Atmospheric duct interference optimization method and device, electronic equipment and storage medium

Abstract

The present disclosure relates to an atmospheric waveguide interference optimization method and device, electronic equipment and storage medium. The method comprises: acquiring multi-source data, the multi-source data at least including remote interference management (RIM) data, base station related data and environment data; dynamically constructing an interference propagation graph according to the RIM data, the base station related data and the environment data, wherein each node in the interference propagation graph corresponds to a base station, and each edge in the interference propagation graph represents interference generated by a disturbing base station on a disturbed base station; using the inference ability of a graph neural network, inputting the interference propagation graph, and outputting a set of cooperatively optimized wireless parameter information for the disturbed and / or disturbing base stations; and executing the wireless parameter information to optimize atmospheric waveguide interference. The present disclosure can realize the cooperative optimization management of atmospheric waveguide interference, improve the intelligence and adaptability of the system, and effectively solve the problems of single optimization dimension, insufficient information perception and static strategy in related technologies.

Description

Technical Field

[0001] This disclosure relates to the field of wireless communication, and in particular to deep learning and reinforcement learning techniques in the field of artificial intelligence, especially to methods, apparatus, electronic devices and storage media for optimizing atmospheric waveguide interference. Background Technology

[0002] In related technologies, atmospheric ducting interference in wireless networks is mainly addressed through distributed detection, centralized coordination, physical model prediction, power and beam optimization, dynamic protection time slot adjustment, and overall processing in carrier aggregation scenarios. For example, distributed base station coordination and core network intervention involve base stations autonomously detecting interference and reporting it to the core network for coordination and adjustment. The network management system centrally controls and tests, initiating downlink loading tests or energy detection to locate the interference source and globally adjust network parameters. Physical model prediction and preventative adjustment predict interference based on environmental and communication parameters and adjust them in advance. Power and beam optimization uses mathematical optimization models to adjust the transmit power or beamforming of the interfering base station, reducing far-end interference. Dynamic protection time slot adjustment notifies the interfering base station to dynamically adjust the GP (Guard Period) length after detection. Overall processing in carrier aggregation scenarios synchronously adjusts the entire cell set to an anti-interference state.

[0003] However, the solutions in related technologies suffer from problems such as limited optimization dimensions, insufficient information perception and utilization, and static strategies, making it difficult to achieve intelligent and global collaborative optimization. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and storage medium for optimizing atmospheric waveguide interference, which can effectively solve problems such as single optimization dimension, insufficient information perception, and static strategy in related technologies.

[0005] In a first aspect, embodiments of this disclosure provide an atmospheric waveguide interference optimization method, including:

[0006] Acquire multi-source data, which includes at least remote interference management (RIM) data, base station-related data, and environmental data; Based on the RIM data, the base station related data, and the environmental data, an interference propagation graph is dynamically constructed, wherein each node in the interference propagation graph corresponds to a base station, and each edge in the interference propagation graph represents the interference caused by the interfering base station to the interfering base station. Using the reasoning capability of graph neural networks, and taking the interference propagation graph as input, a set of collaboratively optimized wireless parameter information is output for the disturbed and / or interfering base stations. The aforementioned wireless parameter information is applied to optimize atmospheric ducting interference.

[0007] Secondly, embodiments of this disclosure provide an atmospheric waveguide interference optimization device, comprising: The data acquisition module is used to acquire multi-source data, which includes at least Remote Interference Management (RIM) data, base station-related data, and environmental data. The graph construction module is used to dynamically construct an interference propagation graph based on the RIM data, the base station related data, and the environmental data. Each node in the interference propagation graph corresponds to a base station, and each edge in the interference propagation graph represents the interference caused by the interfering base station to the interfering base station. The inference module is used to utilize the inference capability of the graph neural network, taking the interference propagation graph as input, to output a set of collaboratively optimized wireless parameter information for the disturbed and / or interfering base station; An execution module is used to execute the wireless parameter information to optimize atmospheric duct interference.

[0008] Thirdly, embodiments of this disclosure provide an electronic device, including: One or more processors; The processor is used to invoke instructions to cause the electronic device to perform the method described in the first aspect above.

[0009] Fourthly, embodiments of this disclosure provide a storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform the method described in the first aspect above.

[0010] Fifthly, embodiments of this disclosure provide a program product including at least one of a program and instructions, wherein when the program and instructions are executed by an electronic device, they implement the steps of the method described in the first aspect.

[0011] According to the technical solution disclosed herein, by fusing multi-source data to construct a dynamic interference propagation map and utilizing graph neural networks, the collaborative optimization and governance of atmospheric duct interference can be achieved, which can improve the intelligence and adaptability of the system and effectively solve the problems of single optimization dimension, insufficient information perception and static strategy in related technologies.

[0012] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0014] Figure 1 This is a flowchart illustrating an atmospheric waveguide interference optimization method according to an exemplary embodiment.

[0015] Figure 2 This is a flowchart illustrating an atmospheric waveguide interference optimization method according to an exemplary embodiment.

[0016] Figure 3 This is a flowchart illustrating an atmospheric waveguide interference optimization method according to an exemplary embodiment.

[0017] Figure 4 This is a block diagram of an atmospheric waveguide interference optimization device according to an exemplary embodiment.

[0018] Figure 5 This is a block diagram of an atmospheric waveguide interference optimization device according to an exemplary embodiment.

[0019] Figure 6 This is a block diagram of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0021] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments. In all embodiments of this disclosure, unless otherwise specified or logically conflicting, the terminology and / or descriptions between the embodiments are consistent and can be mutually referenced. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0022] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0023] In this disclosure, "at least one" means one or more. "More than one" means two or more.

[0024] In some embodiments, terms such as “in response to…”, “in response to determining…”, “in the case of…”, “when…”, “when…”, “if…”, etc. can be used interchangeably. These descriptions all refer to the device making a corresponding action under certain objective circumstances. They do not necessarily limit the time, nor do they require the device to make a judgment action when implementing it, nor do they mean that there must be other limitations.

[0025] In some embodiments, the acquisition of data, information, etc., may comply with the laws and regulations of the country where the location is situated.

