A 5G slice online communication system for antique building digital twin point cloud detection

Abstract

The present application relates to a kind of 5G slice online communication systems for digital twin point cloud detection of antique building, specifically relates to data transmission field, by the real-time service characteristics and wireless environment state of depth fusion, construct predictive resource requirement model, realize the accurate pre-scheduling and dynamic optimization of 5G slice resource, effectively guarantee the high reliability and low latency of massive point cloud data generated in the process of antique building construction, such as three-dimensional laser scanning equipment in the transmission of online quality detection, construction deviation review and digital twin model real-time update etc.Linkage evolution mechanism can continuously calibrate prediction model according to execution feedback, make resource allocation strategy adaptive to network and service change, so as to meet the requirements of deterministic communication, significantly improve slice resource utilization efficiency, reduce adjacent service interference and terminal energy consumption, overall enhance the robustness and execution efficiency of online detection task.

Description

Technical Field

[0001] This invention relates to the field of data transmission, and more specifically, to a 5G slicing online communication system for digital twin point cloud detection of antique buildings. Background Technology

[0002] In the process of structural health monitoring and restoration of large-scale ancient-style building complexes, such as palaces and temples, it is often necessary to use laser scanners or scanning equipment mounted on drones to collect continuous, high-precision three-dimensional point cloud data. Such inspection tasks usually face complex operating environments, with the inspection equipment moving along variable paths, and need to share the same 5G network infrastructure in the same park or area with other services such as security monitoring, visitor guidance, and augmented reality displays. The high walls, deep courtyards, and complex roof structures unique to ancient-style buildings can lead to severe multipath fading and shadowing effects on wireless signals, causing the wireless channel quality to change rapidly and drastically in space. At the same time, the tidal fluctuations in visitor flow can bring unpredictable sudden background traffic impacts to the network, further exacerbating the uncertainty of the wireless transmission environment.

[0003] However, existing 5G network slicing management technologies are mostly based on static or semi-static service-level protocols for resource reservation and allocation, or can only make limited adaptive adjustments based on the overall macro load of the network. This model lacks a deep understanding of the specific service data stream characteristics of online point cloud detection, including its bursty data generation patterns and sensitivity thresholds to transmission latency and packet loss. More importantly, existing solutions fail to achieve deep integration with real-time wireless transmission environment changes, and cannot dynamically perceive and predict signal attenuation changes caused by building obstructions, as well as inter-slice resource disruptions caused by burst traffic from other services. The core issue is that static slicing configuration strategies struggle to adapt to rapid spatial changes in wireless channel quality and instantaneous fluctuations in network load. This leads to dedicated network slice resources reserved for point cloud data transmission being encroached upon or weakened in harsh environments, while remaining idle in favorable environments. Ultimately, this results in high-value point cloud data experiencing stuttering, packet loss, or excessive latency during transmission, causing incomplete or severely delayed data streams on remote online detection platforms and rendering real-time detection functions ineffective. This can lead to the omission of critical structural deformation information and a double decline in network resource utilization and detection assurance effectiveness. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a 5G slicing online communication system for digital twin point cloud detection of ancient-style buildings. The system solves the problems mentioned in the background art through a data fusion and synchronization module, a cross-domain prediction module, a joint decision optimization module, and an execution and model evolution module.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: Specifically, it includes: a data fusion and synchronization module, a cross-domain prediction module, a joint decision optimization module, and an execution and model evolution module connected in sequence; Data fusion and synchronization module: Within the slice controller, a logical model is established to support online detection of point clouds of ancient building components, verification of construction deviations, and dynamic updates of digital twin models. It continuously receives heterogeneous data, including wireless channel state parameters and real-time bit error rate from point cloud data acquisition terminals, point cloud service flow sequences within the current detection service slice from the data plane gateway, and load and interference information of adjacent slices from base stations. The heterogeneous data is timestamped and fused to output a multi-dimensional state vector. Cross-domain prediction module: Receives multi-dimensional state vectors, decomposes the point cloud service flow sequence within the detection service slice to identify its data generation period, trend, and burst components, and analyzes wireless channel state parameters to establish a channel quality spatiotemporal variation model. Then, it performs cross-domain correlation analysis on the identified burst components and the channel quality prediction information obtained based on the channel quality spatiotemporal variation model. Based on the preset delay sensitivity coefficient and bit error tolerance coefficient, it calculates and outputs the equivalent resource demand prediction curve within the future set time window through the equivalent resource demand prediction function. Joint Decision Optimization Module: Using the equivalent resource demand prediction curve as the core constraint, and combining the resource utilization rate of the current detection service slice and the interference information of adjacent slices in the multi-dimensional state vector, it performs rapid optimization calculation with the resource utilization rate within the current detection service slice, the interference to adjacent slices, and the energy consumption of the point cloud data acquisition terminal as optimization objectives. It generates and outputs a joint scheduling strategy that includes a time-frequency resource block pre-allocation scheme, a transmit power adjustment value, and an adaptive modulation and coding level suggestion. The execution and model evolution module distributes the joint scheduling strategy to the base station and the point cloud data acquisition terminal for execution. After the strategy is executed, the actual performance parameters of the point cloud data transmission and the communication resource data consumed are collected and fed back to the data fusion and synchronization module to update the multi-dimensional state vector. At the same time, the deviation between the equivalent resource demand prediction curve and the guaranteed demand calculated based on the actual results is calculated. This deviation is used to drive the channel quality spatiotemporal variation model used for service flow decomposition and channel prediction in the cross-domain prediction module to adjust the parameters. In a preferred embodiment, the specific operation of aligning and merging heterogeneous data using timestamps in the data fusion and synchronization module is as follows: A timeline based on a global clock is established, and an arrival timestamp is recorded for each group of heterogeneous data received. A delay compensation value is assigned to each group of heterogeneous data. An alignment timestamp is obtained by subtracting the delay compensation value from the arrival timestamp. Heterogeneous data are then categorized into unified time slots based on the alignment timestamps. A delay confidence weight is assigned to each type of data within each time slot. The delay confidence weight is calculated based on an exponential function with a negative exponent. The numerator is the absolute value of the difference between the historical average transmission processing delay of this type of data and the current delay compensation value, and the denominator is a preset time sensitivity constant. Heterogeneous data of various types with delay confidence weights within the same time slot are combined into initially aligned data units. Based on this, cross-domain logical association fusion is performed: for the same point cloud data acquisition terminal, its wireless channel state parameters, real-time bit error rate and point cloud service flow sequence are logically bound to form a terminal channel service triplet; at the same time, the load and interference information of adjacent slices are mapped to the wireless resource blocks used by the current detection service slice to generate slice interference mapping relationship; and a spatial-logical association factor is calculated to quantify the consistency between changes in wireless channel quality and changes in interference level.

