Channel knowledge map construction method based on diffusion model in sea area communication

Abstract

The invention discloses a channel knowledge map construction method based on a diffusion model in sea area communication. The method comprises the following steps: extracting channel characteristics and spatial position information of sampling points in a target sea area environment; on the basis of the channel characteristics and the spatial position information, predicting the channel characteristics of the to-be-detected point in the coverage area by using a diffusion model; combining the channel characteristics of the sampling points and the to-be-measured points to construct a sea area channel knowledge map; obtaining channel state information of a target position by searching a sea area channel knowledge map, and guiding antenna beam design and selection to realize sea area coverage; and optimizing the diffusion model by using subsequent measured data, and updating the sea area channel knowledge map. According to the method, the prediction generation capability of the diffusion model is utilized to assist in constructing the sea area channel knowledge map, and the channel state information is obtained by searching the map and is used for guiding multi-antenna beam design and selection in a sea area wireless broadband communication system, so that the application requirements of low-overhead and rapid construction of an efficient coverage network in a wide sea area are met.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a method for constructing a channel knowledge map based on a diffusion model in maritime communications. Background Art

[0002] Broadband communications in maritime areas are crucial for maintaining national maritime security and developing the marine economy. Limited by the marine environment, wireless communications are currently the primary means of network signal coverage at sea. Due to the vast sea area and long signal propagation distances, multi-antenna beamforming technology is often required to achieve broadband communications at sea to enhance communication performance. Accurate channel state information (CSI) is an important basis for antenna beam design and selection, and has a significant impact on broadband communication systems implemented using beams. CSI can generally be obtained using real-time measurements of pilot signals. However, as the number of antennas increases rapidly, the pilot overhead used to obtain multi-antenna CSI also increases, leading to serious system overhead issues.

[0003] A channel knowledge map (CKM) is a database that stores channel characteristics in conjunction with spatial location. It can be used to quickly retrieve or predict channel state information at target locations in the ocean, thereby avoiding repeated pilot signal transmission and reducing system overhead. Existing CKMs are primarily constructed through model-driven or data-driven approaches. However, the former lacks the ability to accurately model nonlinear characteristics, making it difficult to fit the highly nonlinear channel parameters in complex ocean propagation environments, resulting in low accuracy in the generated ocean CKMs. The latter, on the other hand, relies heavily on the density and quality of measurement data. However, due to the high cost of field measurements in the ocean, limited and sparse measurement data makes it difficult to capture the spatial correlation of ocean channels, resulting in poor interpolation generalization and difficulty in generating detailed, spatially continuously varying CKMs. Furthermore, ocean channels are significantly affected by factors such as wind speed and wave height, and channel parameters are highly time-varying, further complicating the construction of CKMs. To better adapt to the channel characteristics of the ocean environment and construct an accurate CKM, a data prediction method with high-fidelity interpolation, nonlinear characterization, and enhanced adaptability to dynamic environments is needed. Summary of the Invention

[0004] Purpose of the Invention: This invention provides a method for constructing a channel knowledge map based on a diffusion model for maritime communications. This method leverages the predictive generation capabilities of the diffusion model to assist in constructing a maritime channel knowledge map. Channel state information is obtained by searching the map and used to guide the design and selection of multiple antenna beams in maritime wireless broadband communication systems, thereby meeting the application requirements of rapidly building efficient, low-cost networks covering vast ocean areas.

[0005] Technical solution: A method for constructing a channel knowledge map based on a diffusion model in maritime communications of the present invention comprises the following steps:

[0006] Step 1: Extract the channel characteristics and spatial location information of the sampling points in the target sea environment;

[0007] Step 2: Based on the channel characteristics and spatial location information, the diffusion model is used to predict the channel characteristics of the test point in the coverage area;

[0008] Step 3: Combine the channel characteristics of the sampling points and the points to be measured to construct a sea area channel knowledge map;

[0009] Step 4: Obtain channel status information at the target location by searching the sea channel knowledge map, and guide antenna beam design and selection to achieve sea coverage;

[0010] Step 5: Use subsequent measured data to optimize the diffusion model and update the sea channel knowledge map.