[0026] In some embodiments, data, information, etc., may be obtained with the user's consent.

[0027] It should be noted that in some embodiments, the terms "scratching station," "scratching site," "scratching base station," and "scratching cell" can be used interchangeably. In some embodiments, the terms "scratched station," "scratched site," "scratched base station," and "scratched cell" can be used interchangeably. The terms "base station," "site," "cell," and "sector" can be used interchangeably.

[0028] The atmospheric waveguide interference optimization method, apparatus, electronic device, and storage medium of this disclosure are described below with reference to the accompanying drawings.

[0029] It should be noted that the execution subject of the atmospheric waveguide interference optimization method in this embodiment can be an atmospheric waveguide interference optimization device. The device can be implemented by software and / or hardware and can be configured in an electronic device, which may include, but is not limited to, a terminal, a server, etc.

[0030] Figure 1 This is a flowchart illustrating an atmospheric waveguide interference optimization method according to an exemplary embodiment. Figure 1 As shown, the atmospheric waveguide interference optimization method may include, but is not limited to, the following steps.

[0031] In step 101, multi-source data is acquired, which may include at least RIM (Remote Interference Management) data, base station-related data, and environmental data.

[0032] In some embodiments, RIM data can be obtained through an RIM mechanism, which enables the disturbed base station to identify the interference source by defining an Interference Reference Signal (RIM-RS) and coordinate between base stations wirelessly or via backhaul. For example, after detecting remote interference, the disturbed base station continuously transmits a RIM-RS1 signal and begins listening for RIM-RS signals. When the disturbing base station receives the RIM-RS1 signal, it implements interference mitigation measures, such as autonomous power reduction, automatic electrical downtilt adjustment, and automatic GAP (gap) expansion, and then broadcasts a RIM-RS2 (or RIM-RS1) signal. After receiving the RIM-RS2 (or RIM-RS1) signal, the disturbed base station checks the uplink channel quality; if the interference continues, it continues to transmit the RIM-RS1 signal. This cycle repeats until the disturbing base station does not receive a RIM-RS1 signal within a timer period, at which point it automatically reverts to its interference mitigation measures and waits for the next cycle.

[0033] In practical networks, for efficient management, base stations can be grouped and assigned Set IDs (Atmospheric Guide Interference Area Identifiers). The RIM information sent by the affected base station will contain this Set ID, which can be used to determine whether it belongs to the group that caused the interference. By parsing the RIM code reported by the affected base station and mapping it to the Set ID, the unique base station identifier (gNB ID) of the interfering station can be accurately located.

[0034] Based on the above, optionally, RIM data can be obtained from the RIM-RS1 signal. By parsing the RIM data, the following information can be obtained: interference source gNB ID (which can be obtained by using Set ID, and by combining the algorithm and provincial correspondence table to convert the corresponding base station identifier gNB ID of the interfering station), detection symbol (indicating which uplink symbol the interfering station is interfering with the interfering station), interference power (indicating the interference power value of the interfering station, characterizing the degree of interference), etc.

[0035] In some embodiments, base station-related data can be divided into static data and dynamic data. Optionally, the static data may include, but is not limited to, at least one of the following: a unique identifier for the base station (gNB ID), a mapped atmospheric waveguide feature sequence code (Set ID), geographical location, antenna height, maximum transmit power, and antenna model. The geographical location may be the latitude and longitude coordinates of the base station, which can be used to calculate the distance and azimuth between base stations, helping to capture spatial correlations. The antenna height can affect key parameters of signal propagation and reception. The maximum transmit power may be the upper limit of the base station's capabilities. Optionally, the antenna model may include, but is not limited to, antenna gain, beamwidth, and radiation pattern.

[0036] Optionally, in some embodiments, the dynamic data may include, but is not limited to, at least one of the following: current transmit power (such as the current actual transmit power level), operating carrier frequency (such as the center frequency of the current serving carrier), antenna tilt angle (such as a combination of electrical downtilt and mechanical downtilt), service load (such as the current cell throughput, number of user connections, etc.), and time slot allocation (such as the current uplink and downlink time slot configuration).

[0037] It should be noted that the environmental data in this disclosure can provide the external conditions for the occurrence of interference. In some embodiments, the environmental data may include, but is not limited to, at least one of meteorological data, time data, etc. The occurrence of atmospheric waveguide phenomena is closely related to the meteorological environment; therefore, meteorological data needs to be introduced as one of the input data. The meteorological data used in this disclosure may include, but is not limited to, at least one of temperature, air pressure, relative humidity, water vapor pressure, wind speed, and precipitation. The time data may include, but is not limited to, at least one of seasonal time information, monthly time information, and diurnal time information, which can be used to capture the periodic patterns of interference.

[0038] Optionally, this disclosure may also incorporate terrain data, i.e., the multi-source data may also include terrain data, such as digital elevation model (DEM) data or three-dimensional maps, which can be used to determine the impact of terrain on signal propagation.

[0039] In step 102, an interference propagation map is dynamically constructed based on RIM data, base station-related data, and environmental data.

[0040] In embodiments of this disclosure, a wireless network can be abstracted into a dynamic interference propagation graph based on RIM data, base station-related data, and environmental data to capture the complex interference relationships between base stations caused by atmospheric waveguides. In these embodiments, each node in the interference propagation graph corresponds to a base station, and each edge in the graph represents the interference caused by the interfering base station to the affected base station.

[0041] Optionally, in some embodiments, interference relationships between nodes (base stations) can be established by parsing RIM data. Based on the RIM data, base station-related data, and environmental data, interference intensity factors, temporal stability factors, and path reliability factors are determined. These factors are then fused to obtain the comprehensive weights of the edges, thereby constructing the interference propagation graph. The comprehensive weight of each edge represents its characteristic. The interference intensity factor can be used to quantify the severity of the interference. The temporal stability factor can utilize dynamic data of nodes (base stations) (such as time slot allocation and service load) and historical interference data to assess the persistence and regularity of interference over time. The path reliability factor comprehensively utilizes static characteristics of nodes (such as geographical location and antenna parameters) and environmental data to assess the reliability and stability of the interference path under atmospheric waveguide conditions.

