Channel information acquisition method and related apparatus

Abstract

The present application provides a channel information acquisition method and a related apparatus. The channel information acquisition method provided by the present application comprises: acquiring physical environment information comprising a point cloud density of sampling points; and acquiring an angle power spectrum on the basis of the physical environment information and a channel model, wherein the channel model is a high-precision NeRF model obtained by training on the basis of real physical environment information and a physical propagation law. An angular power spectrum of a multipath channel is acquired on the basis of a high-precision channel model, thereby improving the accuracy of wireless channel modeling.

Description

Channel information acquisition method and related devices

[0001] This application claims priority to Chinese Patent Application No. 202410992366.5, filed on July 22, 2024, entitled “Method and Apparatus for Acquiring Channel Information”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of communications, and in particular to a method and apparatus for acquiring channel information. Background Technology

[0003] As a crucial stage in the transition from 5G to 6G, 5.5G networks, compared to 5G networks, feature more antennas, higher frequency points, and a wider spectrum. For multi-antenna systems, by modeling the multipath propagation of wireless channels at different locations within the network, the capacity and transmission rate of the wireless communication system can be improved based on wireless channel modeling.

[0004] A wireless channel modeling method is as follows: a terminal device collects channel signals at different locations in space, and a neural radiation field model learns the signal attenuation and the intensity of the outgoing signal at different sampling points in space through measured signal data. Based on the trained neural radiation field model and the location information in space, the signal attenuation and signal intensity at any location in space are obtained for channel signal simulation at any location in space.

[0005] However, the accuracy of the above wireless channel modeling is low. Summary of the Invention

[0006] This application provides a channel information acquisition method and related apparatus to reduce the fitting difficulty of neural radiance fields (NeRF) models and improve the accuracy of wireless channel modeling.

[0007] In a first aspect, this application provides a method for obtaining channel information, the method comprising:

[0008] The system acquires physical environment information, including the point cloud density of sampling points. Based on the physical environment information and the channel model, it obtains the angular power spectrum. The channel model is a neural radiation field model trained based on the physical environment information and measured signal data. The neural radiation field model is used to extract attenuation information and weight information. The attenuation information is used to indicate the amount of attenuation of the signal transmitted by the network device at the sampling point, and the weight information is used to indicate the weight of the scattered signal transmitted along the emission angle at the sampling point.

[0009] The channel model incorporates real-world physical environment information during training, reducing the difficulty of fitting the NeRF model and improving the accuracy of the trained channel model. A high-precision channel model, based on physical environment information including point cloud density at sampling points, can obtain attenuation and weight information, and further determine the angular power spectrum at the sampling points, thus improving the accuracy of wireless channel modeling.

[0010] In some implementations, the angular power spectrum is obtained based on physical environment information and channel models, including:

[0011] The first dataset is input into the first network to obtain attenuation information and first information. The first dataset includes the point cloud density of the sampling points, and the first information is the information reflecting geographical features extracted by the first network based on the first dataset. The first network is a neural network used to extract environmental information. The second dataset is input into the second network to obtain weight information. The second dataset includes the first information, and the second network is a neural network used to extract weight information. The angular power spectrum is obtained based on the attenuation information and weight information, and the angular power spectrum satisfies the following relationship:

[0012] Among them, P RX P represents the geographical location of the receiving point. TX For the geographical location of network devices, This refers to the attenuation of the signal transmitted by the network device at the receiving point. P is the azimuth vector of the signal transmitted by the network device pointing to the receiving point. r Where ω is the geographical location of the sampling point, and W(P) is the emission angle from the sampling point to the receiving point. r ω) represents the weight of the scattered signal emitted at the sampling point along the emission angle, X(P) RX ) represents the angular power spectrum at the receiving point.

[0013] By obtaining attenuation and weight information through a channel model, and then obtaining the angular power spectrum based on the attenuation and weight information, the computational complexity is reduced and the efficiency of wireless channel modeling is improved.

[0014] Secondly, this application provides a model acquisition method, the method comprising:

[0015] The system acquires physical environment information, beam gain information, and measured signal data. The physical environment information includes the point cloud density of the sampling points. Based on the physical environment information, beam gain information, and measured signal data, the system trains a neural radiation field model to obtain a channel model. The neural radiation field model includes a first network and a second network. The first network is a neural network used to extract environmental information, and the second network is a neural network used to extract weight information. The weight information is used to indicate the weight of the scattered signal emitted at the sampling point along the emission angle.

[0016] The channel model is trained using real physical environment information, including the point cloud density of sampling points, which reduces the difficulty of model fitting and improves the accuracy of the channel model.

[0017] In some implementations, the neural radiation field model is trained based on physical environment information, beam gain information, and measured signal data to obtain the channel model, including:

[0018] Step 1: Input the first dataset into the first network to obtain attenuation information and first information. The first dataset includes the point cloud density of the sampling points. The attenuation information is used to indicate the amount of attenuation of the signal transmitted by the network device at the sampling points. The first information is the information reflecting geographical features extracted by the first network based on the first dataset.

[0019] Step 2: Input the second dataset into the second network to obtain weight information. The second dataset includes the first information.

[0020] Step 3: Determine the intensity of the scattered signal emitted by multiple beams at the sampling point along the emission angle based on the weighting information and beam gain information. The intensity of the scattered signal emitted by each beam at the sampling point along the emission angle satisfies the following relationship: S i (P x ,ω)=A i W(P x ,ω)

[0021] Among them, P x Let W(P) be the position of the sampling point, ω be the emission angle from the sampling point to the receiving point, and W(P) be the emission angle from the sampling point to the receiving point. x ω) represents the weight of the scattered signal emitted at the sampling point along the emission angle, and A i S is the beam gain matrix corresponding to each of the multiple beams. i (P x ,ω) represents the intensity of the scattered signal emitted by each of the multiple beams at the sampling point along the emission angle.

[0022] Step 4: Based on the attenuation information and the intensity of the scattered signals emitted by multiple beams at the sampling point along the emission angle, determine the received signal intensity at the receiving point.

[0023] Step 5: Calculate the training error based on the loss function, which includes the difference between the received signal strength at the receiving point and the measured signal data.

[0024] Step 6: Update the neural radiation field model based on the training error.

[0025] Step 7: If the training error does not converge or the training does not reach the maximum number of iterations, repeat the process from step 1.

[0026] Step 8: If the training error converges or the training reaches the maximum number of iterations, determine the neural radiation field model as a channel model.

