A meta-universe-based wireless network optimization system and method

Abstract

The application provides a wireless network optimization system and method based on a meta universe. The method can acquire real-time base station state information, construct a base station twin model through a meta universe simulation center, quickly perform base station test adjustment based on the twin model, and when a wireless network problem is found, a management operator can perform virtual scene operation through a VR device. After the operation is completed, the adjustment instruction is issued to the base station network management through the instruction operation execution module and is executed at the base station equipped with an antenna posture instrument, thereby solving the problems that some base stations cannot be adjusted on site, adjustment is difficult, and adjustment cannot be quickly implemented, and improving the work efficiency of wireless network optimization.

Description

Technical Field

[0001] This invention relates to the field of wireless network technology, and in particular to a wireless network optimization system and method based on the metaverse. Background Technology

[0002] In routine wireless network optimization work, the inability to obtain the most accurate on-site environmental information of the wireless network in real time often makes it difficult to find a suitable and accurate network optimization method to complete the network optimization when a network problem occurs. In addition, in routine wireless network optimization work, some base stations are difficult to adjust on-site due to factors such as aesthetic antennas (exhaust pipes, aesthetic covers) and sensitive property conditions at the site. When encountering urgent wireless network optimization needs, such as traffic diversion for large-scale events and urgent VIP complaints, it is impossible to quickly implement on-site optimization due to factors such as base station distance, property procedures, and nighttime security. Summary of the Invention

[0003] The purpose of this invention is to provide a wireless network optimization system and method based on the metaverse, thereby solving the aforementioned problems existing in the prior art.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] In a first aspect, the present invention provides a wireless network optimization system based on the metaverse, comprising a perception layer, an application layer, and an interaction layer;

[0006] The perception layer includes a base station equipped with an antenna attitude sensor and an electrically adjustable antenna remote control unit, which is used to monitor the base station's operating status in real time and transmit it to the application layer, while receiving remote control commands issued by the application layer and adjusting the status of the antenna attitude sensor through the electrically adjustable antenna remote control unit.

[0007] The application layer includes a metaverse data processing center, an instruction operation execution module, and a base station network management system. The metaverse data processing center is used to receive, process, and store base station operation status data transmitted from the perception layer. The base station network management system sends remote control instructions to the base station based on the processed base station operation status data.

[0008] The interaction layer includes a metaverse scene simulation center, which constructs a metaverse scene based on the base station operation status data stored in the metaverse and presents it in the VR device. The VR device is operated to adjust and update the base station operation status data, and the adjusted instructions are transmitted to the application layer.

[0009] Preferably, the metaverse data processing center includes a 3D vector model library, a 3D vector model data processing module, a vector image processing module, and a base station and environment 3D model database. The 3D vector model data processing module includes a text decoder, a backhaul data decoder, a feature processor, and a spatiotemporal consistency processor, used to process base station backhaul data. Combined with the base station physical objects and environmental physical objects processed by the vector image processing module and the text model library, the base station backhaul data is processed into base station and environment 3D model data and stored in the base station and environment 3D model database.

[0010] Preferably, the 3D vector model data processing module processes the base station backhaul data, combining the base station physical objects and environmental physical objects processed by the vector image processing module with a text model library, to process the base station backhaul data into 3D model data of the base station and environment, specifically including the following steps:

[0011] 1) Data Decoding: The base station's backhaul data contains a set of text descriptions and base station attitude and position data. These are mapped by a text decoder and a backhaul data decoder, respectively, to obtain the text descriptions. and base station attitude and position data Meanwhile, the physical base station obtains vector features through the vector image processing module. The text descriptions, base station attitude and location data, and vector features generated in a short period of time are represented as... ;

[0012] 2) Text position consistency calculation: In the feature processor, for position-text description pairs obtained at the same time, { }, construct the antisample pair { } (i≠j), and input them together into the objective function. ,in express It is generated from the c-th group of base station attitude and position data, where D represents the vector dot product operation. This represents an adjustable constant variable used to dynamically adjust the threshold; based on the output of the objective function, data with consistent text will be clustered, while data with inconsistent text will be separated, resulting in the final processed position-text description pair { }, which is the result of the text position consistency calculation;

[0013] 3) Spatiotemporal Consistency Calculation: Based on the text position consistency results obtained in step 2), extract the position-text description pair within a given S seconds. }, after passing through the objective function The calculation, in which Represents position-vector pairs { } is located within the nth vector diagram, It is a semantically guided cross-modal fusion feature, and its expression is: ,in, , Let { represent an adjustable constant variable, then the obtained { This refers to the target's base station and environmental 3D model data.