[0042] In step 103, the reasoning capability of the graph neural network is used to output a set of collaboratively optimized wireless parameter information for the disturbed and / or disturbing base stations, taking the interference propagation graph as input.

[0043] It should be noted that this disclosure uses the interference propagation graph as input to the graph neural network, which can effectively model the spatial propagation characteristics and temporal evolution of interference.

[0044] In embodiments of this disclosure, the Graph Neural Network (GNN) can be a pre-trained network model. Optionally, an interference propagation map can be constructed using training data (including RIM data, base station-related data, and environmental data, etc.). The interference propagation map is used as the model input, and the co-optimized antenna parameter information of the disturbed and / or instigating base stations is used as the output to train the GNN, so that the GNN learns the interference propagation map G at time t. t The mapping relationship between antenna parameter information and the disturbed and / or disturbing base stations for collaborative optimization. In other words, the goal of GNN is to learn a mapping function. The function uses the disturbance propagation graph G at time t. t As input, it outputs a set of collaboratively optimized radio parameter information for the disturbed and / or disturbing base station (or cell), which may be antenna parameter adjustment information.

[0045] In some embodiments, the wireless parameter information may include, but is not limited to, at least one of: power domain adjustment, spatial domain adjustment, and time domain adjustment suggestion information. The power domain adjustment may be a suggested transmit power change (e.g., -3dB, +1dB, etc.). The spatial domain adjustment may be a suggested antenna downtilt adjustment (e.g., increasing the electrical downtilt angle to compress the vertical beam). The time domain adjustment suggestion information may be a suggested time slot allocation (e.g., adjusting the GP length of a specific subframe) or a scheduling offset.

[0046] It should be noted that this antenna parameter information is coordinated because each node... v i Antenna parameter information a i Based on its neighbor status The method, derived jointly, can solve the "prisoner's dilemma" in distributed decision-making and avoid suboptimal global solutions caused by single-site optimization.

[0047] In step 104, wireless parameter information is processed to optimize atmospheric duct interference.

[0048] In the embodiments of this disclosure, when the reasoning ability of graph neural networks is used to obtain the wireless parameter information for the cooperative optimization of the disturbed and / or disturbing base stations, the corresponding parameters can be adjusted using the wireless parameter information. Since the wireless parameter information is predicted by using RIM data, base station related data and environmental data, the adjustment of the corresponding parameters using the wireless parameter adjustment information can achieve the purpose of optimizing atmospheric duct interference.

[0049] In the above embodiments, by fusing multi-source data to construct a dynamic interference propagation map and utilizing graph neural networks, the collaborative optimization and governance of atmospheric duct interference can be achieved, which can improve the intelligence and adaptability of the system and effectively solve the problems of single optimization dimension, insufficient information perception and static strategy in related technologies.

[0050] Figure 2 This is a flowchart illustrating an atmospheric waveguide interference optimization method according to an exemplary embodiment. Figure 2 As shown, the atmospheric waveguide interference optimization method may include, but is not limited to, the following steps.

[0051] In step 201, multi-source data is acquired, which may include at least remote interference management (RIM) data, base station-related data, and environmental data.

[0052] In the embodiments of this disclosure, step 201 can be implemented in any of the ways described in the various embodiments of this disclosure. This disclosure does not limit this and will not elaborate further.

[0053] In step 202, the nodes and node characteristics in the interference propagation map are determined based on base station-related data and environmental data.

[0054] For example, the interference propagation graph is a directed graph. ,in, V Represents a set of nodes, each node v i ∈ Ve ij For each base station, the number of nodes is equal to the total number of base stations, N. E t This represents the set of directed edges at timestamp t, where each edge... e ij ∈ E t From node v i Pointing to node v j This indicates that base station i interferes with base station j (i.e., v i To harass base stations, v j (The base station is being disrupted).

[0055] It is worth noting that the graph structure of the interference propagation graph is dynamic because the edge set... E t The characteristics of nodes change over time to reflect the evolution of interference relationships and fluctuations in environmental conditions. The scope of this map can be determined by analyzing the atmospheric ducting interference area identifier (Set ID). All base stations can be decomposed into several interference propagation maps. For example, there are regions A, B, C, D, E, and F. If, by analyzing the Set ID, it can be determined that there are pathways in regions A, B, C, and E, and pathways in regions D and F, then two dynamic interference propagation maps can be established. One map includes regions A, B, C, and E, while the other includes regions D and F. Therefore, multiple regions can be decomposed into several dynamic interference propagation maps in this way.

[0056] Each node v i Associate a feature vector ,in d v This is the feature dimension. This vector integrates static, dynamic, and environmental data from the base station to ensure a comprehensive description of the base station's state and context. For example, the node feature extraction process can be as follows: (1) Static features can be extracted from the static data of the base station. These static features may include inherent attributes of the base station that do not change or change slowly over time. For example, these static features are extracted from the base station's ontological data, and may include, but are not limited to, at least one of the following: geographical location (latitude and longitude), antenna height (meters), maximum transmit power (dBm), and antenna model parameters (such as antenna gain (dBi), horizontal beamwidth (degrees), vertical beamwidth (degrees), etc.). The static feature vector can be denoted as... ,in d s This is a static feature dimension.

[0057] (2) Dynamic features can be extracted from the dynamic data of the base station. These dynamic features reflect the real-time status of the base station and can be updated from the base station's own data. Optionally, these dynamic features may include, but are not limited to, at least one of the following: current transmit power (dBm), operating carrier frequency (MHz), antenna tilt angle (degrees), service load, and time slot allocation. The dynamic feature vector can be denoted as... ,in d d It is a dynamic feature dimension that is updated over time t.

[0058] (3) Environmental features can be extracted from environmental data, which can provide external meteorological and temporal context. For each base station, environmental data is associated with its geographical location. Meteorological data may include at least one of the following: temperature (°C), air pressure (hPa), relative humidity (%), water vapor pressure (hPa), wind speed (m / s), and precipitation (mm). Temporal data can be used to capture periodic interference patterns, which may include, but are not limited to, at least one of the following: month (1-12), day / night markers (e.g., sunrise and sunset times). The environmental feature vector can be denoted as: ,in d e This refers to the environmental characteristics dimension.