[0027] In some implementations, the received signal strength at the receiving point satisfies the following relationship:

[0028] Among them, P RX P represents the geographical location of the receiving point. TX For the geographical location of network devices, This refers to the attenuation of the signal transmitted by the network device at the receiving point. P is the azimuth vector of the signal transmitted by the network device pointing to the receiving point. r For the geographical location of the sampling points, RSRP i (P RX ) represents the reference signal received power at the receiving point for each of the multiple beams.

[0029] The received signal is decomposed according to the propagation law, the direct path is rigorously calculated using deterministic physical propagation laws, and the NeRF model is used to focus on fitting the non-direct path. Through the above simulation calculations, the fitting difficulty in the NeRF model training process is reduced, and the accuracy of the channel model is improved.

[0030] In some implementations, the loss function may include at least one of the following: a structural similarity loss function among multiple beams, a sparsity loss function, or an electromagnetic propagation law loss function, wherein the structural similarity loss function is used to control the intensity order of the received signals of multiple beams at the receiving point, and the electromagnetic propagation law loss function is used to control the difference between the intensity of the outgoing signal and the intensity of the incoming signal at the sampling point.

[0031] The loss function incorporates a structural similarity loss function based on measured signal data, a sparsity loss function based on physical laws across the entire space, and an electromagnetic propagation law loss function. These loss functions constrain the training of the NeRF model, thereby improving the generalization ability of the NeRF model and increasing the accuracy of the channel model.

[0032] Thirdly, this application provides a channel information acquisition apparatus, which includes functional modules for implementing any of the channel information acquisition methods mentioned in the first aspect above, or includes functional modules for implementing any of the model acquisition methods mentioned in the second aspect above. In some implementations, each module can be implemented by software and / or hardware.

[0033] Fourthly, this application provides a channel information acquisition apparatus, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions. The processor executes the computer-executable instructions stored in the memory to implement any of the channel information acquisition methods mentioned in the first aspect above, or to implement any of the model acquisition methods mentioned in the second aspect above.

[0034] Fifthly, this application provides a computer-readable storage medium including a computer program that, when run on a computer, causes the computer to implement the methods of the first or second aspect and any possible implementation of the first or second aspect.

[0035] Sixthly, this application provides a computer program product comprising: a computer program (also referred to as code or instructions), which, when executed, causes the computer to perform the methods of the first or second aspect and any possible implementation thereof.

[0036] The third to sixth aspects of this application correspond to the technical solutions of the first or second aspects of this application. The beneficial effects achieved by each aspect and the corresponding feasible implementation are similar, and will not be described again. Attached Figure Description

[0037] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0038] Figure 1 is a schematic diagram of the signal propagation path in a radio frequency radiation field;

[0039] Figure 2 is a schematic diagram of the system applied to both road test and off-road test scenarios according to an embodiment of this application;

[0040] Figure 3 is a schematic diagram of the framework of the NeRF model structure provided in an embodiment of this application;

[0041] Figure 4 is a flowchart illustrating a method for obtaining a channel model based on a NeRF model according to an embodiment of this application;

[0042] Figure 5 is a schematic diagram of the superposition calculation of scattered signals provided in one embodiment of this application;

[0043] Figure 6 is a flowchart illustrating a channel information acquisition method provided in an embodiment of this application;

[0044] Figure 7 is a schematic block diagram of a channel information acquisition device provided in an embodiment of this application;

[0045] Figure 8 is a schematic diagram of the structure of an information acquisition device provided in another embodiment of this application.

[0046] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0047] 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 denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0048] It should be understood that in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the preceding and following related objects, but does not exclude the possibility of indicating an "and" relationship; the specific meaning can be understood in context. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Here, a, b, and c can be single or multiple.

[0049] In this application, the use of prefixes such as "first" and "second" is merely for the purpose of distinguishing and describing different things belonging to the same category, and does not constrain the order, size, or quantity of things. For example, "first parameter" and "second parameter" are simply different parameters, and there is no temporal or quantitative relationship between them.

[0050] This application will present various aspects, embodiments, or features relating to systems that may include multiple devices, components, modules, etc. It should be understood and appreciated that individual systems may include additional devices, components, modules, etc., and / or may not include all the devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these approaches are also possible.

[0051] Furthermore, in the embodiments of this application, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as an "example" in this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the term "example" is intended to present concepts in a concrete manner. In the embodiments of this application, "of," "corresponding, relevant," and "corresponding" may sometimes be used interchangeably, and it should be noted that their intended meanings are consistent unless their distinction is emphasized.

[0052] 5.5G, also known as 5G-Advanced (5G-A), aims to further enhance the performance, functionality, and application scenarios of 5G networks. 5.5G represents the next stage in the evolution of 5G technology and is generally considered a transitional technology between 5G and the future 6G. Compared to 5G networks, 5.5G networks have more antennas, higher frequency bands, and a wider spectrum.

[0053] A multi-antenna system typically implies multipath propagation. Multipath propagation refers to the phenomenon where radio waves, during their propagation, encounter reflections, scattering, and diffraction from obstacles such as buildings and terrain, causing the signal to reach the receiver along multiple paths. Multipath propagation leads to signal superposition, resulting in phase differences and amplitude variations, thus affecting the quality of signal reception.

[0054] Traditional antenna systems often focus on eliminating the negative effects of multipath propagation, such as signal fading and interference. However, multiple-input multiple-output (MIMO) systems use multiple antennas at both the transmitter and receiver, transmitting and receiving signals through different paths, which can then be combined at the receiver. By leveraging multipath propagation, MIMO systems can significantly improve channel capacity and communication rates without increasing bandwidth or transmit power.

[0055] Based on the characteristics of MIMO systems, in order to improve network coverage, capacity, and user experience, it is necessary to determine the multipath information of wireless channels at different locations in the network, thereby testing the impact of network parameters on network performance. Among these methods, constructing a multipath model of wireless channels based on the multipath information of wireless channels at different locations in the network is a necessary approach for network planning during the 5.5G network construction phase.

[0056] Neural radiance fields (NeRF) models are used in computer vision. Based on volume rendering theory, NeRF models extract volume density and color information from sampling points in a 3D environment through neural networks, enabling the generation of high-precision 3D models at arbitrary angles and distances. Existing technologies have transferred NeRF models to the field of channel modeling, allowing the acquisition of wireless channel multipath information. The specific acquisition method is illustrated below with reference to Figure 1.

