equalization modules, and the diffusion weights during training need to consider both the BF weights and the equalization weights. Taking the MMSE equalization model as an example, the weights of the received signal of UE i are... BF processing. The received signal after processing by the BF and equalization modules is:

[0311] When the equilibrium model adopts zero-breaking or other models, the diffusion weights need to be obtained by processing according to the zero-breaking or corresponding model.

[0312] In other alternative implementations, considering the IGDM module replacing all processing modules of the current communication system, and resource coordination based on mutual interference maps (or interference cancellation effectiveness), the interference management process may include: Step 1: Construct a mutual interference impact map; Step 2: Coordinate resources based on the mutual interference impact map or the calculated effectiveness of interference elimination; Step 3: The network side receives the signal sent by the user and performs interference cancellation and interference detection based on the IGDM module.

[0313] It should be noted that if the IGDM module replaces all processing modules in the current communication system, then BF weights and balancing weights do not need to be considered during IGDM training; that is, there are no BF weights or balancing weights. Wp,i .

[0314] In other alternative implementations, considering the replacement of certain modules of the current communication system, such as the equalizer, with an IGDM module, and using BF based on the cross-interference spectrum and resource coordination based on the cross-interference spectrum (or interference cancellation effectiveness), the interference management process may include: Step 1: Construct a mutual interference impact map; Step 2: Coordinate resources based on the mutual interference impact map or the calculated effectiveness of interference elimination; Step 3: The network side receives the signal sent by the user and performs BF processing on it (adjusting the weight of BF based on the mutual interference impact map, and BF processes the received signal). Step 4: The signal processed by BF is input into the IGDM module for interference cancellation and interference detection.

[0315] It should be noted that if the IGDM module replaces some modules of the current communication system, such as the equalizer, then only the BF weights are considered during IGDM training, i.e.: Wp,i = .

[0316] Example 2: Interference cancellation within different cellular coverage areas based on multi-user multiplexing The interference management method based on IGDM proposed in this disclosure can be deployed on the network side of multiple cells (such as...). Figure 10 or Figure 11 (As shown). The matching map (which can be the mutual interference impact map constructed above), interference cancellation effectiveness calculation, IGDM interference sensing and cancellation module, BF based on the mutual interference impact map, and resource coordination module based on the mutual interference impact map (or interference cancellation effectiveness) are all deployed on the network side. Among them, the BF based on the mutual interference impact map and the resource coordination based on the mutual interference impact map (or interference cancellation effectiveness) are optional modules (i.e., signal BF processing and resource coordination can be implemented using existing methods), selected based on whether interference-sensing BF and resource coordination are required. The IGDM module can be placed after the current network-side module that processes the received signal or replace all or part of the current modules.

[0317] Interference in multi-user multiplexing systems across different cellular coverage areas arises from multiple users accessing different access points sharing the same resources, or from multiple users accessing the same access point sharing the same resources. Multiple users simultaneously access the same access point, which can be the same AP in a cell-free system or a gNB in ​​a MIMO system.

[0318] Example 2 differs from Example 1 only in its deployment method; the interference management process remains the same, meaning that each network side executes the interference management process described in Example 1.

[0319] Example 3: Interference cancellation on the UE side with high processing power The interference management method based on IGDM proposed in this disclosure can be deployed simultaneously on the network side and the UE side with strong processing capabilities (such as...). Figure 12 or Figure 13 (As shown). The matching map generation, interference cancellation effectiveness calculation module, and IGDM interference sensing and cancellation module are deployed on the network side for processing interference with received user signals, or deployed on the UE side with strong processing capabilities for processing interference with signals received by the UE.

[0320] Optionally, when used on the network side to process data transmitted by the UE side, both the BF based on the mutual interference impact map and the resource coordination module based on the mutual interference impact map (or interference cancellation effectiveness) are deployed on the network side.

[0321] Optionally, when used on the UE side to process data transmitted from the network side, both the BF based on the mutual interference impact map and the resource coordination module based on the mutual interference impact map (or interference cancellation effectiveness) are deployed on the network side (it should be understood that the signal transmitted by the network side is a signal transmitted on the resources coordinated by the resource coordination module based on the mutual interference impact map after being processed by the BF based on the mutual interference impact map).

[0322] In the above embodiments, BF based on mutual interference impact map and resource coordination based on mutual interference impact map (or interference cancellation effectiveness) are optional modules, which can be selected according to whether interference-aware BF and resource coordination are required.

[0323] Optionally, the IGDM module can be placed on the current network side to process the received signal or to replace all or some of the current modules.