[0059] Optional, node feature vectors It can be obtained by splicing static, dynamic, and environmental features, and can be represented as follows: ,in, This represents a vector concatenation operation, therefore... To ensure numerical stability, discrete data can be encoded using one-hot encoding, and other features can be normalized (such as Min-Max normalization or Z-score standardization).

[0060] In step 203, the interference relationship between base stations is determined based on the RIM data, and each edge in the interference propagation graph and its comprehensive weight are determined based on the interference relationship and the RIM data, thereby constructing the interference propagation graph.

[0061] In the embodiments of this disclosure, interference relationships between nodes (base stations) can be established by parsing RIM data. For example, the RIM data parsing process can be as follows: (1) Detection of the affected base station: The affected base station can identify that it is being interfered with from a distance by monitoring the rise in uplink noise floor and the deterioration of signal-to-interference-plus-noise ratio (SINR).

[0062] (2) Location of the interfering base station: The RIM report sent by the interfering base station may contain the interference source identifier information. The network side parses this report and accurately locates the unique interfering base station through Set ID mapping and gNB ID conversion algorithms.

[0063] (3) Relationship confirmation: When the confidence level of the RIM report (such as signal detection quality, historical accuracy) exceeds the preset threshold, the system establishes a directed edge between the harassing station and the harassed station.

[0064] Optionally, the conditions for confirming interference relationships can be as follows: side e ij The establishment of this can be determined by the following indicator functions: e ij Established I RIM置信度>θ_conf ∩I 干扰功率>θ_power =1 Where: θ_conf is the minimum threshold for RIM report confidence; θ_power is the minimum interference power threshold required to determine it as valid interference; I is an indicator function, which takes the value 1 when the condition is true.

[0065] Once the interference relationship between nodes (base stations) is determined, this interference relationship can be used as an edge between the nodes. Using RIM data, base station-related data, and environmental data, interference intensity factors, temporal stability factors, and path reliability factors can be determined. These factors are then fused to obtain the comprehensive weight of the edges, thereby constructing the interference propagation graph. In the embodiments of this disclosure, both the topology of the interference propagation graph and the comprehensive weight of the edges dynamically evolve over time.

[0066] In step 204, the reasoning capability of the graph neural network is used to output a set of collaboratively optimized wireless parameter information for the disturbed and / or disturbing base stations, taking the interference propagation graph as input.

[0067] In the embodiments of this disclosure, step 204 can be implemented in any of the ways described in the various embodiments of this disclosure. This disclosure does not limit this and will not elaborate further.

[0068] In step 205, wireless parameter information is processed to optimize atmospheric duct interference.

[0069] In the embodiments of this disclosure, step 205 can be implemented in any of the ways described in the various embodiments of this disclosure. This disclosure does not limit this and will not elaborate further.

[0070] In the above embodiments, the wireless network can be abstracted into a dynamic interference propagation map based on RIM data, base station related data, and environmental data to capture the complex interference relationship between base stations caused by atmospheric ducts. This interference propagation map can be used as the input of a graph neural network, which can effectively model the spatial propagation characteristics and temporal evolution of interference. This allows for the output of a set of collaboratively optimized wireless parameter information for the disturbed and / or interfering base stations, thereby achieving collaborative optimization and management of atmospheric duct interference.

[0071] Optionally, in some embodiments, such as Figure 3 As shown, the optional implementation methods for determining each edge and the comprehensive weight of each edge in the interference propagation graph based on interference relationships and RIM data may include, but are not limited to, the following steps: In step 301, the interference relationship between base stations is used as an edge in the interference propagation graph.

[0072] In step 302, for each edge in the interference propagation graph, the intensity of the interference power associated with the edge is obtained from the RIM data, and an interference intensity factor is determined, which can be used to quantify the severity of the interference.

[0073] In embodiments of this disclosure, for each edge in the interference propagation graph, the intensity of the interference power associated with that edge can be obtained from RIM data, and the intensity of this interference power can be normalized to obtain an interference intensity factor. As an example, the intensity of this interference power... The calculation formula can be expressed as follows:

[0074] in, The interference power is obtained from RIM data analysis. It is the set reference power.

[0075] The intensity of this interference power can be normalized, and can be expressed as follows:

[0076] in, The upper and lower limits of the interference intensity are observed or set by the system and are used to... Normalize to the [0,1] interval.

[0077] In step 303, dynamic data of edge-associated nodes and historical interference power data are analyzed to obtain time-series stability indicators and service mode matching degree. Time-domain stability factors are determined based on time-series stability indicators and service mode matching degree.

[0078] For example, the autocorrelation of the interference power time series can be calculated to determine whether the interference path is instantaneous or has a long-term interference relationship. The calculated autocorrelation of the interference power time series can be used as a time series stability index. As an example, the formula for calculating the autocorrelation of the interference power time series is as follows:

[0079] in, This is the time autocorrelation coefficient. It is the interference power at time t.

[0080] The similarity (e.g., cosine similarity) of the service load time series of the interfering and affected stations can be calculated, and this similarity can be used as the service pattern matching degree. A high similarity indicates that both stations may be busy simultaneously, which would exacerbate the interference. As an example, the formula for calculating the similarity of the service load time series of the interfering and affected stations can be expressed as follows:

[0081] in, These are the service load vectors of the harassing station and the harassed station over a period of time, respectively.

[0082] In embodiments of this disclosure, the time-series stability index and the business model matching degree can be weighted and summed to obtain the time-domain stability factor. As an example, the formula for calculating the time-domain stability factor can be expressed as follows:

[0083] in, These are the weighting coefficients.

[0084] In step 304, the reliability and stability of the interference path corresponding to the edge under atmospheric waveguide conditions are evaluated using the geographical location and antenna parameter information of the nodes associated with the edge, as well as environmental data, so as to obtain the path reliability factor.