[0011] Furthermore, step 1 is specifically as follows: determine the sampling density λ according to the spatial variation characteristics of the sea channel environment, and collect some sampling points according to λ in the target sea area; secondly, measure and extract the channel characteristics of the sampling points, including channel gain, propagation angle, delay spread, and multipath structure; then, combine the characteristic parameter v with its position coordinate c to form a key-value pair<c,v> stored in the form of .

[0012] Furthermore, step 2 is specifically as follows: performing a standardization operation on the input channel characteristics and spatial position information; the diffusion model starts from a simple distribution consisting of a small number of sampling points, and gradually introduces system noise through reversible transformation operations to increase the complexity of the data, thereby realizing forward diffusion; then designing and optimizing the loss function to guide the model to convert simple sampled data samples into complex data distributions, realizing reversible conversion, thereby completing model training; finally, after completing the forward diffusion, the model converts the complex data samples back to a simple initial state through an inverse operation, and starting from a point in the simple distribution, gradually generates predicted data similar to the sampled data distribution, thereby realizing reverse diffusion.

[0013] Furthermore, step 2 is specifically as follows: perform a normalization operation on the input channel characteristics and spatial position information; the diffusion model starts from a simple distribution consisting of a small amount of sample data s0, gradually introduces system noise ε through reversible transformation operations, increases the complexity of the data, thereby realizing forward diffusion and obtaining complex data s t ,Right now

[0014]

[0015] in, is the cumulative multiplication of the weighted coefficients of the noise operation introduced in the first t times; then the loss function L is designed and optimized, that is,

[0016] L=||ε t -ε θ (s t ,t)|| 2 ,

[0017] Among them, θ is the model parameter, and guides the model to transform simple sampled data samples into complex data distributions, realize reversible transformation, and thus complete the model training; finally, after completing the forward diffusion, the model transforms the complex data sample s into t Convert back to the simple initial state, and starting from a point in the simple distribution, gradually generate prediction data similar to the distribution of the simple sample data s0, thereby realizing reverse diffusion, that is,

[0018]

[0019] in represents the standard normal distribution, α t is the weighted coefficient of the t-th noise introduction operation, and I is the unit matrix.

[0020] Furthermore, step 3 is as follows: First, the channel characteristics of all sampling points and the channel characteristics of the test point Stored in the same dataset In, that is

[0021]

[0022] Make it the target sea area The corresponding complete data set; secondly, establish the key-value pairs of the position coordinates of each point in the target sea area and the channel characteristics<c,v> , retrieve the corresponding channel feature v through the position coordinate c; then, arrange the data according to the position coordinate to form a regular array Array This is the channel knowledge map to be constructed; the channel knowledge map contains the channel characteristic information of all spatial locations in the target sea area.

[0023] Furthermore, step 4 is specifically as follows: first, the position coordinate c0 of the communication terminal in the target sea area is used to retrieve the constructed channel knowledge map to obtain its corresponding channel characteristics;

[0024]

[0025] The channel state information of the time-frequency domain response H0, multipath structure P0 is obtained from the retrieved channel characteristics, and the channel state information is used as the basis for optimizing the performance of the antenna beam, which can optimize and adjust the parameters of the directivity, width and gain of the beam, and on the other hand, the best code word is selected from the pre-defined codebook

[0026]

[0027] Then, the best code word g * The optimal beam is designed to cover the target sea area communication terminal, so that the beam design selection is accurately matched by blind estimation, thereby improving the system capacity and energy efficiency.

[0028] Further, step 5 is specifically: first, periodically collect a small number of N sample points in the target sea area to measure the channel characteristics And compare it with the channel characteristics v stored in the corresponding position in the current channel knowledge map, calculate the mean square error MSE

[0029]

[0030] Secondly, compare the calculated MSE(v) with the threshold value delta, if MSE(v) >= delta, use the newly collected sample point as the input to fine-tune the current diffusion model Get a new diffusion model Then, using Re-predict the channel characteristics of the to-be-measured point, and update the current channel knowledge map Get a new channel knowledge map Finally, using Optimize the antenna beam selection and design in the sea area wireless broadband communication system to adapt to the changed sea area channel environment.

