Method and apparatus for regionalization of propagation mechanisms for hyper-massive mimo systems

Abstract

The application provides a method and device for dividing a propagation mechanism area of a super large-scale MIMO system. The method provided by the application comprises: acquiring channel delay domain data corresponding to a dual polarization mode of a MIMO system in a dynamic indoor scene with a receiving end moving; constructing a dynamic time delay heat map with a transmitting end antenna serial number and a measurement time as two-dimensional dimensions based on the channel delay domain data, wherein the dynamic time delay heat map represents delay distribution differences in different polarization modes; performing noise smoothing and edge preserving processing on the dynamic time delay heat map, and then performing area segmentation, boundary positioning of different propagation mechanisms, edge detection and morphological optimization; extracting delay distribution features of the segmented areas for unsupervised quantitative clustering to obtain different time delay levels; and based on a combination mode of the time delay levels, combining the physical nature of the propagation mechanism, dividing an occluded line of sight into a transition state subclass to obtain a classified propagation mechanism.

Description

Technical Field

[0001] This application relates to the field of propagation mechanism region division technology, and in particular to a method and apparatus for propagation mechanism region division in ultra-large-scale MIMO systems. Background Technology

[0002] With the evolution of sixth-generation mobile communication technology, ultra-large-scale MIMO systems have achieved higher spectral efficiency and communication capacity through ultra-large-scale antenna arrays, becoming one of the core technologies supporting ultra-dense communication scenarios. Precise division of the propagation mechanism is a key prerequisite for ultra-large-scale MIMO systems to optimize beamforming, adapt polarization strategies, and improve the robustness of communication in dynamic scenarios.

[0003] Currently, the classification of propagation mechanisms in MIMO systems mainly relies on two approaches: one is a coarse classification based on physical scenarios, which simply determines LOS (line-of-sight) or NLOS (non-line-of-sight) based on the spatial location of the receiver (RX) or the presence of significant obstructions; the other is a classification based on static scenarios, which collects data by fixing the RX position. While some studies mention OLOS (obstructed line-of-sight), they do not provide a detailed classification, remaining only at the binary LOS / NLOS level. Existing classification methods have the following shortcomings: First, the classification dimension is singular, relying solely on RX position and ignoring the heterogeneity of VMIMO arrays. Propagation paths of different transmitter (TX) antennas at the same RX position may differ significantly, leading to misjudgments of propagation mechanisms in local areas. Second, adaptability is limited to static scenarios, failing to capture the dynamic switching of propagation mechanisms over time and space when the RX moves, resulting in a disconnect from actual communication environments. Third, the classification granularity is coarse, failing to subdivide the OLOS transition states, which account for a significant proportion in indoor scenarios, making it difficult to support refined polarization adaptation and channel optimization strategies, leading to insufficient communication continuity and stability of the system in complex indoor environments.

[0004] Therefore, there is an urgent need for a method that can adapt to the dynamic indoor scenarios of ultra-large-scale MIMO systems, taking into account the heterogeneity of multiple antennas and the fine-grained regional division of propagation mechanisms, in order to solve the problems of inaccurate division, poor adaptability and coarse granularity of existing methods. Summary of the Invention

[0005] In view of this, this application provides a method and apparatus for dividing the propagation mechanism region of a very large-scale MIMO system, which is adapted to the dynamic indoor scene of a very large-scale MIMO system, takes into account the heterogeneity of multiple antennas and the fine division of the propagation mechanism region, so as to solve the problems of inaccurate division, poor adaptability and coarse granularity of existing methods.

[0006] Specifically, this application is implemented through the following technical solution:

[0007] The first aspect of this application provides a method for partitioning the propagation mechanism region in a very large-scale MIMO system, the method comprising:

[0008] Acquire channel delay domain data corresponding to dual-polarization mode in a dynamic indoor scenario where the receiver of the MIMO system is moving.

[0009] Based on the channel delay domain data, a dynamic delay heatmap is constructed with the transmitter antenna number and measurement time as two dimensions. The dynamic delay heatmap represents the delay distribution differences under different polarization modes.

[0010] The dynamic time-delay heatmap is first subjected to noise smoothing and edge preservation processing, and then region segmentation is performed through edge detection and morphological optimization to locate the boundaries of different propagation mechanisms;

[0011] The delay distribution features of the segmented regions are extracted and unsupervised quantitative clustering is performed to obtain different delay levels;

[0012] Based on the combination patterns of the aforementioned latency levels, and combined with the physical nature of the propagation mechanism, the occlusion distance is divided into transitional subclasses, resulting in the classified propagation mechanism.

[0013] The second aspect of this application provides a region partitioning device for the propagation mechanism of an ultra-large-scale MIMO system, the device comprising an acquisition module, a construction module, a segmentation module, a clustering module, and a classification module;

[0014] The acquisition module is used to acquire channel delay domain data corresponding to the dual-polarization mode of the MIMO system in a dynamic indoor scenario where the receiver moves.

[0015] The construction module is used to construct a dynamic delay heatmap with the transmitter antenna number and measurement time as two dimensions based on the channel delay domain data. The dynamic delay heatmap represents the delay distribution differences under different polarization modes.

[0016] The segmentation module is used to first perform noise smoothing and edge preservation processing on the dynamic time-delay heatmap, and then perform region segmentation through edge detection and morphological optimization to locate the boundaries of different propagation mechanisms.