[0014] Preferably, the metaverse scene simulation center includes a VR device, an image synthesizer, an image display interface, an image decoder, and an instruction decoder. The image synthesizer is used to render the environmental 3D model data processed by the metaverse data processing center into an image, which is then displayed on the VR device through the image display interface. By controlling the VR device to adjust the displayed screen, the image decoder converts the displayed screen into image change data, and finally, the instruction decoder converts the image change data into adjustment instructions.

[0015] Another object of the present invention is to provide a metaverse-based wireless network optimization method, implemented based on the metaverse-based wireless network optimization system described in the first aspect, comprising the following steps:

[0016] S1, the base station sends the latest data on the base station status back to the Metaverse Data Processing Center for processing;

[0017] S2, the Metaverse Data Processing Center combines the received feedback data with the text model library and processes it into base station and environment 3D model data through the 3D vector model data processing module, and stores it in the base station and environment 3D model database.

[0018] S3, the image synthesizer in the metaverse scene simulation center, renders the base station and environment 3D model data from the base station and environment 3D model database into a virtual scene and presents it on the image display interface and VR device;

[0019] S4. Management and operators adjust the displayed screen by operating the VR device. The image decoder converts the displayed screen into image change data, and finally the instruction decoder converts the image change data into adjustment instructions.

[0020] S5. The adjustment command enters the command operation execution module, which translates the adjustment command into a standard network management execution message and pushes it to the base station network management system.

[0021] S6. The base station network management system sends the network management execution message to the base station, adjusts the state of the antenna attitude device through the electrically adjustable antenna remote control unit to complete the adjustment of the real base station, and repeats step S1.

[0022] Preferably, step S2 specifically includes:

[0023] 1) Data Decoding: The base station's backhaul data contains a set of text descriptions and base station attitude and position data. These are mapped by a text decoder and a backhaul data decoder, respectively, to obtain the text descriptions. and base station attitude and position data Meanwhile, the physical base station obtains vector features through the vector image processing module. The text descriptions, base station attitude and location data, and vector features generated in a short period of time are represented as... ;

[0024] 2) Text position consistency calculation: In the feature processor, for position-text description pairs obtained at the same time, { }, construct the antisample pair { } And input them together into the target function. ,in express It is generated from the c-th group of base station attitude and position data, where D represents the vector dot product operation. This represents an adjustable constant variable used to dynamically adjust the threshold; based on the output of the objective function, data with consistent text will be clustered, while data with inconsistent text will be separated, resulting in the final processed position-text description pair { }, which is the result of the text position consistency calculation;

[0025] 3) Spatiotemporal Consistency Calculation: Based on the text position consistency results obtained in step 2), extract the position-text description pair within a given S seconds. }, after passing through the objective function The calculation, where Represents position-vector pairs { } is located within the nth vector diagram, It is a semantically guided cross-modal fusion feature, and its expression is: ,in, , , Let { represent an adjustable constant variable, then the obtained { This refers to the target's base station and environmental 3D model data.

[0026] Preferably, step S3 specifically includes:

[0027] S31. Mapping the 3D model data onto a ray: The image synthesizer emits a ray r = o + td from the origin o of the target new view through a point in the 3D model data, where t represents the distance between sampling points along the ray, and the origin is denoted as o.