[0031] The application also discloses a computer device, which comprises a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to realize the steps of the method.

[0032] The application also discloses a computer readable storage medium, which stores a computer program / instruction, and the computer program / instruction is executed by a processor to realize the steps of the method.

[0033] The application also discloses a computer program product, which comprises a computer program / instruction, and the computer program / instruction is executed by a processor to realize the steps of the method.

[0034] Advantages: compared with the prior art, the application has the following obvious advantages:

[0035] ​(1) The key point of the present application is to map the wireless channel propagation characteristics of a vast sea area into a channel knowledge map, use the powerful prediction generation capability of the diffusion model to predict the channel characteristics of non-sampled positions in the map, and then quickly obtain channel state information by looking up the channel map to assist in the design of a sea area wireless broadband communication system. This scheme not only reduces the channel estimation overhead and improves the system capacity of the target sea area communication coverage network, but also more easily matches the dynamic time-varying characteristics of the sea area channel.

[0036] (2) By using the diffusion model for sea area channel knowledge map construction, channel state information can be obtained with low overhead to assist in the design and selection of antenna beams in a sea area wireless broadband communication system, realizing fast and efficient networking coverage in the target sea area. This method not only effectively reduces the channel estimation overhead, but also improves the prediction accuracy and dynamic updating capability, so it can play an important role in various application scenarios such as sea rescue communication, ocean ranching coverage, and offshore wind power monitoring, providing strong support for the popularization and application of sea area wireless broadband communication systems.

[0037] (3) Considering that a sea area wireless broadband communication system usually deploys a MIMO antenna array, the present application uses the channel state information provided by the channel knowledge map to combine with multi-antenna technology for beam design and selection, realizing signal focusing in the spatial domain and supporting dynamic adjustment and optimization of the beam, thereby significantly improving the beam directivity and system capacity.

[0038] (4) By looking up the channel knowledge map to obtain channel state information, the overhead of repeated large-scale measurement and calculation of pilot signals in traditional channel estimation can be avoided. On the one hand, the system overhead is effectively reduced; on the other hand, the channel knowledge map can quickly provide channel characteristic information over a wide range, speeding up the deployment process of the sea area wireless broadband system and meeting the networking needs of low overhead and high performance.

[0039] (5) The powerful reasoning capability of the diffusion model helps to realize accurate prediction of channel characteristics. With accurate channel characteristics, beamforming design in a sea area wireless broadband system can be optimized, beam pointing accuracy can be improved, system communication quality and signal coverage range can be improved, and thus the overall system capacity can be improved. In addition, it can also adapt to different geographical environments and communication needs, support dynamic adjustment and optimization of beam design, and enhance the adaptability and communication capability of the sea area wireless broadband system in complex environments. The diffusion model can significantly improve data prediction accuracy and robustness through generative interpolation and probabilistic modeling, and performs outstandingly in sparse data and dynamic environment scenarios, which can solve the problems encountered in the construction of a sea area channel knowledge map.

[0040] (6) The channel knowledge map is a database that can store channel characteristics in combination with location information. It can be used to quickly retrieve and obtain the channel state information of the target location, thereby avoiding the repeated transmission of pilot signals and reducing the channel estimation overhead. Considering that the sea surface environment is flat and open with few obstructions, its channels have the characteristics of wide-area continuous and stable changes, which is suitable for predicting the overall channel environment through sparse sampling. It has similarities with the sampling reconstruction of the channel knowledge map. Therefore, the channel characteristics of the target sea area can be combined with the spatial location information to map into a channel knowledge map. By searching the channel knowledge map, the channel state information can be accurately obtained, thereby reducing the pilot overhead while quickly building a wireless broadband coverage network in the sea area. Therefore, the channel knowledge map can be constructed and applied to assist the beam design and selection in the wireless broadband communication system in the sea area, while ensuring communication performance and reducing system overhead.