[0324] Optionally, network-side deployment is not limited to scenarios with the same or multiple cell coverage.

[0325] When IGDM is deployed on the network side, the interference management process can be found in the relevant description in Example 1, and will not be repeated here.

[0326] When IGDM is deployed on the UE side, the interference management process may include: Step 1: The network side sends the trained IGDM model to the UE side.

[0327] Step two: On the UE side, IGDM is run for interference cancellation and interference detection. The obtained UE-side performance and interference magnitude need to be reported to the network side. Performance can be reported directly using the processed reference signal. Interference magnitude is obtained by utilizing signal distinctness. When signals are not distinct, the interference magnitude can be reported as a data signal on the pusch channel.

[0328] Step 3: After receiving performance and interference data, the network side constructs a mutual interference impact map based on channel feature matching.

[0329] Step four: The network side coordinates resources based on the graph and selects the optimal allocation of reused resources for the target.

[0330] Step 5: Design the BF weights for the corresponding users based on the mutual interference impact map and resource coordination results.

[0331] Step 6: After the signal passes through the BF module, it is sent to the UE using the corresponding coordinated resources.

[0332] Step 7: After receiving the signal, the UE performs IGDM interference cancellation and sensing.

[0333] In step seven above, performance can be acquired, interference can be detected and reported to the network side, and then steps three, four, five and six can be executed to continuously update the graph, calculate the effectiveness of interference elimination, and optimize the BF weights and resource coordination process.

[0334] Example 4: An interference cancellation scheme that performs only IGDM and is deployed on the network side.

[0335] This embodiment only includes interference cancellation processing for IGDM, and the communication system structure is as follows: Figure 41 As shown: After the signal undergoes BF and MMSE equalization at the receiving end, it is then subjected to IGDM interference cancellation. IGDM includes condition-guided DM processing based on interference characteristics, multi-state cosine similarity-based diffusion processing, and a diffusion step number determination module based on anti-interference requirements. These three modules are consistent with the aforementioned condition-guided DM, anti-interference requirement diffusion model step number, and cosine similarity multi-state diffusion processing methods, but the specific parameters and calculation methods used in this part of the scheme differ. Condition-guided DM processing considers interference intensity and angle characteristics, guiding the DM to accurately decouple the target user and interfering user signals, improving interference cancellation quality. Multi-state cosine similarity-based diffusion model processing eliminates fluctuations in the multi-step DM processing, enhancing accuracy and stability, and improving the robustness of interference prediction. The diffusion model step number determination module based on anti-interference requirements uses different MCS values ​​to characterize anti-interference capability and determines the required number of diffusion model steps based on the MCS level, reducing the number of diffusion model steps and improving the efficiency and applicability of the IGDM model.

[0336] Assume U users each have power Transmit signal, The received signal after passing through the MMSE is:

[0337] The normalized distribution of the real signal is then:

[0338] Through MMSE equalization and normalization It can be written as:

[0339] in,

[0340] The goal of IGDM is to learn through training. The distribution of the sampled data is determined, and sampling algorithms are used to eliminate interference and noise to restore the original data. .

[0341] First, the interference characteristics are extracted through a conditional DM process based on interference characteristics. These characteristics include interference intensity and interference angle. Since the channels of the user signal and the interference signal are correlated, but the specific channel for each MU UE can be obtained from a reference signal, the channel is estimated. The intensity of UE i can be determined based on the estimated channel. and transmission power intensity The received signal strength is calculated. However, the signals from multiple users in the MU are superimposed, so a normalized intensity ratio is used to evaluate the intensity proportion of each user in the mixed signal, providing intensity information indication for learning interference signals in DM processing. The intensity ratio is defined as the intensity weight:

[0342] The interference angle characteristic θ is due to the complex channel This is caused by the angle characteristic, which can be directly defined as the channel angle of each MU paired with the UE.

[0343] The user signals of paired MUs interfere with each other; the strength and angle characteristics of all MU users can be used to distinguish between interference and user signals. Therefore, all paired users... and This serves as a condition-guided DM training method to obtain accurate interference cancellation results. In the backdiffusion process, the received signal, after preprocessing (BF, MMSE), and They are input together into the DM model for interference cancellation. The conditional distribution is... .

[0344] In traditional DM models, each step is based on the previous step to predict and eliminate noise. The bias in each step accumulates, ultimately leading to a significant deviation between the predicted signal distribution and the actual correct distribution. Therefore, a diffusion processing method based on multi-state cosine similarity is proposed. This method considers n historical states in the diffusion probability distribution and uses a weighted average of the cosine similarities of these n states. As the state distribution at step t+1, i.e. .in Let be the weighted average cosine similarity. Let cosine similarity be represented as follows: .