[0028] S32. Query ray characteristics: For each ray, the distance to the origin o is 𝑡 𝑘The point, its position o + Both d and direction d are sent as input to the MLP model (o + d, d)→( , The model outputs the corresponding density. and RGB colors As the extracted feature for this specific point;

[0029] S33. Color Rendering: Integrate the features of each point on ray r to calculate the color C(r) of the pixel corresponding to ray r. The expression is: ,in M represents the number of sampling points along ray r. For the ray r to point o + The cumulative transmittance of d is the probability that a ray reaches a point without hitting any other point. To render an image with a resolution of K×M, the above steps will be repeated K×M times, corresponding to the number of queries of the MLP model K×M×M, so that the scene can be displayed on the image display interface;

[0030] S34. Change and Adjustment: Users can change the image display interface by operating the VR device. The image decoder will sequentially map the pixels of the changed image onto the ray, and uniformly sample all candidate points along the ray. Then, it will identify pre-existing points through a query process based on the occupancy grid. The method is to tile the occupancy grid into eight subspaces, select the non-zero cube in the subspace closest to the origin of the target view as the pre-existing point, and repeat the above steps to output the image change data.

[0031] S35, Instruction Decoding: The image variation data output from S34 is decoded by the instruction decoder. The decoding process is as follows: the image variation data is translated using the GPT-2 language model, and the output result is sent to the style adapter. For each style from j = 1 to m, a subset of β is first selected, where... Equal to the j-th style, training set The GPT-2 language model parameters are used for training, and the output after training is the required adjustment instruction.

[0032] Preferably, in step S6, the state of the antenna attitude meter is adjusted by the electrically adjustable antenna remote control unit to complete the adjustment of the real base station. Specifically, the electrically adjustable antenna remote control unit measures and rotates between two checkpoints of the minimum and maximum values ​​of the electronic downtilt angle of the supported angle, thereby realizing the reading of the downtilt angle and precise adjustment.

[0033] Preferably, the latest data on the base station status in step S1 includes the latest engineering parameter data of the base station antenna, mainly including equipment model, altitude, elevation angle, azimuth angle, and location latitude and longitude.

[0034] The beneficial effects of this invention are:

[0035] This invention provides a metaverse-based wireless network optimization system and method. The method can acquire real-time base station status information, construct a base station twin model through a metaverse simulation center, and rapidly perform base station trials and adjustments based on the twin model. When a wireless network problem is detected, management personnel can operate in a virtual scenario using VR devices. After the operation is completed, adjustment instructions are sent to the base station network management system via an instruction execution module and executed on the base station equipped with an antenna attitude sensor. This solves the problems of some base stations being unable to be accessed, difficulties in adjustment, and slow implementation, thus improving the efficiency of wireless network optimization. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the metaverse-based wireless network optimization system provided in Example 1;

[0037] Figure 2 This is a schematic diagram of the base station equipped with an antenna attitude sensor in Example 1;

[0038] Figure 3 This is a schematic diagram of the principle of the metaverse data processing center provided in Example 1;

[0039] Figure 4 This is a schematic diagram of the data processing module for the three-dimensional vector model in Example 1;

[0040] Figure 5 This is a schematic diagram of the principle of the metaverse scene simulation center in Example 1. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0042] Example 1

[0043] This embodiment provides a wireless network optimization system based on the metaverse, including a perception layer, an application layer, and an interaction layer, the structure of which is shown in the attached figure. Figure 1 As shown:

[0044] The perception layer includes a base station equipped with an antenna attitude sensor and an electrically adjustable antenna remote control unit, which is used to monitor the base station's operating status in real time and transmit it to the application layer, while receiving remote control commands issued by the application layer and adjusting the status of the antenna attitude sensor through the electrically adjustable antenna remote control unit.

[0045] The application layer includes a metaverse data processing center, an instruction operation execution module, and a base station network management system. The metaverse data processing center is used to receive, process, and store base station operation status data transmitted from the perception layer. The base station network management system sends remote control instructions to the base station based on the processed base station operation status data.

[0046] The interaction layer includes a metaverse scene simulation center, which constructs a metaverse scene based on the base station operation status data stored in the metaverse and presents it in the VR device. The VR device is operated to adjust and update the base station operation status data, and the adjusted instructions are transmitted to the application layer.