[0041] (7) The diffusion model can create new data samples based on the training data, and demonstrates a strong ability in predicting and generating data. The diffusion model can gradually diffuse the desired complex data distribution through the simple data distribution of a small number of samples, thereby generating new data. Therefore, the powerful prediction and generation ability of the diffusion model can be used to predict the state of the sea channel, and the prediction results can be used to construct the sea channel knowledge map. Compared with the traditional spatial interpolation method, the diffusion model can not only improve the prediction accuracy, but also quickly fine-tune the model parameters in combination with subsequent measured results, realize the timely update of the prediction results, and flexibly construct a dynamic channel knowledge map, so that it is easier to match the dynamic time-varying characteristics of the sea channel. In summary, the diffusion model can be used to assist in the construction of the channel knowledge map, while reducing system overhead and improving the prediction accuracy and dynamic update capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The overall flow chart of the channel knowledge map construction based on the diffusion model in the sea area communication of the present invention;

[0043] Figure 2 Schematic diagram of wireless broadband communication coverage in the sea area assisted by channel knowledge map;

[0044] Figure 3 is the curve of the predicted mean square error of the sea channel knowledge map versus SNR;

[0045] Figure 4 The capacity variation curve of the sea area wireless broadband system with the assistance of channel knowledge map. DETAILED DESCRIPTION

[0046] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0047] like Figure 1As shown, a method for constructing a channel knowledge map based on a diffusion model in maritime communication of the present invention includes the following steps:

[0048] Step 1: Extract the channel characteristics and spatial location information of the sampling points in the target sea environment;

[0049] Step 2: Based on the channel characteristics and spatial location information, the diffusion model is used to predict the channel characteristics of the test point in the coverage area;

[0050] Step 3: Combine the channel characteristics of the sampling points and the points to be measured to construct a sea area channel knowledge map;

[0051] Step 4: Obtain channel status information at the target location by searching the sea channel knowledge map, and guide antenna beam design and selection to achieve sea coverage;

[0052] Step 5: Use subsequent measured data to optimize the diffusion model and update the sea channel knowledge map.

[0053] 1. Sea channel data collection and feature extraction

[0054] In order to construct a channel knowledge map mapped by the maritime wireless broadband coverage area, it is necessary to collect the marine channel data in advance and measure the channel characteristics of the sampling points in the area. The specific process includes the following key steps: First, determine the sampling density λ according to the spatial variation characteristics of the marine channel environment, and collect some sample points (i.e., sampling points) in the target sea area according to λ; secondly, measure and extract the channel characteristics of these sampling points, including but not limited to key parameters such as channel gain, propagation angle (angle of arrival, angle of departure), delay spread, and multipath structure; then, combine these characteristic parameters v with their position coordinates c in the form of key-value pairs.<c,v> It is stored in the form of , which facilitates the subsequent construction of the channel knowledge map.

[0055] It should be noted that among the above channel characteristic parameters, the ones most closely related to the beam design selection of the maritime wireless broadband system are channel gain and angle information. These characteristic parameters directly affect the optimization strategy of the maritime wireless broadband system design.

[0056] 2. Diffusion model-assisted prediction

[0057] Taking the location coordinates of all points in the target sea area and the channel characteristics of the sampling points as input, the diffusion model's powerful data prediction and generation capabilities are used to predict the channel characteristics of the remaining locations (i.e., the test points) beyond the sampling points. The specific process of predicting the channel characteristics of the test points using the diffusion model includes the following key steps: First, to enable the model to better process data, the input sampled data needs to be preprocessed and standardized to bring the data to a uniform scale, thereby impartially affecting subsequent model training. Second, the model starts with a simple distribution consisting of a small number of sampled points and gradually introduces systematic noise and increases the complexity of the data through a series of reversible transformation operations, thereby achieving forward diffusion. Then, a loss function is designed and optimized to guide the model to transform the simple sampled data samples into a complex data distribution, achieving reversible transformation and completing model training. Finally, after completing the forward diffusion, the model uses the inverse operation to transform the complex data samples back to the simple initial state. Starting from a point in the simple distribution, the model gradually generates predicted data with a distribution similar to the sampled data, thus achieving reverse diffusion.