[0017] The clustering module is used to extract the delay distribution features of the segmented regions and perform unsupervised quantized clustering to obtain different delay levels;

[0018] The classification module is used to classify the occlusion distance into transitional subclasses based on the combination pattern of the delay level and the physical nature of the propagation mechanism, thereby obtaining the classified propagation mechanism.

[0019] The method and apparatus for classifying propagation mechanism regions in ultra-large-scale MIMO systems provided in this application achieve propagation mechanism region classification by constructing a complete logical chain of "delay domain statistical characteristics → region segmentation → delay level clustering → propagation mechanism mapping". Specifically, by acquiring dual-polarized channel delay domain data and constructing a dynamic delay heatmap, the method accurately captures the delay distribution differences of different antennas and times in ultra-large-scale MIMO systems under dynamic indoor scenarios, providing basic data support covering the polarization dimension for subsequent analysis. Through noise smoothing, edge detection, and morphological optimization of the heatmap for region segmentation, the method specifically addresses the problem of strong correlation between noise and edges in the delay heatmap of ultra-large-scale MIMO systems, accurately locating the delay abrupt boundary (i.e., the transition zone for propagation mechanism switching), clarifying "where the propagation mechanism may be different," and avoiding mechanism region confusion caused by dynamic scenarios and antenna heterogeneity. By extracting... Delay distribution characteristics are used for unsupervised clustering to obtain delay levels, quantifying the degree of delay difference in segmented regions and addressing the question of "how large is the delay difference in these regions?" Delay characteristics are standardized using high / medium / low delay levels to eliminate subjective judgment bias. By combining delay level combinations with physical characteristics to classify propagation mechanisms and OLOS transition state subclasses, the quantified delay levels are accurately mapped to the physical characteristics of propagation mechanisms (LOS direct path, NLOS multipath diffraction, OLOS hybrid propagation). This not only achieves basic LOS / NLOS classification but also refines OLOS subclasses to be line-of-sight biased, typical, and non-line-of-sight biased, breaking the limitations of traditional binary classification. This series of steps works in tandem, utilizing delay statistical characteristics to build a quantitative bridge from data to mechanism. Furthermore, through pre-processing of region segmentation and hierarchical clustering, it addresses the core challenges of antenna heterogeneity and time-varying dynamic scenarios in ultra-large-scale MIMO systems. Ultimately, it achieves refined, objective, and dynamic classification of propagation mechanism regions, providing precise mechanism region basis for polarization adaptation and beamforming optimization in ultra-large-scale MIMO systems. Attached Figure Description

[0020] Figure 1 A flowchart of the region partitioning method for the propagation mechanism of a super-large-scale MIMO system provided in Embodiment 1 of this application;

[0021] Figure 2 This is a schematic diagram of the structure of the region division device for the propagation mechanism of the ultra-large-scale MIMO system provided in Embodiment 2 of this application. Detailed Implementation

[0022] 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 numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0023] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0024] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0025] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0026] Figure 1 The flowchart illustrates the region partitioning method for the propagation mechanism of an ultra-large-scale MIMO system provided in Embodiment 1 of this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0027] S101. Obtain the channel delay domain data corresponding to the dual-polarization mode in a dynamic indoor scenario where the MIMO system is moving at the receiver.

[0028] Specifically, the MIMO system in this application refers to an ultra-large-scale multiple-input multiple-output (XL-MIMO) system, whose core features are "ultra-large-scale antenna configuration" and "dynamic scene adaptation capability". The transmitter deploys a 128-element ultra-large-scale antenna array, employing a 4×32 vertically arranged triangular periodic logarithmic antenna structure. Subarrays are divided by switching control, resulting in a significantly larger number of antennas than traditional large-scale MIMO, enabling higher beamforming gain and multi-user service capabilities. The receiver moves along a preset route, requiring the system to cope with the time-varying channel characteristics caused by user movement, indoor obstructions (such as pillars, desks, and chairs), and local scattering objects, rather than a fixed propagation environment in a static scenario. Due to the large number of antennas and wide coverage, the propagation paths of different transmitter antennas at the same receiver location may differ significantly (some antennas propagate at line-of-sight, while others propagate at obstructed line-of-sight or non-line-of-sight), resulting in obvious spatiotemporal non-stationarity of the channel.

[0029] Furthermore, dual-polarization mode refers to the antenna configuration at the receiver employing two orthogonal polarization methods to capture signals in different polarization states. Specifically, it includes two orthogonal modes: vertical polarization and horizontal polarization. The two broadband antennas at the receiver correspond to the two polarization methods respectively. Channel delay domain data refers to various statistical data reflecting the delay characteristics during signal propagation, specifically including: power delay spectrum (PDP) data (characterizing the power distribution characteristics of multipath signals with arrival time, which can intuitively reflect the quantity and intensity of multipath components), average delay data (the weighted average of the delays of all multipath signals, reflecting the overall delay level of the channel), and root mean square delay data (characterizing the dispersion of multipath signal delays, reflecting the temporal dispersion characteristics of the channel; the higher the value, the more significant the multipath effect).

[0030] In specific implementation, raw channel data in dual-polarization mode is collected; power delay spectrum data is extracted from the raw channel data; based on the power delay spectrum data, average delay data and root mean square delay data are calculated; and the average delay and root mean square delay data are integrated to form channel delay domain data.