[0006] In a preferred embodiment, the specific operation of outputting the multidimensional state vector is as follows: For each time slot, a structured multidimensional state vector is constructed, which includes a time identifier field, a terminal state set field, a service state set field, a network environment set field, and a comprehensive quality factor field. The terminal status set field includes the wireless channel status parameters, real-time bit error rate, and delay confidence weights calculated for the wireless channel status parameters and real-time bit error rate of each point cloud data acquisition terminal. The business status set field contains the aggregated business flow characteristics and their latency confidence weights for this detection business slice within the time slot; The network environment set field includes the resource utilization rate of the slice in this detection service, the interference level in the slice interference mapping relationship, and its delay confidence weight; The value of the comprehensive quality factor field is calculated as follows: the arithmetic mean of the latency confidence weights in the terminal state set field and the network environment set field is multiplied by the spatial-logical correlation factor, and then multiplied by the latency confidence weights in the service state set field. The result is the comprehensive quality factor.

[0007] In a preferred embodiment, the specific operation of analyzing the point cloud service flow sequence and wireless channel state parameters in the cross-domain prediction module is as follows: First, based on the comprehensive quality factor in the multidimensional state vector, only multidimensional state vectors with a comprehensive quality factor not lower than the quality filtering threshold are selected to form a high-quality state sequence. Next, the arrival rate of point cloud data packets in the recorded aggregated business flow features is extracted from the business state set field of each multidimensional state vector in the high-quality state sequence as the instantaneous rate of the point cloud business flow, forming a historical business flow sequence. The historical business flow sequence is then decomposed into a trend component reflecting the progress of long-term tasks, a periodic component reflecting the fixed scanning cycle of the equipment, and a residual component reflecting changes in scene details using time-series decomposition technology. The residual component is identified as a data burst component, and the peak amplitude and duration of the data burst component are then quantified to form a list of burst events described by burst intensity, burst start time, and burst duration. Meanwhile, for each multi-dimensional state vector in the high-quality state sequence, the wireless channel state parameters contained in its terminal state set field are analyzed. Combined with point cloud data to collect terminal location information, a spatiotemporal channel quality variation model describing the channel quality changes with location and time is constructed using a spatiotemporal kriging interpolation algorithm or a graph neural network-based propagation model. This spatiotemporal channel quality variation model is used to predict the channel quality indication at a specified time and predicted location in the future, and the confidence interval of the predicted channel quality indication is output simultaneously to characterize the prediction uncertainty.