[0047] The aforementioned base station equipped with an antenna attitude sensor includes an antenna attitude sensor and a base station configured with an electrically adjustable antenna remote control unit, as shown in the attached figure. Figure 2 As shown;

[0048] The antenna attitude sensor is a device that can monitor the operation of base station antennas in real time, around the clock. The base station equipped with an electrically adjustable antenna remote control unit refers to a base station with an electrically adjustable antenna remote control unit located between the RRU and the antenna. This unit can be remotely controlled by the network management system. When a calibration command is received from the network management system, the electrically adjustable antenna remote control unit measures and rotates between two key points representing the minimum and maximum values ​​of the supported electronic downtilt angle, thereby achieving downtilt angle reading and precise adjustment.

[0049] The Metaverse Data Processing Center described in this embodiment is a data processing center system built on the Metaverse Server, and its principle is as follows: Figure 3 As shown;

[0050] The data processing center system includes a 3D vector model library, a 3D vector model data processing module, a vector image processing module, and a base station and environment 3D model database. The 3D vector model data processing module includes a text decoder, a return data decoder, a feature processor, and a spatiotemporal consistency processor. This module processes base station return data and, combined with the base station and environment physical objects processed by the vector image processing module and the text model library, transforms the data into base station and environment 3D model data, which is then stored in the base station and environment 3D model database. The processing procedure is shown in the attached figure. Figure 4 It includes the following steps:

[0051] 1) Data Decoding: The base station's backhaul data contains a set of text descriptions and base station attitude and position data. These are mapped by a text decoder and a backhaul data decoder, respectively, to obtain the text descriptions. and base station attitude and position data Meanwhile, the physical base station obtains vector features through the vector image processing module. The text descriptions, base station attitude and location data, and vector features generated in a short period of time are represented as... ;

[0052] 2) Text position consistency calculation: In the feature processor, for position-text description pairs obtained at the same time, { }, construct antisample pairs { } (i≠j), and input them together into the objective function. ,in express It is generated from the c-th group of base station attitude and position data, where D represents the vector dot product operation. This represents an adjustable constant variable used to dynamically adjust the threshold; based on the output of the objective function, data with consistent text will be clustered, while data with inconsistent text will be separated, resulting in the final processed position-text description pair { }, which is the result of the text position consistency calculation;

[0053] 3) Spatiotemporal Consistency Calculation: Based on the text position consistency calculation result obtained in 2), extract the position-text description pair within a given S seconds. }, after passing through the objective function The calculation, in which Represents position-vector pairs { } is located within the nth vector diagram, It is a semantically guided cross-modal fusion feature, and its expression is: ,in, , , Let { represent an adjustable constant variable, then the obtained { This refers to the target's base station and environmental 3D model data.

[0054] The principle of the metaverse scene simulation center described in this embodiment is as follows: Figure 5As shown, it includes a VR device, an image synthesizer, an image display interface, an image decoder, and a command decoder. The VR device is a device that displays a virtual scene, and through voice or physical buttons, feedback is provided to the image display interface to change the state of this virtual scene. The image synthesizer is a functional module used to render 3D model data into images. The image display module is a functional module that displays color and shape. The image decoder is a module that can convert the image into image change data. The command decoder is a module that can convert image change data into adjustment commands. The processing flow is as follows:

[0055] 1. Mapping 3D model data onto a ray: The image synthesizer emits a ray r = o + td from the origin o of the target new view through a point in the 3D model data, where t represents the distance between sampling points along the ray, and the origin is denoted as o;

[0056] 2. Query ray features: For each point with a distance of 𝑡𝑘 to o, its position o + Both d and direction d are sent as input to the MLP model (o + d, d)→( , The model outputs the corresponding density. and RGB colors As the extracted feature for this specific point;

[0057] 3. Color Rendering: Integrate the features of each point on ray r to calculate the color C(r) of the pixel corresponding to ray r. The expression is: ,in ), where 𝑁 represents the number of sampling points along ray r, For the ray r to point o + The cumulative transmittance of d is the probability that a ray reaches a point without hitting any other point. To render an image with a resolution of K×K, the above steps will be repeated K×K times, corresponding to the number of queries to the MLP model K×K×K, so that the scene can be displayed on the image display interface;

[0058] 4. Change and Adjustment: Users can change the image display interface by operating the VR device. The image decoder will sequentially map the pixels of the changed image onto the ray, and uniformly sample all candidate points along the ray. Then, it will identify pre-existing points through a query process based on the occupancy grid. The method is to tile the occupancy grid into eight subspaces, select the non-zero cube in the subspace closest to the origin of the target view as the pre-existing point, and repeat the above steps to output the image change data.