[0031] Optionally, the raw channel data acquisition device includes a transmitter unit, a receiver unit, a channel detection unit, and a data processing unit. The transmitter unit employs a 128-element ultra-large-scale MIMO antenna array. The antenna array consists of triangular periodic logarithmic antennas arranged vertically in a 4×32 pattern. Eight switches control 32 antennas to form a square subarray. Antennas within each subarray are activated and numbered sequentially according to a preset order. The receiver unit deploys two broadband antennas with a frequency coverage range of 700MHz to 5GHz. The horizontal distance between the two antennas is 0.43m, and their height above the ground is 1.25m. Configured for vertical and horizontal polarization modes, and installed on a carrier device that moves along a preset route at a constant speed; the channel detection unit includes a transmitting module connected to the transmitting unit and a receiving module connected to the receiving unit. The transmitting module outputs a detection signal with a carrier frequency of 3.4 GHz and a bandwidth of 100 MHz, while the receiving module synchronously acquires the original channel signal in dual polarization mode; the data processing unit is used to perform hardware and system calibration on the original channel signal acquired by the receiving module, and then extracts the power delay spectrum, average delay, and root mean square delay data to form channel delay domain data and store it.

[0032] Specifically, firstly, a 128-element XL-MIMO antenna array is deployed at the transmitting end, and a dual-polarized broadband antenna is deployed at the receiving end (configured with vertical and horizontal polarization respectively, with a horizontal spacing of 0.43m and a height of 1.25m above the ground). The receiving end is installed on the carrier device and moves at a constant speed along a preset route. Hardware and system calibration is performed before data acquisition to eliminate equipment response interference. The transmitting module of the channel detector outputs a probe signal with a carrier frequency of 3.4GHz and a bandwidth of 100MHz. The receiving module synchronously acquires the original channel signal in dual-polarization mode. The acquisition process is time-synchronized using GPS. The original channel signal is demodulated and filtered using signal processing algorithms to extract the power delay spectrum data corresponding to each time node and each transmitting antenna element, and to determine the distribution of multipath signal power with arrival time. Based on the extracted power delay spectrum data, the average delay data is calculated by weighted averaging (the weight is the power of each multipath component), and the root mean square delay data is obtained by second-order moment operation. Finally, the processed average delay and root mean square delay data are summarized and integrated to form complete channel delay domain data and stored.

[0033] For example, in one embodiment, the power delay spectrum can be expressed as:

[0034] ;

[0035] in, Power delay spectrum; This represents the total number of multipath components. For the first The signal power corresponding to each multipath component; For the first Arrival time of each multipath component; This is the Dirac function.

[0036] The average delay can be expressed as:

[0037] ;

[0038] in, This is the average delay; This represents the total number of multipath components. For the first Power delay spectrum of each multipath component; For the first The arrival time of each multipath component.

[0039] The root mean square delay can be expressed as:

[0040] ;

[0041] in, For root mean square delay; This represents the total number of multipath components. For the first Power delay spectrum of each multipath component; For the first The arrival time of each multipath component.

[0042] The difference in average delay between dual-polarization modes can be expressed as:

[0043] ;

[0044] in, This represents the difference in average delay between the two polarization modes. The average delay under vertical polarization at the receiver; This represents the average delay under horizontal polarization at the receiver.

[0045] The difference in root mean square delay in dual-polarization mode can be expressed as:

[0046] ;

[0047] in, This represents the difference in root mean square delay between the two polarization modes. The root mean square delay under vertical polarization at the receiving end; This is the root mean square delay under horizontal polarization at the receiver.

[0048] The method provided in this embodiment, with its formulas for calculating power delay spectrum, average delay, and root mean square delay, is an adaptive choice targeting the core pain point of "ultra-large scale antenna + indoor dynamic scene" in ultra-large scale MIMO systems. For the pain point of large differences in propagation paths among different antennas under ultra-large scale antennas (128 elements), the power delay spectrum formula can quantify the multipath power distribution of a single antenna antenna antenna by antenna and time-by-time. The average delay and root mean square delay formulas can further extract the overall delay level and multipath dispersion of a single antenna, providing a precise delay characteristic data foundation for subsequent subdivision of regions by antenna number, avoiding misjudgments of propagation mechanisms due to antenna heterogeneity. This application specifically solves the special problems of antenna heterogeneity and time-varying dynamic scenes in ultra-large scale MIMO systems, ensuring the accuracy of data quantization and providing reliable feature support for refined and dynamic regional classification of propagation mechanisms, realizing the application of tools adapted to specific scenarios.

[0049] S102. Based on the channel delay domain data, construct a dynamic time delay heatmap with the transmitter antenna number and measurement time as two dimensions.

[0050] The dynamic delay heatmap represents the differences in delay distribution under different polarization modes.

[0051] Specifically, the dynamic delay heatmap is a two-dimensional visualization chart constructed with the transmitter antenna number as the horizontal dimension and the measurement time as the vertical dimension. It associates the average delay data and root mean square delay data under dual polarization modes (vertical polarization and horizontal polarization) with corresponding coordinate points, and presents the delay distribution of different antennas and different times through color mapping (high delay corresponds to bright color, and low delay corresponds to dark color). It can intuitively reflect the difference in delay values ​​under different polarization modes, and at the same time locate the areas of sudden delay changes.