[0057] Figure 1 illustrates the path of signal propagation in a radio frequency (RF) radiation field. As shown in Figure 1, this scenario includes a transmitter (TX) and a receiver (RX). The transmitter emits RF signals into the environment. Due to interference from obstacles in the environment, the RF signals are reflected, scattered, diffracted, or absorbed. In the environment to be modeled, space can be considered as a discretized finite number of three-dimensional sampling points. When the original RF signal arrives at a sampling point from all possible paths, that sampling point can be considered a new radiation source, which re-emits the scattered signal into space.

[0058] As shown in Figure 1, the geographical location in the figure is P. x The sampling point receives radio frequency (RF) signals from paths r1 and r2. At this point, the sampling point can be considered a new transmitting point, and it transmits RF signals again towards the receiving point along path r3. It should be noted that the sampling point is not an omnidirectional radiation source; instead, it emits scattered signals outwards at non-uniform angles. These scattered signals are the RF signals emitted by the sampling point and can therefore be called the outgoing signals at the sampling point. The scattered signal emitted by the sampling point along path r3 towards the receiving point is related to the geographical location P of the receiving point. RX The resulting angle is the emission angle ω of the scattered signal emitted at the sampling point, which can specifically include the azimuth and elevation angles. The intensity S(P) of the scattered signal emitted at the sampling point along the emission angle ω is... x ,ω) can be obtained through actual measurement.

[0059] When acquiring multipath information of a wireless channel using a NeRF model, the following steps are first taken: obtaining the geographical locations of multiple sampling points in the environment to be modeled, the geographical location of the transmitter (TX), the possible emission angles at each sampling point, and the measured intensity of the emitted scattered signal at each sampling point. Based on this information, the NeRF model is trained. The NeRF model can then learn the signal attenuation and the intensity of the emitted signal at different sampling points in the environment to be modeled. The signal attenuation indicates the amount of attenuation of the original radio frequency signal transmitted by the TX as it reaches the sampling point along different paths. After the NeRF model is trained, the geographical locations of the indicated sampling points, the TX, and the emission angles at the sampling points are used as inputs to the NeRF model. This allows the acquisition of the intensity of the emitted scattered signal at the indicated sampling points, thus enabling channel signal simulation of the environment to be modeled.

[0060] However, the NeRF model has a large number of parameters and relies on a large amount of measured data for model fitting during training. In practical applications, due to the limited amount of measured data, the accuracy of the trained NeRF model is limited. Therefore, the simulation accuracy of channel signals at new locations in the environment to be modeled based on the NeRF model is also limited, ultimately leading to low accuracy in wireless channel modeling.

[0061] In addition, the intensity of the scattered signal emitted at the sampling point obtained based on the NeRF model trained above can be regarded as the intensity of the scattered signal emitted after the superposition of multiple paths, for example, in Figure 1 at P x The intensity of the scattered signal emitted at the sampling point is equivalent to the intensity of the scattered signal after the original radio frequency signals on path r1 and path r2 are superimposed. The above method cannot calculate the angular power spectrum of the multipath channel, that is, it cannot calculate parameters such as the departure angle, arrival angle, and path strength of each path in the wireless channel, resulting in low accuracy in wireless channel modeling.

[0062] To address the aforementioned technical problems, this application provides a channel information acquisition method and related apparatus to reduce the fitting difficulty of the NeRF model and improve the accuracy of wireless channel modeling.

[0063] The technical concept of this application is as follows: During the training process of the NeRF model, 3D point cloud information of the environment to be modeled is introduced. By combining the real environment information with the NeRF model, a high-precision NeRF model can be trained, which can reflect the multipath situation in the environment to be modeled. Furthermore, based on the high-precision NeRF model, the angular power spectrum of the multipath channel is calculated, thereby improving the accuracy of wireless channel modeling.

[0064] Figure 2 is a schematic diagram of the system according to the embodiments of this application applied to both road test and off-road test scenarios. Figure 2 shows a schematic diagram of a possible, non-limiting system architecture. As shown in Figure 2, the system includes a terminal device, a radio access network (RAN) device, a geolocation module, an electronic map module, an engineering parameter information module, a wireless propagation modeling module, and an angle power spectrum storage module.

[0065] The electronic map module stores electronic map information, which includes geographic information corresponding to the environment to be modeled. It is understood that the electronic map can originate from commercial maps or be generated through surveying equipment such as LiDAR.

[0066] The engineering parameter information module is used to store engineering parameter information related to wireless channel modeling, such as the location of RAN equipment, antenna gain, antenna position, azimuth angle, downtilt angle, etc.

[0067] The wireless propagation modeling module is the core algorithm module of the system. Based on the measured data provided by the terminal equipment or RAN equipment, the engineering parameter information provided by the engineering parameter information module, and the electronic map information provided by the electronic map module, the wireless propagation modeling module can obtain the angular power spectrum of the multipath channel in the environment to be modeled and model the wireless propagation environment.

[0068] The angle power spectrum storage module is used to store the angle power spectrum output by the wireless propagation modeling module.

[0069] The wireless propagation modeling module can be implemented in software. In drive testing scenarios, it can be deployed on a standalone personal computer (PC) or a cloud server. In drive testing, the terminal device measures the reference signal received power (RSRP) of multiple downlink beams at sampling points along a predetermined acquisition route. This includes the RSRP corresponding to the channel state information (CSI) of multiple downlink beams, as well as the RSRP corresponding to the synchronization signal and physical broadcast channel block (SSB). Simultaneously, it records the latitude, longitude, and altitude information of each sampling point.

[0070] The information data measured and recorded by the terminal device is called drive test (DT) data. DT data is the measured data provided by the terminal device, and the wireless propagation modeling module uses DT data as input. It is understood that DT data can be stored in the terminal device or in the DT data storage module shown in Figure 2, and this application does not limit it in this way.

[0071] The terminal equipment in the system shown in Figure 2 can also be called a terminal, user equipment (UE), mobile station, mobile terminal, etc. As a device with wireless transceiver capabilities, a terminal device can be deployed on land, including indoors or outdoors, handheld, wearable, or vehicle-mounted; it can also be deployed on water (such as on ships); and it can be deployed in the air (such as on airplanes, balloons, and satellites). Terminal devices can be UEs, mobile phones, tablets, computers with wireless transceiver capabilities, virtual reality (VR) terminal devices, augmented reality (AR) terminal devices, wireless terminals in industrial control, vehicle-mounted terminal devices, wireless terminals in self-driving vehicles, wireless terminals in remote medical care, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, wearable terminal devices, etc. Terminal devices can be fixed or mobile.