[0059] 5. Instruction Decoding: The image animation data output by S4 is decoded using an instruction decoder. The decoding process is as follows: the image animation data is translated using the GPT-2 language model, and the output result is sent to the style adapter. For each style from j = 1 to m, a subset of β is first selected, where... Equal to the j-th style, training set The GPT-2 language model parameters are used for training, and the output after training is the required adjustment instruction.

[0060] The instruction execution module described in this embodiment can translate adjustment instructions issued by the metaverse scenario simulation center into standard network management execution messages. It possesses functions such as instruction anomaly monitoring, automatic calibration, intelligent assessment of network management limits, and historical instruction backtracking. The base station network management system is a platform for the reasonable allocation and control of base stations to meet the requirements of service providers and the needs of network users. The base station network management system can collect the operating parameters and status information of base stations, display them to management personnel, accept their processing, and issue control instructions (changing operating status or operating parameters) to the base stations based on the processing results. It also monitors the execution results of the instructions to ensure that the base stations operate according to the requirements of the base station network management system.

[0061] Example 2

[0062] This embodiment provides a metaverse-based wireless network optimization method, implemented using the metaverse-based wireless network optimization system described in Embodiment 1, and includes the following steps:

[0063] S1, according to Figure 2 The base station equipped with an antenna attitude sensor, as described in the document, will transmit the latest data on the base station status back to the Metaverse Data Processing Center.

[0064] The latest data on the base station status includes the latest engineering parameter data of the base station antenna, mainly including equipment model, altitude, elevation angle, azimuth angle, and location latitude and longitude.

[0065] S2, the Metaverse Data Processing Center can process base station backhaul data, and combine it with the base station physical objects and environmental physical objects processed by the 3D vector image processing module, as well as the text model library, to process the base station backhaul data into base station and environmental 3D model data, and store it in the base station and environmental 3D model database;

[0066] The physical base station includes images of the base station equipment, base station equipment model, base station equipment manufacturer, and asset number; the environmental physical objects include images of the actual surrounding environment of the base station, common buildings, and environmental images; the text model library stores three-dimensional vector-related voice text description models; the three-dimensional vector model data processing module converts text descriptions and attitude data into three-dimensional base station and environmental model data through a text decoder, a return data decoder, a feature processor, and a spatiotemporal consistency processor; the base station and environmental three-dimensional model database stores the processed base station and environmental three-dimensional model data.

[0067] In this embodiment, the 3D vector model data processing module uses a text decoder, a return data decoder, a feature processor, and a spatiotemporal consistency processor to convert text descriptions and attitude data into 3D model data of the base station and environment. The specific principle is as follows:

[0068] 1) Data Decoding: The base station's backhaul data contains a set of text descriptions and base station attitude and position data. These are mapped by a text decoder and a backhaul data decoder, respectively, to obtain the text descriptions. and base station attitude and position data Meanwhile, the physical base station obtains vector features through the vector image processing module. The text descriptions, base station attitude and location data, and vector features generated in a short period of time are represented as... ;

[0069] 2) Text position consistency calculation: In the feature processor, for position-text description pairs obtained at the same time, { }, construct the antisample pair { } And input them together into the target function. ,in express It is generated from the c-th group of base station attitude and position data, where D represents the vector dot product operation. This represents an adjustable constant variable used to dynamically adjust the threshold; based on the output of the objective function, data with consistent text will be clustered, while data with inconsistent text will be separated, resulting in the final processed position-text description pair { }, which is the result of the text position consistency calculation;

[0070] 3) Spatiotemporal Consistency Calculation: Based on the text position consistency results obtained in step 2), extract the position-text description pair within a given S seconds. }, after passing through the objective function The calculation, in which Represents position-vector pairs { } is located within the nth vector diagram, It is a semantically guided cross-modal fusion feature, and its expression is: ,in, , , Let { represent an adjustable constant variable, then the obtained { This refers to the target's base station and environmental 3D model data.