[0052] In specific implementation, the average delay data and root mean square delay data under both vertical and horizontal polarization modes are separated from the channel delay domain data; a two-dimensional coordinate system is established with the transmitter antenna number as the horizontal dimension and the measurement time as the vertical dimension; the average delay data and root mean square delay data under the two polarization modes are respectively associated with the corresponding coordinate points of the two-dimensional coordinate system, and a heat map is generated through numerical visualization processing; the heat map is configured with color mapping, with different delay values ​​corresponding to different colors, to obtain a variety of dynamic delay heat maps.

[0053] Specifically, the average delay data and root mean square (RMS) delay data corresponding to vertical and horizontal polarization modes are filtered from the channel delay domain data. A two-dimensional coordinate system is constructed with the transmitter antenna number as the horizontal dimension and the measurement time as the vertical dimension to determine the mapping relationship between each coordinate point and a specific antenna number and time. The vertical polarization average delay data, horizontal polarization average delay data, vertical polarization RMS delay data, and horizontal polarization RMS delay data are matched to the corresponding coordinate points in the two-dimensional coordinate system. The values ​​of each coordinate point are converted into heat map elements using a data visualization tool. Color mapping is configured for the heat map, setting high delay values ​​to correspond to bright colors and low delay values ​​to low colors, ultimately generating four types of dynamic time delay heat maps: vertical polarization average delay heat map, horizontal polarization average delay heat map, vertical polarization RMS delay heat map, and horizontal polarization RMS delay heat map.

[0054] S103. The dynamic time-delay heatmap is first subjected to noise smoothing and edge preservation processing, and then region segmentation is performed through edge detection and morphological optimization to locate the boundaries of different propagation mechanisms.

[0055] Specifically, the propagation mechanism refers to the way a signal propagates from the transmitter to the receiver in a MIMO system. It typically includes three categories: line-of-sight (unobstructed direct path), non-line-of-sight (direct path completely blocked or surrounded by high-density scatterers), and obstructed line-of-sight (a transitional state between line-of-sight and non-line-of-sight). Locating the boundaries of different propagation mechanisms is crucial for identifying the dominant signal propagation mode within different regions. This provides a regional basis for subsequent implementation of differentiated communication optimization strategies (such as polarization adaptation and beamforming adjustment) for different propagation mechanisms, preventing communication quality degradation caused by the mixing of propagation mechanisms.

[0056] In practice, a Gaussian filtering algorithm is used to process the dynamic time delay heatmap to suppress noise interference from indoor scatterers while preserving edge details of abrupt changes in time delay values. The Canny edge detection algorithm is used to traverse the processed dynamic time delay heatmap, identify abrupt changes in time delay distribution, and initially locate the boundaries where the propagation mechanism may switch. Morphological closing operations are performed on the edge detection results to fill the gaps between regions and connect broken edges. Then, through contour extraction and smoothing, isolated edges and redundant points are removed to obtain structured segmented regions.

[0057] Specifically, a Gaussian filtering algorithm is used to filter the generated dynamic time-delay heatmap. Spatial and gray-scale standard deviation parameters are set to suppress noise interference introduced by indoor scatterers, while preserving the edge details of abrupt changes in time-delay values ​​in the dynamic time-delay heatmap. The Canny edge detection algorithm is then called to first smooth the filtered dynamic time-delay heatmap with Gaussian, then calculate the gradient magnitude and direction, and use a double threshold method to filter strong and weak edges, connect weak edges and eliminate false edges. The dynamic time-delay heatmap is traversed to identify abrupt changes in time-delay distribution and initially locate the boundaries where the propagation mechanism may switch. Morphological closing operations are performed on the edge images obtained by Canny edge detection. A 5×5 rectangular structuring element is selected, and the edges are first dilated and then eroded to fill the gaps between regions and connect broken edges. Then, the contours in the image after the closing operation are extracted using a contour extraction algorithm. Polynomial fitting is used to smooth the extracted contours, remove isolated small edges and redundant coordinate points, and finally obtain a structured segmented region.

[0058] The method provided in this embodiment preprocesses the dynamic delay heatmap using a Gaussian filtering algorithm. Leveraging its dual filtering characteristics in both the spatial and grayscale domains, it smooths out small-scale delay noise fluctuations caused by indoor scatterers while accurately preserving edge details of sudden delay changes. This solves the problem of delay edges being easily masked by noise due to multiple scattering in indoor channels of ultra-large-scale MIMO systems. The Canny edge detection algorithm then traverses the filtered heatmap to identify regions of sudden delay distribution changes, initially locating the propagation mechanism switching boundaries. Subsequently, morphological closing operations are used to fill gaps in the regions and connect broken edges. Contour extraction and smoothing processes are combined to remove isolated edges and redundant points, resulting in structured segmented regions. These methods are adaptations and improvements made to the "strong correlation between noise and edges" characteristic of dynamic delay heatmaps in ultra-large-scale MIMO systems. They eliminate interference from irrelevant noise on boundary identification and accurately restore the true edges of propagation mechanism switching. This provides an accurate regional division basis for subsequent extraction of delay distribution features based on segmented regions and refined classification of propagation mechanisms, avoiding misclassification due to blurred edges or noise interference, and ensuring the accuracy and reliability of propagation mechanism classification.

[0059] S104. Extract the delay distribution features of the segmented regions and perform unsupervised quantized clustering to obtain different delay levels.