[0072] In off-road testing scenarios, the wireless propagation modeling module can be deployed in the network management system or on edge nodes that can subscribe to measurement reports (MR) data from RAN devices. In off-road testing scenarios, after the terminal device completes measurements at sampling points on the predetermined acquisition line, it reports the measurement report (MR) to the RAN device. The MR data can be stored in the RAN device or in the MR data storage module shown in Figure 2; this application does not limit this.

[0073] It should be noted that MR data does not include the latitude, longitude, and altitude information recorded by the terminal device at the sampling point. Therefore, in non-road test scenarios, the geolocation module needs to determine the latitude, longitude, and altitude information at the sampling point based on the MR data, and use the latitude, longitude, and altitude information at the sampling point along with the MR data as input to the wireless propagation modeling module.

[0074] The RAN equipment in the system shown in Figure 2 can be a base station, an evolved NodeB (eNodeB), an access point (AP), a transmission reception point (TRP), a next-generation NodeB (gNB), a next-generation base station in a 6G mobile communication system, a base station in a future mobile communication system, or an access node in a Wi-Fi system. RAN nodes can be macro base stations, micro base stations or indoor stations, relay nodes or donor nodes, or radio controllers in a cloud radio access network (CRAN) scenario.

[0075] It should be noted that the wireless propagation modeling module is used to execute the channel information acquisition method provided in the embodiments of this application. The premise of acquiring the angular power spectrum is to train the NeRF model based on measured data, engineering parameter information and electronic map information to obtain the channel model. The method for acquiring the channel model is described below.

[0076] Figure 3 is a schematic diagram of the NeRF model structure provided in one embodiment of this application. As shown in Figure 3, the NeRF model may include a first network and a second network, both of which are multi-layer fully connected neural networks.

[0077] The first network is a neural network used to extract environmental information. As shown in Figure 3, the input to the first network is the geographical location information of the sampling points and the point cloud density of the sampling points. For example, in the road test scenario shown in Figure 2, when the terminal device measures the RSRP at the sampling points on the predetermined acquisition route, it records the latitude, longitude, and altitude information of the sampling points. The DT data input to the wireless propagation modeling module includes the latitude, longitude, and altitude information of the sampling points, which is the geographical location information of the sampling points.

[0078] In another example, as shown in Figure 2 in the non-road test scenario, the geolocation module can determine the latitude, longitude, and altitude information of the sampling point based on the MR data, and input the latitude, longitude, and altitude information of the sampling point as the geographical location information of the sampling point into the wireless propagation modeling module.

[0079] It should be noted that the electronic map information stored in the electronic map module is usually basic data reflecting geospatial information. To obtain the point cloud density of the sampling points, the electronic map can be sampled to obtain three-dimensional point cloud information. The three-dimensional point cloud information includes multiple three-dimensional coordinates, where an object exists at each three-dimensional coordinate location. For three-dimensional point cloud data, continuous point cloud data in space can be divided according to a certain spatial resolution (e.g., 2 meters (m) * 2m * 2m as the spatial resolution) to form a series of cubic grids. Each grid contains all point cloud data points within the spatial range of that grid, thereby calculating the point cloud density within the grid where the sampling point is located.

[0080] Understandably, the electronic map module can store basic data reflecting geospatial information, calculate the point cloud density of sampling points using the methods described above, or directly store the point cloud density of sampling points. In some implementations, the electronic map module can also output electronic map information to the wireless propagation modeling module, which then calculates and obtains the point cloud density of the sampling points.

[0081] The output of the first network consists of attenuation information and first information. The attenuation information indicates the amount of attenuation of the radio frequency signal transmitted by the network device, i.e., TX, at the sampling point. The first information is information reflecting geographical features extracted by the first network based on the geographical location information and point cloud density of the input sampling points. The point cloud density of the sampling points can be regarded as real physical environment information. Based on the real physical environment information, the first network extracts environmental information in the environment to be modeled.

[0082] Understandably, since the point cloud density of the sampling points is introduced, which is equivalent to a new input dimension in the first network, the dimension of the input layer of the first network needs to be adaptively adjusted.

[0083] The second network is a neural network used to extract weight information. As shown in Figure 1, multiple original radio frequency (RF) signals corresponding to different beams converge at the same sampling point, which then re-emits a scattered signal into space along the emission angle. This re-emitted scattered signal can be considered as a convergence of multiple original RF signals at the emission angle, and the intensity of the scattered signal emitted by the sampling point along the emission angle is influenced by multiple beams. At the same sampling point, different beams correspond to different beam gain matrices but share the same weight information. The weight information extracted by the second network is the weight of the scattered signal emitted along the emission angle at the sampling point, and this weight information is shared by multiple beams converging at the same sampling point.

[0084] It should be noted that the weight information extracted by the second network is shared weight information that converges at the sampling points from different path directions. This weight information is a multi-dimensional vector, with each dimension corresponding to a different path direction in the space of the environment to be modeled. The dimension of the output vector is determined by the dimension of the output layer of the second network. Therefore, when designing the output layer structure of the second network, the degree of dispersion in the horizontal and vertical directions can be determined based on the angular resolution in the horizontal and vertical directions, thereby determining the dimension of the output layer in the second network.

[0085] As shown in Figure 3, the inputs to the second network are the geographical location information of the network device, the first information output by the first network, and the emission angle. The geographical location information of the network device is typically its latitude, longitude, and altitude, which can be obtained through the engineering parameter information module. The emission angle can be determined based on the geographical location information of the sampling point and the receiving point. Based on these inputs, the second network can extract the weights of the scattered signal emitted at the sampling point along the emission angle.

[0086] Figure 4 is a flowchart illustrating a method for obtaining a channel model based on a NeRF model according to an embodiment of this application. This model acquisition method is applicable to the NeRF model shown in Figure 3, and as shown in Figure 4, it specifically includes steps S401 to S407.

[0087] S401, input the first dataset into the first network to obtain attenuation information and first information. The first dataset includes the point cloud density of the sampling points. The attenuation information is used to indicate the attenuation of the signal transmitted by the network device at the sampling points. The first information is information reflecting geographical features extracted by the first network based on the first dataset.

[0088] As shown in Figure 3, the NeRF model uses the first dataset as input data in the training dataset of the first network. This dataset can include the geographical location information of the sampling points and the point cloud density of those points. In some implementations, the first dataset can be cleaned first to remove outliers, thereby improving its quality and accuracy.