[0008] In a preferred embodiment, the specific operation of performing cross-domain correlation analysis and calculating the equivalent resource demand forecast curve is as follows: Perform cross-domain correlation analysis, and perform spatiotemporal alignment and correlation analysis between the list of burst events obtained based on the burst component quantization of data and the channel quality prediction information including the predicted channel quality indicator and its confidence interval obtained by using the spatiotemporal variation model of channel quality; calculate a cross-domain risk coefficient for each future moment within each predicted burst period; Based on the results of cross-domain correlation analysis, the equivalent resource demand prediction curve is calculated using the equivalent resource demand prediction function. The calculation process is as follows: First, the predicted value of the future basic service data rate is obtained by extrapolating the trend and periodic components in the historical service flow sequence. Then, the predicted value of the future basic service data rate is multiplied by the value 1 plus the cross-domain risk coefficient adjusted by the preset first weighting factor to obtain the first intermediate value. Next, the predicted channel quality indicator is converted into the corresponding spectral efficiency estimate through a preset mapping relationship that maps the channel quality indicator to spectral efficiency. Then, the first intermediate value is divided by the spectral efficiency estimate to obtain the second intermediate value. Subsequently, the adjustment term is calculated, which is equal to the value 1 plus the delay sensitivity coefficient adjusted by the preset second weighting factor, and then minus the bit error tolerance coefficient adjusted by the preset third weighting factor. Finally, the second intermediate value is multiplied by the adjustment term, and the result is the equivalent resource demand at that future moment. This calculation process is repeated for each future moment within the set time window to generate the equivalent resource demand prediction curve.

[0009] In a preferred embodiment, the calculation process of the cross-domain risk coefficient is as follows: A1. Normalize the burst intensity obtained by quantizing the burst components of the data at this moment to obtain the normalized burst intensity; A2. Calculate the exponential function value with the natural constant as the base, where the exponent of the exponential function is negative. The numerator is the difference between the predicted channel quality indicator and a preset channel quality threshold, and the denominator is a preset scaling factor; A3. Calculate the difference between the value 1 and a ratio, where the ratio is the difference between the upper and lower bounds of the confidence interval of the predicted channel quality indicator divided by the entire range of values ​​of the channel quality indicator; Finally, multiply the normalized burst intensity obtained from A1, the exponential function value obtained from A2, and the difference obtained from A3. The product is the cross-domain risk coefficient at this future moment.

[0010] In a preferred embodiment, the specific process of performing fast optimization calculations in the joint decision optimization module is as follows: First, the received equivalent resource demand prediction curve is fused with the latest multidimensional state vector, and a dynamic multi-objective constrained optimization problem with a future-oriented time window is constructed based on the two. The equivalent resource demand prediction curve is the core constraint, and the optimization variables include the time-frequency two-dimensional resource block set pre-allocated to each point cloud data acquisition terminal, the transmit power adjustment sequence, and the adaptive modulation and coding level combination. Next, a unified performance function is constructed as the objective function, which is the first weighting coefficient multiplied by the resource utilization efficiency, minus the second weighting coefficient multiplied by the interference performance loss, and then minus the third weighting coefficient multiplied by the total energy consumption. Among them, the calculation of resource utilization efficiency introduces an evaluation correction factor for the satisfaction of the equivalent resource demand prediction curve, the calculation of interference performance loss is based on the ratio of the predicted interference to the current interference level of adjacent slices, and the total energy consumption is proportional to the sum of the power consumption of all point cloud data acquisition terminals. The three weighting coefficients are dynamic weights, which are dynamically adjusted according to the changing trend of the equivalent resource demand prediction curve and the prediction uncertainty evaluation value.

[0011] In a preferred embodiment, the specific operation of generating and outputting the joint scheduling strategy including the time-frequency resource block pre-allocation scheme, transmit power adjustment value, and adaptive modulation and coding level suggestion is as follows: An improved constrained multi-objective particle swarm optimization algorithm is employed to rapidly solve dynamic multi-objective constrained optimization problems. This algorithm comprises three mechanisms: prediction-guided particle initialization, feasibility rule-driven particle updating, and Pareto optimal solution selection based on historical performance evaluation. The prediction-guided particle initialization prioritizes allocating initial resource blocks to candidate solutions during high-demand periods based on the equivalent resource demand prediction curve. The feasibility rule-driven particle updating penalizes candidate solutions that do not meet the constraints of the equivalent resource demand prediction curve and guides their search direction. The Pareto optimal solution selection based on historical performance evaluation selects the most robust solution from the Pareto optimal solution set as the final result based on historical channel prediction accuracy. The final solution is transformed into a joint scheduling strategy that includes a time-frequency resource block pre-allocation scheme, transmit power adjustment value, and adaptive modulation and coding level recommendations.