[0058] 3. Channel Knowledge Map Construction

[0059] After obtaining the channel characteristics of all sampling points and test points in the target sea area, the sea area channel knowledge map can be constructed by combining the location information of each point. The specific process includes the following key steps: First, the channel characteristics of all sampling points are combined. and the channel characteristics of the test point Stored in the same dataset In, that is

[0060]

[0061] Make it the target sea area The corresponding complete data set; secondly, it is necessary to establish the key-value pairs of the position coordinates of each point in the target sea area and the channel characteristics<c,v> , the corresponding channel feature v can be retrieved by the position coordinate c; then, the data is arranged according to the position coordinate to form a regular array This array is the channel knowledge map to be constructed. The channel knowledge map contains the channel feature information of all spatial locations in the target sea area.

[0062] It should be noted that since the dimension of the array can be flexibly selected, each dimension can correspond to different channel characteristics (such as signal reception strength, angle, number of valid paths, etc.), so the constructed channel knowledge map can be a two-dimensional, three-dimensional, or even higher-dimensional map.

[0063] 4. Channel Knowledge Map Application

[0064] After obtaining the channel knowledge map of the target sea area, the channel state information of the target location can be retrieved from the map to guide the design and selection of antenna beams. The specific process includes the following key steps: First, the position coordinates c0 of the communication terminal in the target sea area are used to retrieve the constructed channel knowledge map to obtain its corresponding channel characteristics.

[0065]

[0066] From the retrieved channel characteristics, we can obtain key channel state information such as time-frequency domain response H0 and multipath structure P0. Using this information as the basis for optimizing antenna beam performance, we can optimize and adjust the beam directivity, width, gain and other parameters on the one hand, and on the other hand, we can select the optimal channel state information from the predefined codebook. Select the optimal beam

[0067]

[0068] Then, using the optimal beam g * Covering the target sea area communication terminals enables beam design selection to achieve precise matching through blind estimation, thereby improving system capacity and energy efficiency.

[0069] 5. Model fine-tuning and map updates

[0070] As the communication environment in the sea area changes slowly, the channel state information may also change dynamically. The sea area wireless broadband communication system can fine-tune the diffusion model by collecting new sample data, and then re-predict the channel characteristics of the test point and update the channel knowledge map, continuously optimize the beam design, and ensure that the communication system can still operate effectively in a dynamic coverage environment. The specific process includes the following key steps: First, periodically collect a small number of N sample points in the target sea area to measure their channel characteristics. And compare it with the channel feature v stored in the corresponding position in the current channel knowledge map, and calculate its mean square error (AMSE), that is,

[0071]

[0072] Secondly, the calculated MSE(v) is compared with the threshold δ. If MSE(v) ≥ δ, the newly collected samples are used as input to fine-tune the current diffusion model. Get a new diffusion model Then, use Re-predict the channel characteristics of the test point and update the current channel knowledge map Get a new channel knowledge map Finally, use Optimize the antenna beam selection and design in the marine wireless broadband communication system to adapt it to the changed marine channel environment.

[0073] The diffusion model-assisted channel knowledge map construction method for maritime communications described in this invention, by combining the channel knowledge map with location information, can not only effectively reduce channel estimation overhead, but also improve the accuracy of beam design and communication quality. It has broad application prospects in actual deployment, especially in scenarios such as maritime emergency communications and marine ranch coverage. The following case will explain this in detail.

[0074] Implementation Case 1

[0075] In maritime emergency communication scenarios, faced with emergencies and unknown ocean environments, the powerful prediction and generation capabilities of the diffusion model can be used to quickly construct a channel knowledge map. This channel knowledge map can then be used to search for channel status information and rapidly deploy an emergency network, thereby providing efficient and reliable communication services and ensuring the smooth implementation of emergency operations.