[0089] In some implementations, the point cloud density of the sampled points in the first dataset can be pre-input into a multilayer perceptron (MLP). The MLP can automatically extract the effective features from the point cloud density of the sampled points and input the geographical location information of the sampled points and the point cloud density of the sampled points processed by the MLP into the first network.

[0090] As one possible implementation, as shown in Figure 3, the point cloud density of the sampled points can also be multiplied by the intermediate layer of the first network, so that the first network can better utilize the information of the real physical environment. It can be understood that the point cloud density of the sampled points multiplied by the intermediate layer of the first network can also be pre-input into the MLP, and the point cloud density of the sampled points processed by the MLP is multiplied by the intermediate layer of the first network.

[0091] As shown in Figure 3, based on the first dataset above, the output of the first network is attenuation information and first information. To avoid redundancy, the attenuation information and first information output by the first network will not be explained in detail here.

[0092] S402, input the second dataset into the second network to obtain weight information. The second dataset includes the first information.

[0093] As shown in Figure 3, the second dataset serves as the input data in the training dataset of the second network. It may include the geographical location information of the network device, the emission angle between the sampling point and the receiving point, and the first information output by the first network in step S401.

[0094] In this step, the second dataset is input into the second network, and the second network outputs the weight information of the scattered signal emitted at the sampling point along the emission angle. Multiple different beams share this weight information at the sampling point.

[0095] It is understandable that the second network includes a multi-dimensional output layer. Therefore, the weight information output by the second network is equivalent to a multi-dimensional vector, where the weight information in each dimension corresponds to different paths in the space of the environment to be modeled. For example, when the network device transmits radio frequency signals in the environment to be modeled, the minimum resolution angle of the beam in the horizontal direction is 5 degrees (°), and the minimum resolution angle in the vertical direction is 2°. Then the weight information corresponds to 72*91=6552 dimensions.

[0096] S403 determines the intensity of the scattered signals emitted by multiple beams at the sampling point along the emission angle based on the weight information and beam gain information.

[0097] As an example, at the same sampling point, different beams correspond to different beam gain matrices but share the same weighting information. By weighted summation, the intensity of the scattered signal emitted by each beam at the sampling point along the emission angle can be determined, specifically satisfying the following relationship: S i (P x ,ω)=A i W(P x ,ω)

[0098] Among them, P xLet W(P) be the position of the sampling point, ω be the emission angle from the sampling point to the receiving point, and W(P) be the emission angle from the sampling point to the receiving point. x ,ω) is at sampling point P x The weight of the scattered signal emitted at the emission angle ω, A i S is the beam gain matrix corresponding to each of the multiple beams. i (P x ,ω) represents the intensity of the scattered signal emitted by each of the multiple beams at the sampling point along the emission angle.

[0099] It is understandable that the beam gain matrix is ​​part of the engineering parameter information. The wireless propagation modeling module shown in Figure 2 can determine the beam gain matrix corresponding to each beam among multiple beams at the sampling point by using the engineering parameter information input from the engineering parameter information module.

[0100] S404 determines the received signal strength at the receiving point based on attenuation information and the intensity of the scattered signals emitted by multiple beams at the sampling point along the emission angle.

[0101] In this step, for any receiving point in the environment space to be modeled, the simulated signal at the receiving point can be calculated based on the attenuation information output by the first network in step S401 and the intensity of the scattered signal emitted by multiple beams at the sampling point along the emission angle in step S403.

[0102] In some implementations, the received signal strength at the receiving point satisfies the following relationship:

[0103] Among them, P RX P represents the geographical location of the receiving point. TX For the geographical location of network devices, This refers to the attenuation of the signal transmitted by the network device at the receiving point. P is the azimuth vector of the signal transmitted by the network device pointing to the receiving point. r For the geographical location of the sampling points, RSRP i (P RX ) represents the reference signal received power at the receiving point for each of the multiple beams.

[0104] It is understandable that the RSRP of multiple downlink beams measured by the terminal device at the sampling point is the measured signal data, which is the target value in the NeRF model training process. The simulated signal at the receiving point needs to be compared with the target value. Therefore, the RSRP of each beam at the receiving point is used as the received signal strength at the receiving point.

[0105] It should be noted that, As a multidimensional vector, each dimension corresponds to a direction in the environment space to be modeled. Only the dimension corresponding to the direct direction from the network device to the receiving point has a value of 1, and the dimension corresponding to the other directions has a value of 0.

[0106] In the above relation, the first term on the right-hand side of the equation Used to represent receiving point P RX The strength of the received signal in the direct direction from the network device to the receiving point. According to the formula, the received signal in this direction can be determined based on the receiving point P. RX The attenuation information and free space law at the location are determined.

[0107] In the above relation, the second term on the right-hand side of the equation This is used to represent the intensity of the received signal in the non-direct direction. Figure 5 is a schematic diagram of the superposition calculation of scattered signals provided in an embodiment of this application. As an example, as shown in Figure 5, the geographical location of the receiving point is P. RX The receiving point receives the scattered signals emitted by four sampling points P1, P2, P3, and P4 along the emission angle -ω in the -ω direction. ω corresponds to the non-direct direction, therefore the receiving point P... RX The intensity of the received signal in the -ω direction can be obtained by superimposing the scattered signals emitted from the above four sampling points at the receiving point. The intensity of the received signal in the -ω direction satisfies the second term of the above equation.

[0108] The above equation is equivalent to decomposing the received signal at the receiving point according to the propagation laws. The first term on the right-hand side of the equation can be considered as rigorously calculating the direct path using physical laws, while the second term on the right-hand side can be considered as focusing on fitting the non-direct path using the NeRF model. In this step, decomposing the received signal based on deterministic physical laws can reduce the difficulty of fitting the NeRF model and improve the accuracy of the obtained channel model.

[0109] S405, calculate the training error based on the loss function, which includes the difference between the received signal strength at the receiving point and the measured signal data.

[0110] In this step, for the receiving point where there is measured signal data, that is, the sampling point of the terminal device on the predetermined acquisition route, the RSRP difference between the received signal and the measured signal data at the receiving point can be calculated based on the strength of the simulated signal obtained in step S404.

[0111] In some implementations, the loss function may include at least one of the following: a structural similarity loss function among multiple beams, a sparsity loss function, or an electromagnetic propagation law loss function, wherein the structural similarity loss function is used to constrain the intensity order of the received signals of multiple beams at the receiving point, and the electromagnetic propagation law loss function is used to constrain the difference between the outgoing signal intensity at the sampling point of the signal transmitted by the network device and the incident signal intensity at the receiving point.