[0085] Optionally, in some embodiments, reliability assessment can be performed based on the number of successful interference predictions and the total number of interference events to obtain a reliability assessment index based on historical data; the probability of atmospheric duct occurrence can be predicted based on the environmental data of the edge-associated nodes, and the probability of atmospheric duct occurrence can be used as a path reliability index based on weather data; topology stability analysis can be performed based on the geographical location and antenna parameter information of the edge-associated nodes to obtain a topology stability index; and the reliability assessment index based on historical data, the path reliability index based on weather data, and the topology stability index can be fused to obtain a path reliability factor.

[0086] For example, the calculation formula for the above reliability assessment index based on historical data can be expressed as follows: .

[0087] in, As a reliability assessment metric based on historical data, The coefficients are known.

[0088] Since meteorological data is usually at the district / county level, the following strategy is adopted for base station-level mapping and calculation to obtain path reliability indicators based on weather data.

[0089]

[0090] in, Route reliability index based on weather data; function It is a classifier or regressor based on an atmospheric waveguide physical model. The input is the meteorological characteristics of the district / county where the base station is located, and the output is a score that represents the probability of atmospheric waveguide occurrence, which can be used as an indicator of the reliability of the path.

[0091] Optionally, for topology stability metrics, the static geography and antenna relationships of the base station pairs can be considered. For example, at what antenna tilt angle, antenna gain, or antenna height is it easier to form a stable interference path? As an example, the topology stability metric can be determined using the following formula. : .

[0092] In embodiments of this disclosure, a path reliability factor can be obtained by weighted summation of reliability assessment indicators based on historical data, path reliability indicators based on weather data, and topology stability indicators. As an example, the fusion formula for this path reliability factor can be expressed as follows:

[0093] Wherein, the weight coefficients satisfy It can be adjusted according to the climate and terrain characteristics of different regions.

[0094] It should be noted that in some embodiments, steps 302, 303, and 304 can be executed in an interchangeable order or simultaneously.

[0095] In step 305, the interference intensity factor, temporal stability factor, and path reliability factor are fused to obtain the comprehensive weight of the edge.

[0096] In embodiments of this disclosure, the interference intensity factor, temporal stability factor, and path reliability factor can be weighted and summed to obtain the comprehensive weight of the edge. For example, each edge... e ij The comprehensive weighting integrates interference intensity, spatiotemporal stability, and path reliability, allowing for a more refined characterization of the quality of interference, rather than simply its existence. As an example, each edge... e ij The formula for calculating the overall weight can be expressed as follows:

[0097] in, These are adjustable weight coefficients for each dimension, reflecting the emphasis of network optimization on different factors.

[0098] Optionally, in some embodiments, when the overall weight of the edges in the interference propagation graph evolves dynamically over time, an exponentially weighted moving average is used to smooth the overall weight of the edges to be evolved.

[0099] An example is the topology (edge ​​set) of the interference propagation graph. E t The combined weight W of the edges ij All of these evolve dynamically over time. For example, the structure of the interference propagation map can typically be updated once an hour. New RIM measurement reports (such as hourly ones), periodic environmental data updates, or base station configuration updates will all trigger updates to the interference propagation map.

[0100] Optionally, the lifecycle management of edges in the interference propagation graph can be as follows: a sliding time window can be used for data processing. For example, an edge is only retained if an interference relationship is continuously detected in the RIM data within the most recent ΔT time period. To avoid drastic fluctuations in weights, an exponentially weighted moving average can be used to smooth the overall weights. For example, the formula for this smoothing process can be as follows:

[0101] in η It is a smoothing factor. This is the edge e ij The newly calculated overall weight, At time t-1, edge e ij The overall weight; For the smoothed edges e ij The overall weight.

[0102] In the above embodiments, by fusing interference intensity factors, temporal stability factors, and path reliability factors, a comprehensive weight of the edges in the interference propagation graph is generated, which can be used to accurately characterize the "quality" of the interference relationship and construct an information-rich dynamic interference propagation graph. This enables the system to "see through" the causes of interference and predict its evolution, and the decision-making can be upgraded from passive response to active prevention and precision strike, which can greatly improve the accuracy and timeliness of interference suppression.

[0103] Optionally, the graph neural network can be trained online using the maximum entropy reinforcement learning (SAC) framework. In some embodiments, the graph neural network can be used as a policy network and combined with a value network, and the graph neural network can be trained online using the maximum entropy reinforcement learning (SAC) framework. The graph neural network includes multiple message-passing layers, which enable nodes in the graph to aggregate information from their neighbors and connected edges to learn deep patterns inherent in the network topology and node relationships. Optionally, the reward function used by the maximum entropy reinforcement learning (SAC) framework can be a multi-objective reward function, which may include at least an interference suppression reward function, a throughput improvement reward function, a power efficiency reward function, a user experience reward function, and an oscillation penalty function.

[0104] For example, graph neural networks (GNNs) can learn deep patterns in network topology and node relationships by enabling nodes in a graph to aggregate information from their neighbors and connected edges through message passing mechanisms. In this disclosure, the goal of a GNN is to learn a mapping function. The function uses the disturbance propagation graph G at time t. t As input, it outputs a set of collaboratively optimized wireless parameter adjustment strategies for the disturbed and / or disturbing base stations.

[0105] Optionally, the GNN inference process can contain multiple message-passing layers. In the... Layer, for each node in the graph v i Perform the following operations: (1) Message aggregation: node v i From all its neighboring countries (i.e., all the interfering stations) and outgoing neighbors. Aggregated information (i.e., all affected stations interfered with by it). The aggregation process can consider edge weights W. ji and W ik This allows edges with high interference intensity and high reliability to have a greater weight in the decision-making process.

[0106]

[0107] in: It is the neighbor node V j In the Hidden states (feature representations) of layers; It is the edge eigenvectors; It is the edge The fusion weights; the AGGREGATE function can be a weighted summation function, an attention-weighted average (such as GAT) function, or a pooling operation, etc.