[0060] Specifically, delay distribution features are core statistical features extracted from the channel delay data of segmented regions that reflect the temporal distribution patterns of multipath signals. These include key quantiles (10%, 30%, 50%, 70%, and 90%) of the cumulative distribution function derived from the power delay spectrum, as well as indicators such as average delay and root mean square delay. Unsupervised quantization clustering refers to the process of automatically grouping the extracted delay distribution features using clustering algorithms without requiring manual labeling. Delay levels are the result of unsupervised quantization clustering, specifically dividing the segmented regions into three levels: high, medium, and low, based on delay distribution features. Low delay levels correspond to line-of-sight propagation regions (dominated by direct paths, with low delay and weak dispersion); medium delay levels correspond to typical obstructed line-of-sight regions (a mixture of line-of-sight and non-line-of-sight, with moderate delay and dispersion); and high delay levels correspond to non-line-of-sight or partially non-line-of-sight obstructed line-of-sight regions (dominated by multipath scattering, with high delay and strong dispersion).

[0061] In specific implementation, based on the delay distribution feature data within the segmented region, the distribution pattern of the delay distribution feature data is analyzed through the probability density function to derive the cumulative distribution function; multiple key quantiles are extracted from the cumulative distribution function as core region features; the K-means unsupervised clustering algorithm is used to perform clustering operations on the core region features, iteratively optimizing the homogenization of intra-cluster characteristics and the heterogenization of inter-cluster characteristics, and dividing the segmented region into different delay levels.

[0062] Specifically, the average delay data, root mean square delay data, and delay distribution feature data derived from the power delay spectrum are extracted from each segmented region. All delay data from each segmented region are integrated into a one-dimensional dataset. Based on the one-dimensional dataset, a probability density function (PDF) is constructed using the kernel density estimation (KDE) method. A suitable kernel function (such as a Gaussian kernel) and bandwidth parameter are selected to fit the probability distribution shape of the delay data. The probability density function is integrated to obtain the cumulative distribution function (CDF). Furthermore, five key quantiles (10%, 30%, 50% (median), 70%, and 90%) are extracted from the CDF. These quantiles accurately characterize the central tendency, dispersion, and skewness of the delay distribution, collectively forming the core feature vector of each segmented region. The K-means unsupervised clustering algorithm is initialized, with the number of clusters preset to 3 (corresponding to high, medium, and low delay levels) according to the propagation mechanism classification requirements. Three samples are randomly selected from the core feature vectors of all segmented regions as initial cluster centers. The Euclidean distance between the core feature vector of each segmented region and the three initial cluster centers is calculated. Each segmented region is assigned to the nearest cluster. After the first round of sample allocation, the mean of all core feature vectors within each cluster is calculated to obtain new cluster centers. The Euclidean distance between the core feature vectors of all segmented regions and the new cluster centers is calculated again, and samples are reassigned to the nearest clusters. The iterative process of "sample allocation - updating cluster centers" is repeated, and the iteration termination condition is set to stop when the change in the position of the cluster centers is less than a preset threshold, or when the number of iterations reaches a preset maximum value (e.g., 100 times). After the iteration is completed, the three clusters are... The core features are statistically analyzed: clusters with low latency quantiles and small dispersion are defined as low latency level, corresponding to line-of-sight (LOS) dominated segmentation regions; clusters with medium latency quantiles and moderate dispersion are defined as medium latency level, corresponding to typical occluded line-of-sight (OLOS) segmentation regions; clusters with high latency quantiles and large dispersion are defined as high latency level, corresponding to non-line-of-sight (NLOS) or occluded line-of-sight biased segmentation regions; finally, the latency level classification results corresponding to all segmentation regions are output.

[0063] S105. Based on the combination mode of the aforementioned delay level and combined with the physical nature of the propagation mechanism, the occlusion line-of-sight distance is divided into transitional subclasses to obtain the classified propagation mechanism.

[0064] Specifically, the physical essence of a propagation mechanism refers to the path characteristics and signal behavior patterns determined by the propagation environment (such as obstructions and scattering objects) during the signal propagation from the transmitter to the receiver. It can be divided into three core essences. The physical essence of line-of-sight (LOS) propagation is that the signal has a direct propagation path, there are no obstructions between the transmitter and receiver, and there are few multipath components with energy mainly concentrated in the direct path, resulting in low delay and weak temporal dispersion (low root mean square delay). The physical essence of non-line-of-sight (NLOS) propagation is that the direct path is completely blocked, and the signal must reach the receiver through reflection, diffraction, or scattering by obstacles such as walls and furniture. It has abundant and chaotic multipath components, resulting in high delay and strong temporal dispersion (high root mean square delay). The physical essence of obstructed line-of-sight (OLOS) propagation is that the direct path is partially blocked (such as partial obstruction or near-field scattering interference). The signal simultaneously contains direct components and scattered / reflected components, representing a hybrid state of LOS and NLOS, with delay and dispersion levels between the two, and channel stability at an intermediate level.