[0071] S3, the image synthesizer in the metaverse scene simulation center, renders the base station and environment 3D model data from the base station and environment 3D model database into a virtual scene and presents it on the image display interface, and there is a corresponding screen display on the VR device. The specific steps are as follows:

[0072] S31. Mapping the 3D model data onto a ray: The image synthesizer emits a ray r = o + td from the origin o of the target new view through a point in the 3D model data, where t represents the distance between sampling points along the ray, and the origin is denoted as o.

[0073] S32. Query ray characteristics: For each ray, the distance to the origin o is 𝑡 𝑘 The point, its position o + Both d and direction d are sent as input to the MLP model (o + d, d)→( , The model outputs the corresponding density. and RGB colors As the extracted feature for this specific point;

[0074] S33. Color Rendering: Integrate the features of each point on ray r to calculate the color C(r) of the pixel corresponding to ray r. The expression is: ,in M represents the number of sampling points along ray r. For the ray r to point o + The cumulative transmittance of d is the probability that a ray reaches a point without hitting any other point. To render an image with a resolution of K×M, the above steps will be repeated K×M times, corresponding to the number of queries of the MLP model K×M×M, so that the scene can be displayed on the image display interface;

[0075] S34. Change and Adjustment: Users can change the image display interface by operating the VR device. The image decoder will sequentially map the pixels of the changed image onto the ray, and uniformly sample all candidate points along the ray. Then, it will identify pre-existing points through a query process based on the occupancy grid. The method is to tile the occupancy grid into eight subspaces, select the non-zero cube in the subspace closest to the origin of the target view as the pre-existing point, and repeat the above steps to output the image change data.

[0076] S35, Instruction Decoding: The image variation data output from S34 is decoded by the instruction decoder. The decoding process is as follows: the image variation data is translated using the GPT-2 language model, and the output result is sent to the style adapter. For each style from j = 1 to m, a subset of β is first selected, where... Equal to the j-th style, training set The GPT-2 language model parameters are used for training, and the output after training is the required adjustment instruction.

[0077] S4. Management and operators operate the VR device to adjust the displayed screen. The image decoder converts the displayed screen into image change data, and finally the instruction decoder converts the image change data into adjustment instructions.

[0078] The adjustment command mentioned in this embodiment refers to a Chinese description of the adjustment, such as "adjust the azimuth angle of base station A from 30 degrees to 90 degrees".

[0079] S5. The adjustment command enters the command operation execution module, which translates the adjustment command into a standard network management execution message and pushes it to the base station network management system.

[0080] Standard network management execution messages refer to a JSON string that follows the HTTP protocol and conforms to the RESTful design standard, either common to various equipment manufacturers or adapted to the actual network management situation of each manufacturer.

[0081] S6. The base station network management executes the message and sends it to the base station equipped with the antenna attitude device to complete the adjustment of the real base station. Then, it returns to step S1 and repeats the above steps S1-S6.

[0082] By adopting the above-disclosed technical solution of this invention, the following beneficial effects are obtained:

[0083] Prior to this invention, when a wireless network issue required on-site base station adjustments, it was necessary to apply for a work order, arrange vehicles, schedule appointments for personnel (including optimization staff and tower technicians), and contact property management to obtain access permission to enter the equipment room and complete the adjustments. The typical workflow took one day, on-site coordination took another day, and vehicle and personnel costs were considerable.

[0084] When applied to a problem site, this invention allows management personnel to perform virtual scene operations via VR devices when a wireless network issue is detected. After the operation, adjustment instructions are sent to the base station network management system via the instruction execution module and executed at the base station equipped with an antenna attitude sensor. The entire process takes only a few minutes and requires only the labor cost of the management personnel. Compared to before and after implementing this invention, both time and labor costs are reduced significantly. This invention can acquire real-time base station status information, solving problems such as some base stations being unable to be accessed, difficulties in adjustment, and slow implementation, thus improving the efficiency of wireless network optimization.