[0108] (2) State update: node v i By combining the state of its parent layer and the aggregated messages, its current state is updated, which can be represented as follows:

[0109] The UPDATE function is typically a learnable neural network, such as a fully connected layer followed by an activation function. Initial state That is, the original feature vector of the node.

[0110] After L layers of iteration, each node finally obtains a global representation that contains its K-hop neighbor topological relationships and interference context. .

[0111] After obtaining the final representations of all nodes, a policy generation head network (such as a fully connected network) can be used for the disturbed nodes (base stations). v i Decode the specific optimization actions a i :

[0112] in, This is the power domain adjustment amount; For airspace adjustment; Provides time-domain adjustment suggestions. This strategy is collaborative because each node... v i Antenna parameter information a i Based on its neighbor status The method, derived jointly, can solve the "prisoner's dilemma" in distributed decision-making and avoid suboptimal global solutions caused by single-site optimization.

[0113] To achieve self-evolution and continuous optimization of graph neural networks, this disclosure employs a deep reinforcement learning (DRL) closed loop that closely interacts with the environment. This framework uses a constructed dynamic perturbation propagation graph G. t The input is used to generate and evaluate collaborative strategies through an Actor-Critic architecture, constantly balancing between "exploration" and "exploitation" to eventually converge to the long-term optimal interference governance strategy.

[0114] For example, the maximum entropy reinforcement learning SAC framework can be used, which adds the policy entropy maximization objective to the standard Actor-Critic framework, encourages exploration, and performs more stably and efficiently in high-dimensional continuous action spaces such as complex wireless environments.

[0115] The network architecture of this maximum entropy reinforcement learning SAC framework may include: an actor network. Critic Network And Value Network Among them, the actor network It can be the policy generator of the aforementioned GNN, with input state s t (Dynamic interference propagation diagram G) t The last layer of the GNN outputs a random policy, meaning it outputs parameters for the action distribution (such as the mean of a Gaussian distribution) for each base station. and standard deviation The actor network It can be represented as follows: The actual actions performed. a t Sampling from this distribution introduces an exploratory element to the strategy.

[0116] Critics Network It can be used as a valuation network to evaluate the state s t Next action a t The long-term expected return. To improve stability, two independent Critic networks are typically used ( And take the smaller value as the update target.

[0117] Value Network It can be a state-valued function network to evaluate state s t Its own long-term value. It learns by minimizing the error between itself and the soft value objective derived by the Critic and Actor.

[0118] The core objective of SAC is to maximize the expected cumulative reward with entropy regularization:

[0119] Where H is the entropy of the policy, used to measure randomness; Temperature parameter, and the importance of automatically adjusting the entropy term.

[0120] Empirical tuples can be sampled from the playback buffer D. Perform batch training. The loss function may include: Critic loss L. critic Actor loss L actor Temperature parameters The adaptive loss. Wherein, the Critic loss L... critic The two Critic networks can be updated by minimizing the temporal difference (TD) error, optionally with the Critic loss L. critic It can be represented as follows: , Among them, the target for: , For the target network parameters, through soft updates Stable learning.

[0121] By minimizing the Actor loss L actor To update the Actor (GNN) parameters θ, actions that achieve high Q-values ​​while maintaining a degree of randomness can be selected. As an example, the Actor loss L... actor The formula can be expressed as follows:

[0122] It is worth noting that in order to maintain the entropy value at the target level Temperature parameters α It can also be optimized; as an example, the temperature parameter The adaptive loss can be expressed as follows: .

[0123] By alternately minimizing the aforementioned loss, the Actor-Critic framework drives the GNN strategy. Continuously evolve towards achieving higher and more stable long-term rewards.

[0124] Reward function R t It serves as the "command stick" guiding self-learning evolution. This disclosure allows for the design of a reward function that integrates five objectives, ensuring the comprehensiveness of the optimization direction. This multi-objective reward function can be expressed as follows:

[0125] in, These are adjustable weights used to balance the priorities of different optimization objectives; The function is a function to suppress interference and reward. A reward function for increasing throughput; For power efficiency reward function; For user experience reward function; This is the oscillation penalty function.

[0126] The interference suppression reward function measures the percentage decrease in average interference noise (IN) in the interference dynamic graph. Negative values ​​indicate increased interference, while positive values ​​indicate mitigation. As an example, this interference suppression reward function can be expressed as follows: .

[0127] Throughput Improvement Reward Function: To encourage growth in total network throughput (TPut), a logarithmic form is used to smooth the impact of large numerical changes. As an example, the throughput improvement reward function can be expressed as follows: .

[0128] Power efficiency reward function: Evaluates improvements in energy efficiency (bit / joule) and encourages reductions in total power consumption while increasing or maintaining throughput. As an example, the power efficiency reward function can be expressed as follows: .

[0129] User experience reward function: Measured by user-level metrics (QoE), balancing throughput gains and latency penalties. As an example, the user experience reward function can be expressed as follows: .

[0130] Oscillation penalty function: This function penalizes drastic fluctuations in base station parameters within adjacent decision periods, ensuring network stability. As an example, the oscillation penalty function can be expressed as follows: .

[0131] This disclosure enables online evolution of the GNN decision model using the SAC algorithm, ensures comprehensive optimization directions through a multi-objective reward function, and constructs an atmospheric duct interference optimization device based on the aforementioned learning framework. This device incorporates real-time monitoring and safety mechanisms to guarantee reliable deployment in production environments. This allows the system to not only respond to interference in real time but also continuously learn and approach the optimal adjustment scheme for atmospheric duct interference. To ensure the system's robustness in production environments, this disclosure establishes a multi-dimensional monitoring system within the device, including: a multi-dimensional KPI tracking dashboard, model performance monitoring, safety constraints, and rollback mechanisms.

[0132] The multi-dimensional KPI tracking dashboard includes dimensions such as interference, service quality, energy efficiency, and stability. Interference dimension: It calculates the `InterferenceScore` to monitor its deviation from the target value. Service quality dimension: It comprehensively calculates the `QoEScore` to provide a global understanding of user experience. Energy efficiency dimension: It can generate a real-time trend chart of `EnergyEfficiency` across the entire network. Stability dimension: It can statistically analyze the frequency and magnitude of parameter change requests.