[0345] in, Furthermore, each state is an independent sampling process. For simplicity, let's use... express Then we have:

[0346] in, Let be the variance of the predicted distribution at state t+i. , This represents the prediction bias. The corresponding weighted average distribution is then:

[0347] This represents the predicted probability distribution during the DM sampling process, due to the inference loss of the AI ​​model. There is a deviation from the actual distribution p. The AI ​​model loss follows a Gaussian distribution with standard deviation, i.e. Considering multiple states, a weighted average of sampled statistical values ​​is required. Cosine similarity-based weighted averaging takes into account the vector characteristics of complex signals, thus capturing their properties. Furthermore, a register is used to store n historical states (n>2). The principal vector direction of the n historical states is defined as... The vector direction of each historical state i is The cosine similarity is:

[0348] Will As weights, the weighted average state of the first n states is used as the previous state of the t-th state for DM sampling to obtain... ,like Figure 43 As shown.

[0349] However, since DM is an iterative process, it requires many steps to obtain the final result, resulting in high processing complexity. To reduce this complexity, a step reduction method based on anti-interference requirements is proposed. First, the UE's MCS determines its modulation order (QPSK, 16QAM, 64QAM, etc.). Different modulation orders require different soft decision regions, meaning the region for determining a symbol as a corresponding modulation constellation point is different. The minimum distance between adjacent constellation points in the constellation diagram is defined as... The smaller the modulation order The larger, such as Figure 42 As shown.

[0350] During the interference removal process, when a result at a certain step falls into the soft decision region, it can be identified as the corresponding constellation point. The larger the soft decision region, the greater the allowable interference removal error, meaning lower accuracy requirements for interference removal and lower user demands for anti-interference capabilities. IGDM can achieve satisfactory results with fewer steps, eliminating the need for high-precision interference removal. The decision bias is defined as:

[0351] in Used to assess the deviation of the real and imaginary parts from the actual constellation points. This represents the variance. The decision region is half the minimum constellation point spacing. To ensure the accuracy of interference cancellation, the Kullback-Leibler (KL) divergence between the signal distribution obtained in the m-th step of DM processing and the true distribution is less than the allowable deviation threshold. .

[0352]

[0353] in, , It is the variance of the original transmitted signal distribution under a fixed MCS. When When smaller, The smaller the value, the more precise the IGDM denoising needs to be, and the more DM steps are required. Therefore, considering the anti-interference needs of different users, users can divide the MCS into C categories, each with the same modulation level, and use the corresponding number of denoising steps to avoid introducing additional accuracy loss due to too much or too little denoising processing.

[0354] Finally, the interference-free signal is obtained based on the training and reverse processes of IGDM.

[0355] According to the general diffusion model formula, the initial transmitted signal is: For user i, the DM model is:

[0356] Assuming It is estimated at step t. The goal of IGDM training is to minimize the following loss function:

[0357] Considering the coefficients, the loss function can be rewritten as:

[0358] The training algorithm is as follows:

[0359] The output of the reverse process in the final step 1 is the interference cancellation result. During sampling, the number of steps is determined by the MCS (Mean Differential Sampling), and the information from the previous step is obtained by weighted averaging. The specific reverse processing algorithm is as follows: The input information is: the values ​​of the first n states for each state t, n, and their corresponding cosine similarities.

[0360] To ensure latency requirements in communication, the diffusion model can be distilled to reduce its processing complexity and latency. Specific distillation or knowledge transfer methods include, but are not limited to, those mentioned above. Figure 37 As shown. The effects of distillation include a reduction in the number of parameters in the model, a reduction in the number of diffusion steps, and a reduction in processing time.

[0361] This disclosure also provides an electronic device including a processor, and optionally, a transceiver and / or memory coupled to the processor, the processor being configured to perform the steps of the method provided in any optional embodiment of this disclosure.

[0362] Figure 14 The diagram shows a structural schematic of an electronic device to which an embodiment of the present invention applies, such as... Figure 14 As shown, Figure 14 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of this disclosure. Optionally, the electronic device may be a first network node, a second network node, or a third network node.

[0363] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with this disclosure. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0364] Bus 4002 may include a pathway for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 14 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0365] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing computer programs and capable of being read by a computer, without limitation herein.

[0366] The memory 4003 is used to store computer programs that execute embodiments of the present disclosure, and is controlled by the processor 4001 to execute them. The processor 4001 is used to execute the computer programs stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.