[0076] Implementation Case 2

[0077] In the marine ranch coverage scenario, facing a specified ranch area, the channel knowledge map can assist in the design of wireless broadband systems, generate beams based on user needs to enhance coverage, and improve system capacity. Figure 2 This figure illustrates a general scenario for wireless broadband communication coverage in the ocean, assisted by a channel knowledge map. The figure illustrates three common methods for ocean coverage: communication satellites, drones (or aircraft), and shore-based coverage. Taking drone coverage as an example, a drone equipped with multiple antennas uses a multi-beam approach to cover a marine ranch. The marine ranch generates a channel knowledge map through sampling measurements and interpolation predictions. The channel state information stored in the channel knowledge map provides guidance for beam design and selection.

[0078] Taking the wireless broadband network enhancement of a marine ranch as an example, the system uses sampling measurements and diffusion model predictions to obtain channel state information within the coverage area. This information is then combined with location information to create a channel knowledge map. By retrieving data from the channel knowledge map, the wireless broadband system can quickly and accurately determine channel status and flexibly generate precise beam design solutions, effectively improving signal quality and system capacity.

[0079] The powerful prediction and generation capabilities of the diffusion model effectively assist in constructing a channel knowledge map, thereby optimizing beam design in the marine ranch wireless broadband system and ensuring rapid and efficient network coverage in the marine ranch area. Compared to traditional spatial interpolation methods, the diffusion model not only improves prediction accuracy but also allows for rapid fine-tuning of model parameters based on subsequent field measurements, enabling timely updates of predictions and the flexible construction of a dynamic channel knowledge map, making it easier to adapt to the dynamic and slowly changing characteristics of the marine ranch channel environment.

[0080] Implementation Case 3

[0081] In the same wireless broadband coverage scenario in the ocean, channel knowledge maps were constructed based on two prediction schemes: traditional spatial interpolation and diffusion model prediction. The prediction mean square error of the generated channel knowledge map was calculated. The curves of the prediction mean square error achieved by the two schemes under different signal-to-noise ratios (SNR) are shown in the figure below. Figure 3 As shown in the figure, under different SNRs in the same coverage area, the channel knowledge map constructed based on the diffusion model prediction scheme has higher accuracy, and this advantage becomes more obvious as the SNR decreases.

[0082] Implementation Case 4

[0083] In the same sea area wireless broadband coverage scenario, two schemes are used to construct channel knowledge maps based on traditional spatial interpolation and diffusion model prediction, and antenna beams are designed based on the generated channel knowledge maps. The curves of system capacity achieved by the two schemes as a function of SNR are shown in the figure below. Figure 4 As shown in the figure, under different SNRs in the same coverage area, the channel knowledge map constructed based on the diffusion model can help the maritime wireless broadband system achieve higher system capacity, especially when the SNR is small.

Claims

1. A method for constructing a channel knowledge map based on a diffusion model in maritime communications, characterized in that: It includes the following steps: Step 1: Extract the channel characteristics and spatial location information of sampling points in the target sea area environment; Step 2: Based on the channel characteristics and spatial location information, use the diffusion model to predict the channel characteristics of the points to be measured within the coverage area; Step 3: Combine the channel characteristics of sampling points and points to be measured to construct a sea area channel knowledge map; Step 4: Obtain the channel state information of the target location by searching the sea area channel knowledge map, and guide the design and selection of antenna beams to achieve sea area coverage; Step 5: Use subsequent measured data to optimize the diffusion model and update the sea area channel knowledge map.

2. The method for constructing a channel knowledge map based on a diffusion model in maritime communication according to claim 1, characterized in that: Specifically, Step 1 is as follows: Determine the sampling density λ according to the spatial change characteristics of the sea area channel environment, and collect some sampling points in the target sea area according to λ; Secondly, measure and extract the channel characteristics of the sampling points, including channel gain, propagation angle, delay spread, and multipath structure; Then, combine the feature parameter v with its position coordinate c and store it in the form of a key-value pair <c, v>.