[0133] Model performance monitoring can include prediction accuracy, policy stability, and the health of the learning process. Specifically, prediction accuracy can be assessed by comparing the GNN's predictions of the disturbance map state at the next time step. To assess its perception and modeling capabilities by comparing it with the actual y, the comparison formula can optionally be expressed as follows: .

[0134] Policy stability: This assesses the volatility of the policy itself, preventing network instability caused by drastic policy changes. Optionally, the assessment formula can be expressed as follows: .

[0135] Learning process health: Monitor reward curves, Critic loss, policy entropy, etc., to determine whether the learning is converging or diverging.

[0136] It should be noted that security constraints can refer to action boundary constraints, such as using activation functions like `tanh` at the output layer of the Actor network to strictly limit actions to the adjustment range allowed by the network management system (e.g., power adjustment ±10dB). Rollback mechanisms can refer to setting hard thresholds for key indicators such as decreased throughput or increased call drop rate. Once triggered, learning is immediately paused, and the policy is rolled back to the previous security version, or switched to a rule-based conservative policy. This ensures the robustness and engineering feasibility of deploying the intelligent system in the live network and has high practical value.

[0137] Figure 4 This is a block diagram illustrating an atmospheric waveguide interference optimization device according to an exemplary embodiment. Figure 4 As shown, the atmospheric waveguide interference optimization device may include: a data acquisition module 401, a graph construction module 402, an inference module 403, and an execution module 404.

[0138] The data acquisition module 401 is used to acquire multi-source data, which includes at least remote interference management (RIM) data, base station-related data, and environmental data.

[0139] The graph construction module 402 is used to dynamically construct an interference propagation graph based on RIM data, base station related data and environmental data. Each node in the interference propagation graph corresponds to a base station, and each edge in the interference propagation graph represents the interference caused by the interfering base station to the interfering base station.

[0140] The inference module 403 is used to utilize the inference capability of the graph neural network to output a set of co-optimized wireless parameter information for the disturbed and / or disturbing base station, taking the interference propagation graph as input.

[0141] Execution module 404 is used to execute wireless parameter information to optimize atmospheric duct interference.

[0142] In some embodiments, the graph construction module 402 is used to: determine the nodes and node characteristics in the interference propagation graph based on base station related data and environmental data; determine the interference relationship between base stations based on RIM data, and determine each edge in the interference propagation graph and the comprehensive weight of each edge based on the interference relationship and RIM data; the topology of the interference propagation graph and the comprehensive weight of the edges evolve dynamically over time.

[0143] In some embodiments, the graph construction module 402 is further configured to: smooth the overall weight of the edges to be evolved by using an exponentially weighted moving average when the overall weight of the edges in the interference propagation graph evolves dynamically over time.

[0144] In some embodiments, the graph construction module 402 is used to: treat the interference relationship between base stations as edges in the interference propagation graph; for each edge in the interference propagation graph, obtain the intensity of the interference power associated with the edge from the RIM data, determine the interference intensity factor, and the interference intensity factor is used to quantify the severity of the interference; analyze the dynamic data and historical interference power data of the nodes associated with the edge to obtain the time-series stability index and service mode matching degree, and determine the time-domain stability factor based on the time-series stability index and service mode matching degree; evaluate the reliability and stability of the interference path corresponding to the edge under atmospheric waveguide conditions using the geographical location and antenna parameter information and environmental data of the nodes associated with the edge to obtain the path reliability factor; and fuse the interference intensity factor, the time-domain stability factor and the path reliability factor to obtain the comprehensive weight of the edge.

[0145] In some embodiments, the graph construction module 402 is used to: perform reliability assessment based on the number of successful interference predictions and the total number of interference events to obtain a reliability assessment index based on historical data; predict the probability of atmospheric duct occurrence based on the environmental data of the edge-associated nodes and use the probability of atmospheric duct occurrence as a path reliability index based on weather data; perform topology stability analysis based on the geographical location and antenna parameter information of the edge-associated nodes to obtain a topology stability index; and fuse the reliability assessment index based on historical data, the path reliability index based on weather data, and the topology stability index to obtain a path reliability factor.

[0146] In some embodiments, such as Figure 5 As shown, the atmospheric waveguide interference optimization device may further include a training module 505. This training module 505 is used to: use a graph neural network as a policy network and combine it with a value network, employing a maximum entropy reinforcement learning (SAC) framework to train the graph neural network online; wherein the graph neural network includes multiple message-passing layers, and through these layers, the graph neural network enables nodes in the graph to aggregate information from their neighbors and connected edges, thereby learning deep patterns inherent in the network topology and node relationships. Figure 5 Medium 501-504 and Figure 4 401-404 have the same function and structure.

[0147] In some embodiments, the reward function used by the maximum entropy reinforcement learning SAC framework is a multi-objective reward function, which includes at least an interference suppression reward function, a throughput improvement reward function, a power efficiency reward function, a user experience reward function, and an oscillation penalty function.

[0148] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0149] According to embodiments of this disclosure, this disclosure also provides an electronic device and a readable storage medium.

[0150] like Figure 6 The diagram shown is a block diagram of an electronic device according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as wearable devices (e.g., smartwatches, fitness trackers, etc.) and servers (e.g., digital human servers). The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0151] like Figure 6As shown, the electronic device includes one or more processors 601, a memory 602, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise as required. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 6 Take the 601 processor as an example.

[0152] The memory 602 is the non-transitory computer-readable storage medium provided in this disclosure. The memory stores instructions executable by at least one processor to cause the at least one processor to perform the atmospheric duct interference optimization method provided in this disclosure. The non-transitory computer-readable storage medium of this disclosure stores computer instructions for causing a computer to perform the atmospheric duct interference optimization method provided in this disclosure.