[0112] It is understandable that by measuring the RSRP of multiple downlink beams at sampling points along a predetermined acquisition route, the terminal device can determine the intensity order of multiple beams received at each sampling point. During the NeRF model training process, the intensity order of multiple beams predicted by the NeRF model should be consistent with the intensity order of multiple beams determined by the actual measurement of the terminal device.

[0113] For example: the terminal device is located at geographical location P x The sampling point at point P receives four downlink beams: beam 1, beam 2, beam 3, and beam 4. The intensity order of these four downlink beams is beam 4 > beam 3 > beam 2 > beam 1. Based on the NeRF model prediction, step S404 can be used to obtain the intensity at point P. x At that location, the strength of the received signal corresponding to beam 1 is RSRP1(P x The strength of the received signal corresponding to beam 2 is RSRP2(P). x The strength of the received signal corresponding to beam 3 is RSRP3(P). x The strength of the received signal corresponding to beam 4 is RSRP4(P). x The structural similarity loss function is used to constrain the intensity ranking of the received signals corresponding to the four downlink beams to satisfy RSRP4(P). x )>RSRP3(P x )>RSRP2(P x )>RSRP1(P x ).

[0114] Theoretically, the intensity of the received signal can be calculated for any point in the environment space to be modeled. According to step S402, the intensity of the emitted signal at that point can be calculated. The intensity of the received signal and the intensity of the emitted signal satisfy the law of conservation of energy and the laws of physical propagation in terms of energy distribution. The electromagnetic propagation loss function is used to constrain the difference between the intensity of the emitted signal and the intensity of the incident signal at that point, ensuring that the difference satisfies the law of conservation of energy and the laws of physical propagation.

[0115] Sparsity generally refers to a state where most elements in a data or parameter vector are zero or close to zero. In wireless channel modeling, sparsity refers to the sparseness of multipath information. Sparsity loss functions are typically implemented through regularization terms, the two most common forms being L1 regularization and L0 regularization, which will not be elaborated upon here. Introducing a sparsity loss function into the loss function allows the NeRF model to automatically select important features and ignore unimportant features during training.

[0116] In this step, the training error can be calculated based on the above loss functions. The above sparse loss function and electromagnetic propagation law loss function can be regarded as loss functions determined based on physical propagation laws. By using prior physical laws, the dependence on measured signal data can be reduced, thereby improving the generalization ability of the NeRF model.

[0117] S406, update the neural radiation field model based on training error.

[0118] For example, in this step, the gradient of the loss function with respect to the parameters of all neural networks in step S405 can be obtained by the chain rule, and then the preset parameters of the first and second networks in the NeRF model can be adjusted in reverse according to the gradient descent method.

[0119] S407 determines whether the training error has converged or whether the training has reached the maximum number of iterations.

[0120] When the training error converges or the training reaches the maximum number of iterations, the training is complete and the channel model is obtained. Otherwise, return to step S401 and repeat the above steps.

[0121] In this embodiment, real physical environment information is used as the input to the first network in the NeRF model, and the NeRF model is trained in combination with the physical propagation law, thereby reducing the requirement for the amount of measured signal data, reducing the difficulty of fitting the NeRF model, and improving the accuracy of the obtained channel model.

[0122] It is understandable that after the wireless propagation modeling module shown in Figure 2 obtains the measured data provided by the terminal device or RAN device, the engineering parameter information provided by the engineering parameter information module, and the electronic map information provided by the electronic map module, it can be used to execute the above-mentioned method of obtaining the channel model based on the NeRF model. That is, the wireless propagation modeling module can train the NeRF model, and the channel model after training is stored in the wireless propagation modeling module.

[0123] The above embodiments train the NeRF model to obtain a channel model. Based on the channel model, attenuation and weight information at any point in the environment space to be modeled can be obtained, thereby further calculating the angular power spectrum at that point. The method of obtaining the angular power spectrum using the channel model is described in detail below.

[0124] Figure 6 is a flowchart illustrating a channel information acquisition method according to an embodiment of this application. As shown in Figure 6, the NeRF model trained in Figure 3 is used as the channel model, and the channel information acquisition method includes steps S601 to S604.

[0125] S601, acquire physical environment information, including the point cloud density of the sampling points.

[0126] In this step, the physical environment information is the input information of the channel model. As can be seen from the above embodiments, the physical environment information may include the point cloud density of the sampling points, the geographical location information of the sampling points, the geographical location information of the network devices, and the emission angle between the sampling points and the receiving points.

[0127] S602, the first dataset is input into the first network to obtain attenuation information and first information. The first dataset includes the point cloud density of the sampling points, and the first information is the information reflecting geographical features extracted by the first network based on the first dataset. The first network is a neural network used to extract environmental information.

[0128] As shown in the embodiment in Figure 4, the input of the first network in the channel model includes the geographical location information of the sampling points and the point cloud density of the sampling points. After obtaining the physical environment information in step S601, the first dataset can be constructed using the geographical location information of the sampling points and the point cloud density of the sampling points.

[0129] In some implementations, the first dataset can be cleaned to remove outlier data, thereby improving the quality and accuracy of the first dataset.

[0130] Similar to the training process of the NeRF model, in some implementations, the point cloud density of the sampling points in the first dataset can be pre-input into the MLP, and the geographical location information of the sampling points and the point cloud density of the sampling points after processing by the MLP are input into the first network.

[0131] As one possible implementation, the point cloud density of the sampled points can also be multiplied by the intermediate layer of the first network. Understandably, the point cloud density of the sampled points multiplied by the intermediate layer of the first network can also be pre-input into the MLP, and the point cloud density of the sampled points processed by the MLP can be multiplied by the intermediate layer of the first network.

[0132] In this step, the first dataset is input into the first network in the channel model, and the first network outputs the attenuation information of the original radio frequency signal transmitted by the network device at the sampling point, as well as the first information reflecting geographical features extracted from the first dataset.

[0133] S603, input the second dataset into the second network to obtain weight information. The second dataset includes the first information, and the second network is a neural network used to extract weight information.

[0134] As shown in the embodiment in Figure 4, the input of the second network in the channel model includes the geographical location information of the network device, the emission angle between the sampling point and the receiving point, and the first information. Based on the physical environment information obtained in step S601 and the first information output in step S602, the second dataset can be constructed using the geographical location information of the network device, the emission angle between the sampling point and the receiving point, and the first information.