[0065] Furthermore, the transitional subclass is a refined classification of the OLOS propagation mechanism, specifically including three categories. The physical essence of the line-of-sight biased OLOS subclass is that the direct path dominates, with only a small amount of scattering components, and the delay level is characterized by "low average delay + low-medium root mean square delay". The physical essence of the typical OLOS subclass is that the direct and scattering components are roughly equal in proportion, with no obvious dominant path, and the delay level is characterized by "medium average delay + medium root mean square delay", representing a typical mixed propagation state. The physical essence of the non-line-of-sight biased OLOS subclass is that scattering / reflection components dominate, with weak (or severely blocked) direct components, and the delay level is characterized by "medium-high average delay + high root mean square delay", with propagation characteristics closer to the non-line-of-sight mechanism.

[0066] In practical implementation, the physical propagation characteristics of various propagation mechanisms are analyzed to determine the unobstructed characteristics of direct paths at line-of-sight distances, the completely blocked characteristics of direct paths at non-line-of-sight distances, and the occlusion of line-of-sight distance as a transitional state. Three types of delay levels—average delay and root mean square delay—are identified as core combination dimensions. Combined with the transitional characteristics of occlusion of line-of-sight distances, three transitional state subclasses are defined: line-of-sight biased OLOS, typical OLOS, and non-line-of-sight biased OLOS. A correspondence rule between delay level combinations and propagation mechanisms is established. Based on this correspondence rule, the delay level combinations of each segmented region are matched and determined to obtain the propagation mechanism classification results.

[0067] Optionally, the corresponding rules are as follows: when the delay level combination is all at level 1, it is determined to be line-of-sight; when the delay level combination is all at level 3, it is determined to be non-line-of-sight; when the delay level combination is at level 1 and level 2, it is determined to be line-of-sight biased OLOS; when the delay level combination is at level 3 and level 1 or when the delay level combination is at level 2, it is determined to be typical OLOS; when the delay level combination is at level 3 and level 2, it is determined to be non-line-of-sight biased OLOS; the level 1, level 2, and level 3 are arranged in ascending order.

[0068] Specifically, we first analyze the physical propagation characteristics of the three basic propagation mechanisms. The core physical characteristic of line-of-sight (LOS) propagation is that there is an unobstructed direct propagation path between the transmitter and receiver, and the signal energy is mainly concentrated in the direct path with very few multipath components. The core physical characteristic of non-line-of-sight (NLOS) propagation is that the direct path is completely blocked by obstacles such as walls or large furniture, and the signal needs to propagate through reflection, diffraction, or scattering, resulting in abundant and chaotic multipath components. The core physical characteristic of obstructed line-of-sight (OLOS) propagation is that the direct path is partially obstructed (such as by partial obstruction or near-field scattering interference), and the signal contains both direct and scattered / reflected components, representing a transitional state between LOS and NLOS.

[0069] Furthermore, using the average latency level (first level, second level, third level, with the level and value increasing sequentially) and root mean square latency level (first level, second level, third level, with the level and value increasing sequentially) obtained from unsupervised clustering as the core combination dimension, and combining the transition characteristics of OLOS, OLOS is further subdivided into three transitional subclasses: line-of-sight biased OLOS, typical OLOS, and non-line-of-sight biased OLOS. The physical characteristics of each subclass are clarified: line-of-sight biased OLOS is dominated by direct components and supplemented by scattering components; typical OLOS has a relatively equal proportion of direct and scattering components; and non-line-of-sight biased OLOS is dominated by scattering components and has weak direct components.

[0070] Subsequently, a complete set of rules for the correspondence between latency level combinations and propagation mechanisms is established. If the average latency level of a segmented region is level 1 and the root mean square latency level is level 1, the region is determined to correspond to the line-of-sight (LOS) propagation mechanism. If the average latency level of a segmented region is level 3 and the root mean square latency level is level 3, the region is determined to correspond to the non-line-of-sight (NLOS) propagation mechanism. If the average latency level of a segmented region is level 1 and the root mean square latency level is level 2, or the average latency level is level 2 and the root mean square latency level is level 1, the region is determined to correspond to the biased line-of-sight mechanism. The OLOS transition state subclass is determined as follows: If the average delay level of a segmented region is level 3 and the root mean square delay level is level 1, or the average delay level is level 1 and the root mean square delay level is level 3, or both the average delay level and the root mean square delay level are level 2, then the region is determined to correspond to the typical OLOS transition state subclass; if the average delay level of a segmented region is level 3 and the root mean square delay level is level 2, or the average delay level is level 2 and the root mean square delay level is level 3, then the region is determined to correspond to the OLOS transition state subclass biased towards non-line-of-sight. Next, the average delay level and root mean square delay level of all structured segmented regions are extracted, and a unique delay level combination is generated for each segmented region (e.g., "average delay level 1 + root mean square delay level 2"). According to the preset correspondence rules, the delay level combination of each segmented region is matched with the rules one by one, and the propagation mechanism type corresponding to each segmented region is determined one by one (LOS, NLOS, line-of-sight biased OLOS, typical OLOS, non-line-of-sight biased OLOS). Finally, the determination results of all segmented regions are summarized to form a complete propagation mechanism classification result.