[0367] This disclosure provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.

[0368] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.

[0369] The terms “first,” “second,” “third,” “fourth,” “1,” “2,” etc. (if present) in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in a sequence other than that shown in the figures or text.

[0370] It should be understood that although arrows indicate various operation steps in the flowcharts of the embodiments of this disclosure, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of the embodiments of this disclosure, the implementation steps in each flowchart can be executed in other orders as required. Furthermore, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on the actual implementation scenario. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage can also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured as required, and the embodiments of this disclosure do not limit this.

[0371] The above text and accompanying drawings are provided as examples only to help the reader understand this disclosure. They are not intended and should not be construed as limiting the scope of this disclosure in any way. Although certain embodiments and examples have been provided, it will be apparent to those skilled in the art, based on the content disclosed herein, that changes can be made to the illustrated embodiments and examples, and other similar implementations based on the technical concept of this disclosure can be adopted without departing from the scope of this disclosure, and these modifications and modifications are also within the protection scope of the embodiments of this disclosure.

Claims

1. A method performed by a network node in a wireless communication network, characterized in that, The method includes: Based on the first information of the first multiplexed MU user equipment (UE) and the second MU UE, interference-related information corresponding to the first MU UE is determined, and the second MU UE and the first MU UE perform MU MIMO multiplexing transmission. The first information includes at least one of the following: MU pairing information, signal strength-related information, and channel-related information. Based on the interference-related information corresponding to the first MU UE, the first artificial intelligence (AI) network is used to perform interference cancellation processing on the signal sent by the first MU UE.

2. The method according to claim 1, characterized in that, The interference-related information corresponding to the first MU UE includes: Interference weight information, indicating the interference strength of the second MU UE to the first MU UE on each frequency domain resource; and / or, Interference channel information indicates the interference channels of the second MU UE to the first MU UE on each frequency domain resource.

3. The method according to claim 2, characterized in that, The step of determining the interference-related information corresponding to the first MU UE includes: Based on the MU pairing information and the signal strength related information, determine the interference weight information corresponding to the first MU UE; and / or, Based on the MU pairing information and the channel-related information, the interference channel information corresponding to the first MU UE is determined.

4. The method according to claim 1, characterized in that, The first AI network is a diffusion network, and the interference cancellation processing of the signal transmitted by the first MU UE using the first AI network includes: The number of processing steps for interference cancellation in the diffusion network is determined. Using the interference-related information corresponding to the first MU UE as a constraint condition of the diffusion network, and based on the determined number of processing steps, the first AI network is used to perform interference cancellation processing on the signal sent by the first MU UE.

5. The method according to claim 4, characterized in that, The step of determining the number of processing steps for interference cancellation processing of the diffusion network includes: Based on the environmental information of the first MU UE, the second AI network is used to determine the maximum number of steps for the diffusion network to perform interference cancellation processing; Based on the maximum number of steps and the signal strength information of the first MU UE, the number of processing steps for interference cancellation by the diffusion network is determined.

6. The method according to claim 4, characterized in that, The interference cancellation process performed on the signal transmitted by the first MU UE using the first AI network includes: The interference-related information corresponding to the first MU UE is used as the constraint condition of the diffusion network, and the interference cancellation process is performed iteratively T times, where T is the determined number of processing steps; The t-th interference cancellation process in the T-th interference cancellation process includes: Determine the desired outcome of the historical interference removal process; The results of historical interference elimination are weighted, and the weighted result is used as the result of the (t-1)th interference elimination process. Based on the result of the (t-1)th interference cancellation process and the interference-related information corresponding to the first MU UE, the interference of the first MU UE is predicted; Based on the predicted interference, the interference cancellation process is performed for the t-th time.

7. The method according to claim 1, characterized in that, The signal strength-related information includes: modulation and coding scheme (MCS) and / or signal-to-interference-plus-noise ratio (SINR); The channel-related information includes the UE's channel estimation results.

8. A method performed by a network node in a wireless communication network, characterized in that, The method includes: Based on the channel estimation results of the first user equipment relationship and the reference signal, the beamforming (BF) weights are determined, wherein the first user equipment relationship is constructed based on the resource reuse relationship between multiple user equipments and the channel feature index of each user equipment; The first signal is processed based on the BF weight, wherein the first signal is a signal received from the first user equipment, or a signal sent to the first user equipment.

9. The method according to claim 8, characterized in that, The first user equipment relationship includes: the performance of each user equipment being reused, and the magnitude of interference when reusing resources.