[0135] In this step, the second dataset is input into the second network in the channel model, and the weight information output by the second network is the weight of the scattered signal emitted along the emission angle at the sampling point.

[0136] S604 obtains the angular power spectrum based on attenuation and weighting information.

[0137] In this step, the angular power spectrum at any location in the environment space to be modeled can be determined based on the attenuation and weight information obtained from the channel model. The angular power spectrum satisfies the following relationship:

[0138] Among them, P RX P represents the geographical location of the receiving point. TX For the geographical location of network devices, This refers to the attenuation of the signal transmitted by the network device at the receiving point. P is the azimuth vector of the signal transmitted by the network device pointing to the receiving point. r Where ω is the geographical location of the sampling point, and W(P) is the emission angle from the sampling point to the receiving point. r ω) represents the weight of the scattered signal emitted at the sampling point along the emission angle, X(P) RX ) represents the angular power spectrum at the receiving point.

[0139] It is understandable that the geographical location of the receiving point can correspond to any location in the environment space to be modeled. Based on the above relationship, it can be seen that the attenuation information obtained based on the channel model... The attenuation of the signal transmitted by the network device at the receiving point is the amount of signal loss. Therefore, when obtaining the attenuation information based on the first network in the channel model in the above steps, it is necessary to obtain the geographical location information and point cloud density of the receiving point to construct the first dataset and input it into the first network.

[0140] As a multidimensional vector, each dimension corresponds to a direction in the environment space to be modeled. Only the dimension corresponding to the direct direction from the network device to the receiving point has a value of 1, and the dimension corresponding to the other directions has a value of 0.

[0141] It should be noted that X(P) RX As a multidimensional vector, its dimension and weight information W(P) r , ω) are consistent. Here, different dimensions represent the path values ​​of the corresponding paths in different directions from the network device to the receiving point, including the departure angle, arrival angle, and path strength of each path in the wireless channel at the receiving point. Due to sparsity constraints, X(P RX Most of the elements in the array are 0, with only a small number of non-zero elements corresponding to multipath information.

[0142] The angular power spectrum can be output in tabular form. As an example, its specific contents are shown in Table 1 below.

[0143] Table 1

[0144] In wireless channel modeling, a custom coordinate system is typically defined in the environment to be modeled. The location of the receiver point within this custom coordinate system can be obtained by converting its latitude, longitude, and altitude information. As shown in Table 1, (0, 0, 1.5) represents the geographical location of the receiver point in the custom coordinate system. The cell where the receiver point is located is identified by its identity document (ID). As shown in Table 1, the cell ID corresponding to the cell where the receiver point is located is 3924751.

[0145] For example, Table 1 shows four possible transmission paths at the receiving point, each path can be identified by a path ID. Table 1 shows four paths with path IDs of 1, 2, 3, and 4 at the receiving point. Table 1 also shows the departure angle and path strength of different paths at the receiving point.

[0146] For example, according to Table 1, the departure angle of the transmission path with path ID 1 at the receiving point is 5 degrees (°) in the horizontal direction and -2 degrees (°) in the vertical direction. Path strength is identified by path loss, and the path loss of the transmission path with path ID 1 at the receiving point is -91.2 dB.

[0147] In this embodiment, the angular power spectrum is obtained based on physical environment information and channel model, reducing the complexity of scheme calculation and improving the efficiency of wireless channel modeling. Simultaneously, it combines data-driven algorithms with physical laws to improve the accuracy of wireless channel modeling.

[0148] It is understood that the wireless propagation modeling module shown in Figure 2 is based on the stored channel model and can be used to execute the channel information acquisition method provided in the embodiment shown in Figure 6, and output the acquired angle power spectrum to the angle power spectrum storage module. To avoid redundancy, further explanation is not provided here.

[0149] Figure 7 is a schematic block diagram of a channel information acquisition device provided in an embodiment of this application. As shown in Figure 7, the channel information acquisition device 700 includes a processing module 710 and a transceiver module 720. The channel information acquisition device 700 can be used to implement the function of the wireless propagation modeling module in the above method embodiments, and therefore can also achieve the beneficial effects of the above method embodiments.

[0150] The transceiver module 720 can implement corresponding communication functions and can also be referred to as an input / output interface or communication unit. The processing module 710 can be used to perform processing operations. It should be understood that if the device 700 is a component configured in a PC or RAN device, such as a chip, the transceiver module 720 can be an input / output interface.

[0151] Optionally, the device 700 may further include a storage module for storing instructions and / or data, and the processing module 710 may read the instructions and / or data from the storage module to enable the device to implement the method embodiments shown in FIG4 or FIG6.

[0152] A more detailed description of the processing module 710 and the transceiver module 720 can be obtained directly from the relevant descriptions in the embodiments shown in Figures 2 to 6, and will not be repeated here.

[0153] It should be noted that the transceiver module can also be called a transceiver unit, transceiver, transceiver machine, or transceiver device, etc. The processing module can also be called a processor, processing board, processing unit, or processing device, etc.

[0154] In another possible design, the aforementioned transceiver module and / or processing module can be implemented using virtual modules. For example, the processing module can be implemented using software functional modules or virtual devices, and the transceiver module can also be implemented using software functional modules or virtual devices. In another possible design, the processing module or transceiver module can also be implemented using physical devices. For example, if the device is implemented using a chip / chip circuit, the transceiver module can be an input / output circuit and / or a communication interface, performing input operations (corresponding to the aforementioned receiving operation) and output operations (corresponding to the aforementioned sending operation); the processing module is an integrated processor, microprocessor, or integrated circuit.

[0155] It should be understood that the module division in the embodiments of this application is illustrative and only represents a logical functional division. In actual implementation, there may be other division methods. Furthermore, the functional modules in the various embodiments of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0156] Figure 8 is a schematic diagram of an information acquisition device provided in another embodiment of this application. The device 800 shown in Figure 8 can be used to execute any of the methods described above performed by the information acquisition device.

[0157] As shown in Figure 8, the device 800 of this embodiment includes: a memory 801, a processor 802, a communication interface 803, and a bus 804. The memory 801, the processor 802, and the communication interface 803 are interconnected via the bus 804.

[0158] The memory 801 can be a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 801 can store programs, and when the program stored in the memory 801 is executed by the processor 802, the processor 802 performs any of the aforementioned methods.

[0159] The processor 802 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for executing relevant programs.

[0160] The processor 802 can also be an integrated circuit chip with signal processing capabilities. In implementation, the various related steps in the embodiments of this application can be completed by the integrated logic circuitry in the processor 802 or by software instructions.