[0071] The method provided in this embodiment, firstly, offers an adaptive selection for the calculation formulas of average delay and root mean square delay, specifically addressing the core pain point of "ultra-large scale antennas + indoor dynamic scenarios" in ultra-large scale MIMO systems. For the pain point of large differences in propagation paths among different antennas under ultra-large scale antennas (128 elements), the power delay spectrum formula can quantify the multipath power distribution of a single antenna antenna antenna by antenna and time by time. The average delay and root mean square delay formulas can further extract the overall delay level and multipath dispersion of a single antenna, providing a precise delay characteristic data foundation for subsequent subdivision of regions by antenna number, avoiding misjudgments of propagation mechanisms due to antenna heterogeneity. This application specifically solves the special problems of antenna heterogeneity and time-varying dynamic scenarios in ultra-large scale MIMO systems, ensuring the accuracy of data quantization and providing reliable feature support for refined and dynamic regional classification of propagation mechanisms, thus realizing the application of tools adapted to specific scenarios.

[0072] Secondly, by selecting the 10%, 30%, 50%, 70%, and 90% quantiles in the cumulative distribution function as input features for K-means clustering, instead of the conventional mean and variance, we can accurately capture the "tail characteristics" and "asymmetric differences" of the delay distribution under different propagation mechanisms in ultra-large-scale MIMO systems: For the NLOS propagation mechanism, its high delay originates from the extreme delay values ​​of a few multipath diffraction paths, and the 90% quantile can effectively characterize this tail high delay feature; For the line-of-sight biased OLOS propagation mechanism, most paths exhibit low delay, and only a few scattering paths have medium delay, the 50% quantile can reflect the low delay level of the main body, and the 70% quantile can reflect the medium delay characteristics of the scattering paths, while the mean will mask this distribution heterogeneity and cannot distinguish the difference between "overall low delay + local medium delay" and "overall medium delay". This selection of quantile features is a customized improvement to address the asymmetric characteristics of multipath distribution in ultra-large-scale MIMO. It can objectively quantify the morphological differences in time delay distribution under different propagation mechanisms, avoid misjudgment of asymmetric distribution by mean features, and enable K-means clustering to more accurately classify time delay levels. This provides a reliable quantitative basis for subsequent matching of the physical nature of propagation mechanisms and subdividing OLOS transition state subclasses, ultimately improving the accuracy and refinement of propagation mechanism region classification.

[0073] Corresponding to the aforementioned embodiment of a method for partitioning the propagation mechanism region of an ultra-large-scale MIMO system, this application also provides an embodiment of a device for partitioning the propagation mechanism region of an ultra-large-scale MIMO system.

[0074] Figure 2 This is a schematic diagram of the structure of the region partitioning device for the propagation mechanism of the ultra-large-scale MIMO system provided in Embodiment 2 of this application. Please refer to... Figure 2 The apparatus provided in this embodiment includes an acquisition module 210, a construction module 220, a segmentation module 230, a clustering module 240, and a classification module 250.

[0075] The acquisition module 210 is used to acquire channel delay domain data corresponding to the dual polarization mode in a dynamic indoor scenario where the MIMO system moves at the receiving end.

[0076] The construction module 220 is used to construct a dynamic time delay heatmap with the transmitting antenna number and measurement time as two dimensions based on the channel delay domain data. The dynamic time delay heatmap represents the delay distribution differences under different polarization modes.

[0077] The segmentation module 230 is used to first perform noise smoothing and edge preservation processing on the dynamic time delay heatmap, and then perform region segmentation through edge detection and morphological optimization to locate the boundaries of different propagation mechanisms.

[0078] The clustering module 240 is used to extract the delay distribution features of the segmented regions and perform unsupervised quantitative clustering to obtain different delay levels.

[0079] The classification module 250 is used to classify the occlusion distance into transitional subclasses based on the combination mode of the delay level and the physical nature of the propagation mechanism, thereby obtaining the classified propagation mechanism.

[0080] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0081] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0082] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0083] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for area division of propagation mechanism in a hyper-massive MIMO system, characterized in that, The method comprises: obtaining channel delay domain data corresponding to a dual-polarization mode of a MIMO system in a dynamic indoor scene with a receiving end moving; based on the channel delay domain data, constructing a dynamic time delay heat map with the antenna serial number of the transmitting end and the measurement time as two-dimensional dimensions, the dynamic time delay heat map representing the delay distribution difference under different polarization modes; performing noise smoothing and edge preserving processing on the dynamic time delay heat map, and then performing region segmentation through edge detection and morphological optimization to locate the boundaries of different propagation mechanisms; extracting the delay distribution features of the segmented regions for unsupervised quantitative clustering to obtain different time delay levels; based on the combination mode of the time delay levels and the physical nature of the propagation mechanism, dividing the obstructed line of sight into transition state subclasses to obtain the classified propagation mechanism.

2. The method of claim 1, wherein, performing noise smoothing and edge preserving processing on the dynamic time delay heat map, and then performing region segmentation through edge detection and morphological optimization to locate the boundaries of different propagation mechanisms, including: using a Gaussian filter algorithm to process the dynamic time delay heat map to suppress noise interference caused by indoor scatterers while preserving the edge details of sudden changes in time delay values; using a Canny edge detection algorithm to process the processed dynamic time delay heat map to identify sudden change regions of time delay distribution and preliminarily locate the boundaries where the propagation mechanism may switch; performing morphological closing operation on the results obtained by edge detection to fill the gaps between regions and connect broken edges, and then removing isolated edges and redundant points through contour extraction and smoothing processing to obtain structured segmented regions.