10. The method according to claim 8 or 9, characterized in that, The method further includes: Send first information including a first resource to the first user equipment, wherein the first resource is a resource allocated by the network device for transmission to the first user equipment; Receive the first signal sent by the first user equipment on the first resource; or, The first signal is sent to the first user equipment on the first resource.

11. The method according to claim 10, characterized in that, The method further includes: Based on the first user equipment relationship, determine the first resource allocated to the first user equipment; or... Based on the obtained interference magnitude, estimated interference cancellation parameters, and performance indicators of the first user equipment, the first resource allocated to the first user equipment is determined.

12. The method according to claim 11, characterized in that, The determination of the first resource allocated to the first user equipment based on the first user equipment relationship includes: Based on the resource reuse information in the first user equipment relationship, obtain the performance of the first user equipment when it reuses resources with other user equipment; Based on the performance indicators satisfying the first constraint, the resources allocated for transmission to each user equipment are determined.

13. The method according to claim 11, characterized in that, The determination of the first resource allocated to the first user equipment based on the acquired interference magnitude, estimated interference cancellation parameters, and performance indicators of the first user equipment includes: Based on the magnitude of the interference received by the first user equipment and the interference cancellation parameters, the magnitude of the interference that was not eliminated is determined; Based on the performance indicators of the first user equipment and the fact that the amount of uneliminated interference meets the conditions, the first resource is determined to be allocated to the first user equipment.

14. The method according to any one of claims 8-13, characterized in that, The determination of BF weights based on the channel estimation results of the first user equipment relationship and reference signal includes: The BF weights are determined based on the first BF weights and the magnitude of interference related to the first user equipment in the first user equipment relationship; wherein the first BF weights are determined based on the feature matrix obtained by channel estimation of the reference signal; or, The BF weight is determined based on the directionality of the second BF weight and the magnitude of interference related to the first user equipment in the relationship with the first user equipment; wherein, the second BF weight is a combining weight obtained by channel estimation of the reference signal.

15. A method performed by a network node in a wireless communication network, characterized in that, The method includes: The first resource allocated for transmission to the first user equipment is determined; Send first information to the first user equipment, the first information including the first resource; Receive a first signal sent by the first user equipment on the first resource, or send a first signal to the first user equipment on the first resource; Wherein, the first resource is determined based on the first user equipment relationship, and the first user equipment relationship is constructed based on the resource reuse relationship between multiple user equipments and the channel feature index of each user equipment; or, The first resource is determined based on the magnitude of interference experienced by the first user equipment, the estimated interference cancellation parameters, and the performance indicators of the first user equipment.

16. The method according to claim 15, characterized in that, The first user equipment relationship includes: the performance of each user equipment being reused, and the magnitude of interference when reusing resources.

17. The method according to claim 15 or 16, characterized in that, The method further includes: Based on the channel estimation results of the first user equipment relationship and the reference signal, the beamforming (BF) weights are determined. The first signal is processed based on the BF weights.

18. The method according to claim 17, characterized in that, The determination of BF weights based on the channel estimation results of the first user equipment relationship and reference signal includes: The BF weights are determined based on the first BF weights and the magnitude of interference related to the first user equipment in the first user equipment relationship; wherein the first BF weights are determined based on the feature matrix obtained by channel estimation of the reference signal; or, The BF weight is determined based on the directionality of the second BF weight and the magnitude of interference related to the first user equipment in the relationship with the first user equipment; wherein, the second BF weight is a combining weight obtained by channel estimation of the reference signal.

19. The method according to any one of claims 15-18, characterized in that, Based on the first user equipment relationship, the first resource allocated to the first user equipment is determined, including: Based on the resource reuse information in the first user equipment relationship, obtain the performance of the first user equipment when it reuses resources with other user equipment; Based on the performance indicators satisfying the first constraint, the resources allocated for transmission to each user equipment are determined.

20. The method according to any one of claims 15-19, characterized in that, Based on the obtained interference magnitude experienced by the first user equipment, the estimated interference cancellation parameters, and the performance indicators of the first user equipment, the first resource allocated to the first user equipment is determined, including: Based on the magnitude of the interference received by the first user equipment and the interference cancellation parameters, the magnitude of the interference that was not eliminated is determined; Based on the performance indicators of the first user equipment and the fact that the amount of uneliminated interference meets the conditions, the first resource is determined to be allocated to the first user equipment.

21. An electronic device, characterized in that, It includes at least one processor, said at least one processor being configured to perform the method according to any one of claims 1 to 20.

22. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, performs the method described in any one of claims 1 to 20.

Images