[0161] The processor 802 described above can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor.

[0162] The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory 801, and processor 802 reads the information in memory 801 and, in conjunction with its hardware, completes the functions required by the units included in the device of this application.

[0163] The communication interface 803 can use, but is not limited to, transceivers to enable communication between the device 800 and other devices or apparatuses.

[0164] Bus 804 may include a pathway for transmitting information between various components of device 800 (e.g., memory 801, processor 802, communication interface 803).

[0165] This application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the steps of the methods described above.

[0166] This application also provides a computer program product, including computer instructions that, when executed by a processor, implement the various steps in the methods described above.

[0167] It should be noted that the modules or components shown in the above embodiments can be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), one or more microprocessors, or one or more field-programmable gate arrays (FPGAs). Furthermore, when a module is implemented by a processing element calling program code, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling program code, such as a controller. Additionally, these modules can be integrated together and implemented as a System-on-a-Chip (SoC).

[0168] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, software modules, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0169] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application 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 intent of this application are indicated by the following claims.

[0170] It should be understood that this application is not limited to the precise structure 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 application is limited only by the appended claims.

Claims

1. A channel information acquisition method characterized by comprising: The method comprises: obtaining physical environment information, the physical environment information comprising point cloud density of sampling points; obtaining an angle power spectrum based on the physical environment information and a channel model, the channel model being a neural radiance field model trained according to the physical environment information and measured signal data, the neural radiance field model being used to extract attenuation information and weight information, the attenuation information being used to indicate an attenuation amount of a signal transmitted by a network device at the sampling points, and the weight information being used to indicate a weight of a scattered signal transmitted along an exit angle at the sampling points.

2. The method of claim 1, wherein, The obtaining of the angle power spectrum based on the physical environment information and the channel model comprises: inputting a first data set to a first network to obtain the attenuation information and first information, the first data set comprising the point cloud density of the sampling points, the first information being information reflecting geographical features extracted by the first network according to the first data set, and the first network being a neural network used to extract environment information; inputting a second data set to a second network to obtain the weight information, the second data set comprising the first information, and the second network being a neural network used to extract the weight information; acquire the angle power spectrum based on the attenuation information and the weight information, the angle power spectrum satisfying a relationship as follows: wherein P RX is a geographic location of a receiving point, P TX is a geographic location of the network device, an amount of attenuation of a signal transmitted by the network device at the reception point, a direction vector, P, pointing to the receive point for a signal transmitted by the network device r a geographic location of the sampling point, ω is an exit angle from the sampling point pointing to a receive point, W(P r , ω) is a weight of a scattered signal transmitted at the sampling point along the exit angle, X(P RX ) is an angular power spectrum at the receive point.

3. A model acquisition method characterized by comprising: The method comprises: obtaining physical environment information, beam gain information and measured signal data, the physical environment information comprising point cloud density of sampling points; training a neural radiance field model based on the physical environment information, the beam gain information and the measured signal data to obtain a channel model, the neural radiance field model comprising a first network and a second network, the first network being a neural network used to extract environment information, and the second network being a neural network used to extract weight information, the weight information being used to indicate a weight of a scattered signal transmitted along an exit angle at the sampling points.

4. The method of claim 3, wherein, The training of the neural radiance field model based on the physical environment information, the beam gain information and the measured signal data to obtain the channel model comprises: Step one: inputting a first data set to the first network to obtain attenuation information and first information, the first data set comprising the point cloud density of the sampling points, the attenuation information being used to indicate an attenuation amount of a signal transmitted by a network device at the sampling points, and the first information being information reflecting geographical features extracted by the first network according to the first data set; Step two: inputting a second data set to the second network to obtain the weight information, the second data set comprising the first information; Step three: determining intensities of scattered signals of a plurality of beams transmitted along the exit angle at the sampling points according to the weight information and the beam gain information, the intensity of the scattered signal of each beam of the plurality of beams transmitted along the exit angle at the sampling points satisfying the following relationship: S i (P x , ω) = A i W(P x , ω) where P x is the position of a sampling point, ω is an exit angle pointing from the sampling point to a receiving point, W(P x , ω) is a weight of a scattered signal emitted along the exit angle at the sampling point, A i is a beam gain matrix corresponding to each beam of the plurality of beams, S i (P x , ω) is an intensity of a scattered signal emitted along the exit angle at the sampling point by each beam of the plurality of beams; Step four: determining a received signal intensity at the receiving point based on the attenuation information and the intensities of the scattered signals of the plurality of beams transmitted along the exit angle at the sampling points; Step five: calculating a training error according to a loss function, the loss function comprising a difference between the received signal intensity at the receiving point and the measured signal data. Step six: updating the neural radiated field model according to the training error; Step seven: if the training error does not converge or the training does not reach the maximum number of iterations, repeating the execution from the step one; Step eight: if the training error converges or the training reaches the maximum number of iterations, determining the neural radiated field model as the channel model.

5. The method of claim 4, wherein, The received signal strength at the receiving point satisfies the following relationship: wherein P RX is the geographic position of the receiving point, P TX is the geographic position of the network device, an amount of attenuation of a signal transmitted by the network device at the reception point, a bearing vector, P, for the network device transmitted signal to point towards the reception point r a geographical position, RSRP, for the sampling point i (P RX ) a reference signal received power for each of the plurality of beams at the reception point.

6. The method according to claim 4 or 5, characterized in that, The loss function further comprises at least one of a structural similarity loss function between the plurality of beams, a sparsity loss function, or an electromagnetic propagation law loss function, wherein the structural similarity loss function is used to control the intensity order of the received signals of the plurality of beams at the receiving point, and the electromagnetic propagation law loss function is used to control the difference between the exit signal intensity and the incident signal intensity of the sampling point.

7. A channel information acquisition apparatus characterized by comprising: The channel information acquisition device comprises a module for implementing the channel information acquisition method of claim 1 or 2, or a module for implementing the channel information acquisition method of any one of claims 3 to 6.

8. A channel information acquisition apparatus characterized by comprising: Comprise: A processor and a memory; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory, so that the channel information acquisition device executes the channel information acquisition method of claim 1 or 2, or the channel information acquisition method of any one of claims 3 to 6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to implement the channel information acquisition method of claim 1 or 2, or the channel information acquisition method of any one of claims 3 to 6.

10. A computer program product, characterised in that, The computer program is executed by the processor to implement the channel information acquisition method of claim 1 or 2, or the channel information acquisition method of any one of claims 3 to 6.

Images