[0153] Memory 602, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the atmospheric waveguide interference optimization method in the embodiments of this disclosure (e.g., appendix). Figure 4 The data acquisition module 401, graph construction module 402, inference module 403, and execution module 404 shown are attached. Figure 5 The data acquisition module 501, graph construction module 502, inference module 503, execution module 504, and training module 505 are shown. The processor 601 executes various server functions and data processing by running non-transient software programs, instructions, and modules stored in the memory 602, thereby implementing the atmospheric waveguide interference optimization method in the above method embodiments.

[0154] Memory 602 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, memory 602 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 602 may optionally include memory remotely located relative to processor 601, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0155] The electronic device may also include an input device 603 and an output device 604. The processor 601, memory 602, input device 603, and output device 604 can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.

[0156] Input device 603 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the electronic device, such as touch screens, keypads, mice, trackpads, touchpads, joysticks, one or more mouse buttons, trackballs, joysticks, etc. Output device 604 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The display device may include, but is not limited to, liquid crystal displays (LCDs), light-emitting diode (LED) displays, and plasma displays. In some embodiments, the display device may be a touch screen.

[0157] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, application-specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0158] These computational programs (also referred to as programs, software, software applications, or code) include machine instructions for a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) used to provide machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0159] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0160] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet, and blockchain networks.

[0161] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0162] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0163] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for optimizing atmospheric waveguide interference, characterized in that, include: Acquire multi-source data, which includes at least remote interference management (RIM) data, base station-related data, and environmental data; Based on the RIM data, the base station related data, and the environmental data, an interference propagation graph is dynamically constructed, wherein each node in the interference propagation graph corresponds to a base station, and each edge in the interference propagation graph represents the interference caused by the interfering base station to the interfering base station. Using the reasoning capability of graph neural networks, and taking the interference propagation graph as input, a set of collaboratively optimized wireless parameter information is output for the disturbed and / or interfering base stations. The aforementioned wireless parameter information is applied to optimize atmospheric ducting interference.

2. The method according to claim 1, characterized in that, The step of dynamically constructing an interference propagation map based on the RIM data, the base station-related data, and the environmental data includes: Based on the base station-related data and the environmental data, the nodes and node characteristics in the interference propagation map are determined; The interference relationship between base stations is determined based on the RIM data, and each edge and its comprehensive weight in the interference propagation graph are determined based on the interference relationship and the RIM data; the topology of the interference propagation graph and the comprehensive weight of the edges evolve dynamically over time.

3. The method according to claim 2, characterized in that, The method further includes: When the overall weight of the edges in the interference propagation graph evolves dynamically over time, an exponentially weighted moving average is used to smooth the overall weight of the edges to be evolved.

4. The method according to claim 2 or 3, characterized in that, The step of determining each edge and its overall weight in the interference propagation graph based on the interference relationship and the RIM data includes: The interference relationships between the base stations are used as edges in the interference propagation graph; For each edge in the interference propagation graph, the intensity of the interference power associated with the edge is obtained from the RIM data, and an interference intensity factor is determined, which is used to quantify the severity of the interference. By analyzing the dynamic data and historical interference power data of the nodes associated with the edge, a time-series stability index and a service mode matching degree are obtained, and a time-domain stability factor is determined based on the time-series stability index and the service mode matching degree. Using the geographical location and antenna parameter information of the nodes associated with the edge, as well as environmental data, the reliability and stability of the interference path corresponding to the edge under atmospheric waveguide conditions are evaluated to obtain the path reliability factor. The interference intensity factor, the temporal stability factor, and the path reliability factor are fused together to obtain the comprehensive weight of the edge.

5. The method according to claim 4, characterized in that, The method of utilizing the geographical location and antenna parameter information of the nodes associated with the edge, as well as environmental data, to evaluate the reliability and stability of the interference path corresponding to the edge under atmospheric waveguide conditions, in order to obtain the path reliability factor, includes: Reliability assessment is performed based on the number of successful interference predictions and the total number of interference events to obtain reliability assessment indicators based on historical data. Based on the environmental data of the nodes associated with the edge, the probability of atmospheric waveguide occurrence is predicted, and the probability of atmospheric waveguide occurrence is used as a path reliability index based on weather data. Based on the geographical location and antenna parameter information of the nodes associated with the edge, a topology stability analysis is performed to obtain a topology stability index; The reliability assessment index based on historical data, the path reliability index based on weather data, and the topology stability index are fused together to obtain the path reliability factor.

6. The method according to claim 1, characterized in that, The method further includes: The graph neural network is used as a policy network and combined with a value network. The graph neural network is trained online using the maximum entropy reinforcement learning (SAC) framework. The graph neural network includes multiple message passing layers. Through the message passing layers, the graph neural network enables nodes in the graph to aggregate information from their neighbors and connected edges, so as to learn the deep patterns contained in the network topology and node relationships.

7. The method according to claim 6, characterized in that, The reward function used in the maximum entropy reinforcement learning SAC framework is a multi-objective reward function, which includes at least an interference suppression reward function, a throughput improvement reward function, a power efficiency reward function, a user experience reward function, and an oscillation penalty function.

8. An atmospheric waveguide interference optimization device, characterized in that, include: The data acquisition module is used to acquire multi-source data, which includes at least Remote Interference Management (RIM) data, base station-related data, and environmental data. The graph construction module is used to dynamically construct an interference propagation graph based on the RIM data, the base station related data, and the environmental data. Each node in the interference propagation graph corresponds to a base station, and each edge in the interference propagation graph represents the interference caused by the interfering base station to the interfering base station. The inference module is used to utilize the inference capability of the graph neural network, taking the interference propagation graph as input, to output a set of collaboratively optimized wireless parameter information for the disturbed and / or interfering base station; An execution module is used to execute the wireless parameter information to optimize atmospheric duct interference.

9. An electronic device, characterized in that, include: One or more processors; The processor is used to invoke instructions to cause the electronic device to perform the method of any one of claims 1-7.

10. A storage medium storing instructions, characterized in that, When the instructions are executed on an electronic device, the electronic device causes the electronic device to perform the method of any one of claims 1-7.

11. A program product comprising at least one of a program and instructions, characterized in that, When at least one of the programs or instructions is executed by an electronic device, it implements the steps of the method according to any one of claims 1-7.

Images