3. The method for constructing a channel knowledge map based on a diffusion model in maritime communication according to claim 1, characterized in that: Specifically, Step 2 is as follows: Perform a normalization operation on the input channel characteristics and spatial location information; The diffusion model starts from a simple distribution composed of a small number of sampling points, gradually introduces system noise through reversible transformation operations, and increases the complexity of the data, thereby achieving forward diffusion; Then design and optimize the loss function to guide the model to convert simple sampling data samples into complex data distributions and achieve reversible conversion, thereby completing model training; Finally, after completing forward diffusion, the model converts complex data samples back to a simple initial state through reverse operations, and starts from a point in the simple distribution, gradually generating prediction data similar to the sampling data distribution, thereby achieving reverse diffusion.

4. The method for constructing a channel knowledge map based on a diffusion model in maritime communication according to claim 3, characterized in that: Step 2 is as follows: perform standardization on the input channel characteristics and spatial position information; the diffusion model starts from a simple distribution consisting of a small amount of sample data s0, gradually introduces system noise ε through reversible transformation operations, increases the complexity of the data, thereby achieving forward diffusion and obtaining complex data s t ,Right now in, is the cumulative multiplication of the weighted coefficients of the noise operation introduced in the first t times; then the loss function L is designed and optimized, that is, L=||e t -e θ (s t ,t)|| 2 , Among them, θ is the model parameter, and guides the model to transform simple sampled data samples into complex data distributions, realize reversible transformation, and thus complete the model training; finally, after completing the forward diffusion, the model transforms the complex data sample s into t Convert back to the simple initial state, and starting from a point in the simple distribution, gradually generate prediction data similar to the distribution of the simple sample data s0, thereby realizing reverse diffusion, that is, in represents the standard normal distribution, α t is the weighted coefficient of the t-th noise introduction operation, and I is the unit matrix.

5. The method for constructing a channel knowledge map based on a diffusion model in maritime communication according to claim 1, characterized in that: Step 3 is as follows: First, the channel characteristics of all sampling points and the channel characteristics of the test point Stored in the same dataset In, that is Make it the target sea area The corresponding complete data set; secondly, establish the key-value pairs of the position coordinates of each point in the target sea area and the channel characteristics<c,v> , retrieve the corresponding channel feature v through the position coordinate c; Then, arrange the data according to the position coordinates to form a regular array Array This is the channel knowledge map to be constructed; the channel knowledge map contains the channel characteristic information of all spatial locations in the target sea area.

6. The method for constructing a channel knowledge map based on a diffusion model in maritime communication according to claim 1, characterized in that: Specifically, Step 4 is as follows: First, use the position coordinate c0 of the communication terminal in the target sea area to retrieve the constructed channel knowledge map and obtain its corresponding channel characteristics; The channel state information of the time-frequency domain response H0 and the multipath structure P0 is obtained from the retrieved channel characteristics. The channel state information is used as the basis for optimizing the antenna beam performance. On the one hand, the parameters of the beam directionality, width, and gain are optimized and adjusted. On the other hand, the predefined codebook is used to calculate the channel state information. Select the best codeword; Then, using the optimal codeword g * The optimal beam is designed to cover the communication terminals in the target sea area, so that the beam design selection can achieve precise matching through blind estimation.

7. The method for constructing a channel knowledge map based on a diffusion model in maritime communication according to claim 1, characterized in that: Step 5 is as follows: First, periodically collect a small number of N sample points in the target sea area to measure its channel characteristics. And compare it with the channel feature v stored at the corresponding position in the current channel knowledge map, and calculate its mean square error MSE; Secondly, the calculated MSE(v) is compared with the threshold δ. If MSE(v) ≥ δ, the newly collected samples are used as input to fine-tune the current diffusion model. Get a new diffusion model Then, use Re-predict the channel characteristics of the test point and update the current channel knowledge map Get a new channel knowledge map Finally, use Optimize the antenna beam selection and design in the marine wireless broadband communication system to adapt it to the changed marine channel environment.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the method described in claim 1.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by the processor, the steps of the method described in claim 1 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by the processor, the steps of the method described in claim 1 are implemented.

Images