[0085] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A wireless network optimization system based on the metaverse, characterized in that, It includes the perception layer, application layer, and interaction layer; The perception layer includes a base station equipped with an antenna attitude sensor and an electrically adjustable antenna remote control unit, which is used to monitor the base station's operating status in real time and transmit it to the application layer, while receiving remote control commands issued by the application layer and adjusting the status of the antenna attitude sensor through the electrically adjustable antenna remote control unit. The application layer includes a metaverse data processing center, an instruction operation execution module, and a base station network management system. The metaverse data processing center is used to receive, process, and store base station operation status data transmitted from the perception layer. The base station network management system sends remote control instructions to the base station based on the processed base station operation status data. The interaction layer includes a metaverse scene simulation center, which constructs a metaverse scene based on the base station operation status data stored in the metaverse and presents it in the VR device. The VR device is operated to adjust and update the base station operation status data, and the adjusted instructions are transmitted to the application layer. The metaverse data processing center includes a 3D vector model library, a 3D vector model data processing module, a vector image processing module, and a base station and environment 3D model database. The 3D vector model data processing module includes a text decoder, a backhaul data decoder, a feature processor, and a spatiotemporal consistency processor. It is used to process base station backhaul data and, in conjunction with the base station physical objects and environmental physical objects processed by the vector image processing module and the text model library, process the base station backhaul data into base station and environment 3D model data and store it in the base station and environment 3D model database. The processing procedure includes the following steps: 1) Data Decoding: The base station's backhaul data contains a set of text descriptions and base station attitude and position data. These are mapped by a text decoder and a backhaul data decoder, respectively, to obtain the text descriptions. and base station attitude and position data Meanwhile, the physical base station obtains vector features through the vector image processing module. The text descriptions, base station attitude and location data, and vector features generated in a short period of time are represented as... ; 2) Text position consistency calculation: In the feature processor, for position-text description pairs obtained at the same time, { }, construct antisample pairs { } (i≠j), and input them together into the objective function. ,in express It is generated from the c-th group of base station attitude and position data, where D represents the vector dot product operation. This represents an adjustable constant variable used to dynamically adjust the threshold; based on the output of the objective function, data with consistent text will be clustered, while data with inconsistent text will be separated, resulting in the final processed position-text description pair { }, which is the result of the text position consistency calculation; 3) Spatiotemporal Consistency Calculation: Based on the text position consistency results obtained in step 2), extract the position-text description pair within a given S seconds. }, after passing through the objective function The calculation, where Represents position-vector pairs { } is located within the nth vector diagram, It is a semantically guided cross-modal fusion feature, and its expression is: ,in, , Let { represent an adjustable constant variable, then the obtained { This refers to the target's base station and environmental 3D model data; The metaverse scene simulation center includes a VR device, an image synthesizer, an image display interface, an image decoder, and an instruction decoder. The image synthesizer is used to render the environmental 3D model data processed by the metaverse data processing center into an image, which is then displayed on the VR device through the image display interface. By controlling the VR device to adjust the displayed screen, the image decoder converts the displayed screen into image change data, and finally, the instruction decoder converts the image change data into adjustment instructions.

2. A wireless network optimization method based on the metaverse, characterized in that, The implementation of the metaverse-based wireless network optimization system according to claim 1 includes the following steps: S1. The base station transmits the latest base station status data back to the Metaverse Data Processing Center for processing. S2. The Metaverse Data Processing Center combines the received feedback data with the text model library and processes it into base station and environment 3D model data through the 3D vector model data processing module, and stores it in the base station and environment 3D model database. S3, the image synthesizer in the metaverse scene simulation center, renders the base station and environment 3D model data from the base station and environment 3D model database into a virtual scene and presents it on the image display interface and VR device; S4. Management and operators adjust the displayed screen by operating the VR device. The image decoder converts the displayed screen into image change data, and finally the instruction decoder converts the image change data into adjustment instructions. S5. The adjustment command enters the command operation execution module, which translates the adjustment command into a standard network management execution message and pushes it to the base station network management system. S6. The base station network management system sends the network management execution message to the base station, adjusts the state of the antenna attitude device through the electrically adjustable antenna remote control unit to complete the adjustment of the real base station, and repeats step S1.