3. The method of claim 1, wherein, extracting the delay distribution features of the segmented regions for unsupervised quantitative clustering to obtain different time delay levels, including: based on the delay distribution feature data in the segmented regions, analyzing the distribution form of the delay distribution feature data through a probability density function to derive a cumulative distribution function; extracting a plurality of key quantiles as core region features from the cumulative distribution function; using a K-means unsupervised clustering algorithm to perform clustering operation on the core region features to iteratively optimize the homogeneity of intra-cluster characteristics and the heterogeneity of inter-cluster characteristics, and divide the segmented regions into different time delay levels.

4. The method of claim 1, wherein, based on the combination mode of the time delay levels and the physical nature of the propagation mechanism, dividing the obstructed line of sight into transition state subclasses to obtain the classified propagation mechanism, including: analyzing the physical propagation characteristics of various propagation mechanisms to determine the direct path unobstructed characteristics of line of sight, the direct path completely blocked characteristics of non-line of sight, and the obstructed line of sight as a transition state; determining the three types of time delay levels of mean delay and root mean square delay as the core combination dimensions, combining the transition characteristics of the obstructed line of sight, and dividing the three types of transition state subclasses of OLOS biased towards line of sight, typical OLOS, and OLOS biased towards non-line of sight; establishing a corresponding rule between the combination of time delay levels and the propagation mechanism; according to the corresponding rule, matching and determining the time delay level combination of each segmented region to obtain the classification result of the propagation mechanism.

5. The method of claim 4, wherein, The corresponding rules are: when the time delay level combination is all first level, it is determined to be line of sight; when the time delay level combination is all third level, it is determined to be non-line of sight; when the time delay level combination is first level and second level, it is determined to be OLOS biased towards line of sight; when the time delay level combination is third level and first level or the time delay level combination is all second level, it is determined to be typical OLOS; when the time delay level combination is third level and second level, it is determined to be OLOS biased towards non-line of sight; the first level, the second level and the third level are arranged in ascending order.

6. The method of claim 1, wherein, The channel delay domain data includes mean delay and root mean square delay data.

7. The method of claim 1, wherein, The channel delay domain data corresponding to the dual-polarized mode in the dynamic indoor scene of the MIMO system moving at the receiving end is obtained, including: Collecting original channel data in the dual-polarized mode; Extracting power delay profile data from the original channel data; Based on the power delay profile data, mean delay data and root mean square delay data are calculated; Integrating the mean delay and root mean square delay data to form channel delay domain data.

8. The method of claim 1, wherein, Based on the channel delay domain data, a dynamic time delay heat map is constructed with the transmitting end antenna serial number and the measurement time as two-dimensional dimensions, including: Separating the mean delay data and the root mean square delay data in the vertical polarization and the horizontal polarization from the channel delay domain data; Establishing a two-dimensional coordinate system with the transmitting end antenna serial number as the horizontal dimension and the measurement time as the vertical dimension; Correlating the mean delay data and the root mean square delay data in the two polarization modes to the corresponding coordinate points of the two-dimensional coordinate system respectively, and generating a heat map through numerical visualization processing; Color mapping configuration is performed on the heat map, different delay values correspond to different colors, and a plurality of dynamic time delay heat maps are obtained.

9. The method of claim 7, wherein, The collecting device of the original channel data includes a transmitting end unit, a receiving end unit, a channel detection unit and a data processing unit. The transmitting end unit adopts a super large scale MIMO antenna array with 128 elements. The antenna array is composed of a triangular periodic logarithmic antenna arranged in 4×32 vertical mode. 32 antennas are controlled by 8 switches to form a square subarray. The antennas in each subarray are activated and numbered in a predetermined order. The receiving end unit is deployed with two broadband antennas with a frequency coverage range of 700MHz to 5GHz. The horizontal distance between the two antennas is 0.43m, and the height from the ground is 1.25m. The two antennas are configured in vertical polarization and horizontal polarization modes respectively, and are installed on a bearing device moving along a predetermined route with a constant moving speed. The channel detection unit includes a transmitting module connected with the transmitting end unit and a receiving module connected with the receiving end unit. The transmitting module outputs a detection signal with a carrier frequency of 3.4GHz and a bandwidth of 100MHz. The receiving module synchronously collects original channel signals in the dual-polarized mode. The data processing unit is used for hardware and system calibration of the original channel signals collected by the receiving module, and then extracts power delay profile, mean delay and root mean square delay data to form channel delay domain data and store it.

10. A region partitioning device for the propagation mechanism of an ultra-large-scale MIMO system, characterized in that, The device includes an acquisition module, a construction module, a segmentation module, a clustering module and a classification module. The acquisition module is configured to acquire channel delay domain data corresponding to a dual-polarization mode in a dynamic indoor scene in which a MIMO system moves at a receiving end; The construction module is configured to construct a dynamic time-delay heat map with a two-dimensional dimension of a transmitting-end antenna serial number and a measurement time based on the channel delay domain data, and the dynamic time-delay heat map represents delay distribution differences in different polarization modes; The segmentation module is configured to perform noise smoothing and edge preserving processing on the dynamic time-delay heat map, and then perform region segmentation through edge detection and morphological optimization to locate boundaries of different propagation mechanisms; The clustering module is configured to extract delay distribution features of segmented regions for unsupervised quantitative clustering to obtain different time-delay levels; The classification module is configured to divide an occluded line of sight into a transition state subclass based on a combination mode of the time-delay levels and a physical nature of the propagation mechanism to obtain a classified propagation mechanism.

Images