3. The wireless network optimization method based on the metaverse according to claim 2, characterized in that, Step S2 specifically includes: 1) Data Decoding: The base station's backhaul data contains a set of text descriptions and base station attitude and position data. These are mapped by a text decoder and a backhaul data decoder, respectively, to obtain the text descriptions. and base station attitude and position data Meanwhile, the physical base station obtains vector features through the vector image processing module. The text descriptions, base station attitude and location data, and vector features generated in a short period of time are represented as... ; 2) Text position consistency calculation: In the feature processor, for position-text description pairs obtained at the same time, { }, construct the antisample pair { } And input them together into the target function. ,in express It is generated from the c-th group of base station attitude and position data, where D represents the vector dot product operation. This represents an adjustable constant variable used to dynamically adjust the threshold; based on the output of the objective function, data with consistent text will be clustered, while data with inconsistent text will be separated, resulting in the final processed position-text description pair { }, which is the result of the text position consistency calculation; 3) Spatiotemporal Consistency Calculation: Based on the text position consistency results obtained in step 2), extract the position-text description pair within a given S seconds. }, after passing through the objective function The calculation, where Represents position-vector pairs { } is located within the nth vector diagram, It is a semantically guided cross-modal fusion feature, and its expression is: ,in, , , Let { represent an adjustable constant variable, then the obtained { This refers to the target's base station and environmental 3D model data.

4. The wireless network optimization method based on the metaverse according to claim 3, characterized in that, Step S3 specifically includes: S31. Mapping the 3D model data onto a ray: The image synthesizer emits a ray r = o + td from the origin o of the target new view through a point in the 3D model data, where t represents the distance between sampling points along the ray, and the origin is denoted as o. S32. Query ray characteristics: For each ray, the distance to the origin o is 𝑡 𝑘 The point, its position o + Both d and direction d are sent as input to the MLP model (o + d, d)→( , The model outputs the corresponding density. and RGB colors Features extracted for specific points; S33. Color Rendering: Integrate the features of each point on ray r to calculate the color C(r) of the pixel corresponding to ray r. The expression is: ,in M represents the number of sampling points along ray r. For the ray r to point o + The cumulative transmittance of d is the probability that a ray reaches a point without hitting any other point. In order to render an image with a resolution of d×k, the above steps will be repeated d×k times, corresponding to the number of queries of the MLP model d×k×M, so that the scene can be displayed on the image display interface. S34. Change and Adjustment: Users can change the image display interface by operating the VR device; the image decoder will map the pixels of the changed image onto the ray in sequence, and uniformly sample all candidate points along the ray. Then, it will identify the pre-existing points through a query process based on the occupancy grid. The method is to tile the occupancy grid into eight subspaces, select the non-zero cube in the subspace closest to the origin of the target view as the pre-existing point, and repeat the above steps to output the image change data. S35, Instruction Decoding: The image variation data output from S34 is decoded by the instruction decoder. The decoding process is as follows: the image variation data is translated using the GPT-2 language model, and the output result is sent to the style adapter. For each style from j = 1 to m, a subset of β is first selected, where... Equal to the j-th style, training set The GPT-2 language model parameters are used for training, and the output after training is the required adjustment instruction.

5. The wireless network optimization method based on the metaverse according to claim 4, characterized in that, In step S6, the state of the antenna attitude meter is adjusted by the electrically adjustable antenna remote control unit to complete the adjustment of the real base station. Specifically, the electrically adjustable antenna remote control unit measures and rotates between two checkpoints of the minimum and maximum values ​​of the electronic downtilt angle of the supported angle, thereby realizing the reading of the downtilt angle and precise adjustment.

6. The wireless network optimization method based on the metaverse according to claim 4, characterized in that, The latest data on the base station status mentioned in step S1 includes the latest engineering parameter data of the base station antenna, including equipment model, altitude, elevation angle, azimuth angle, and location latitude and longitude.

Images