[0012] In a preferred embodiment, the specific operation of collecting actual performance parameters of point cloud data transmission and communication resource consumption data and calculating the deviation in the execution and model evolution module is as follows: The joint scheduling strategy is issued and executed, and the actual transmission delay, actual throughput, actual block error rate, actual number of physical resource blocks used, actual average transmit power, actual channel quality indication and actual interference level are collected synchronously from the performance layer, resource layer and environment layer. Based on the collected actual performance parameters and consumed communication resource data, the deviation of the guarantee requirement satisfaction is calculated. The calculation process is as follows: the actual guarantee requirement is deduced from the actual throughput, block error rate, and number of resource blocks; the equivalent resource requirement prediction value at the same time is retrieved; the ratio of the absolute difference between the prediction value and the actual value to the actual value is calculated to obtain the basic deviation ratio; an adjustment factor is calculated, which is 1 plus the sum of the first penalty term triggered by the actual delay exceeding the delay tolerance threshold and the second penalty term triggered by the actual block error rate exceeding the block error rate tolerance threshold; the basic deviation ratio is multiplied by the adjustment factor to obtain the final deviation. This calculation process is performed for each time point within the set time window to obtain a deviation sequence composed of the deviations at each time point.

[0013] In a preferred embodiment, the specific operation of adjusting the parameters using the channel quality spatiotemporal variation model for traffic flow decomposition and channel prediction in the deviation-driven cross-domain prediction module is as follows: Source analysis is performed on the deviation sequence, linking high deviation moments with historical state characteristics. If the burst component characteristics of the data in the historical state at that moment are significant, the traffic flow decomposition model is deemed insufficient in prediction. If the difference between the actual channel quality and the predicted value exceeds the tolerance, the spatiotemporal variation model of channel quality is deemed to be invalid. A learning confidence score is calculated for each deviation data point, which is the product of the exponential function value with the natural constant base and the comprehensive quality factor of the multidimensional state vector on which the deviation is based. The exponent of the exponential function is negative, the numerator is the deviation value, and the denominator is the recent deviation moving average. Using the learning confidence score as weight, the internal parameters of the model determined to be invalid are updated incrementally through an online gradient descent algorithm. At the same time, the collected actual data is fed back to the data fusion and synchronization module in a specified format to update the multidimensional state vector.

[0014] The beneficial effects of this invention are as follows: by deeply integrating real-time service characteristics and wireless environment status, a predictive resource demand model is constructed, enabling precise pre-scheduling and dynamic optimization of 5G slice resources. This effectively ensures high reliability and low latency in the transmission of point cloud data in complex detection scenarios. Its closed-loop evolution mechanism can continuously calibrate the predictive model based on execution feedback, making the resource allocation strategy adaptable to network and service changes. Thus, while meeting deterministic communication requirements, it significantly improves the utilization efficiency of slice resources, reduces interference to adjacent services and terminal energy consumption, and enhances the robustness and execution efficiency of online detection tasks overall. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a block diagram of the system structure of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0018] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application. Example

[0019] This embodiment provides, for example Figure 1-2 The above describes a 5G slicing online communication system for digital twin point cloud detection of ancient buildings, which specifically includes: a data fusion and synchronization module, a cross-domain prediction module, a joint decision optimization module, and an execution and model evolution module connected in sequence. Data fusion and synchronization module: Within the slice controller, a logical model is established to support online detection of point clouds of ancient building components, verification of construction deviations, and dynamic updates of digital twin models. It continuously receives heterogeneous data, including wireless channel state parameters and real-time bit error rate from point cloud data acquisition terminals, point cloud service flow sequences within the current detection service slice from the data plane gateway, and load and interference information of adjacent slices from base stations. The heterogeneous data is timestamped and fused to output a time-synchronized multidimensional state vector. Cross-domain prediction module: Receives multi-dimensional state vectors, decomposes the point cloud service flow sequence within the detection service slice to identify its data generation period, trend, and burst components, and analyzes wireless channel state parameters to establish a channel quality spatiotemporal variation model. Then, it performs cross-domain correlation analysis on the identified burst components and the channel quality prediction information obtained based on the channel quality spatiotemporal variation model. Based on the preset delay sensitivity coefficient and bit error tolerance coefficient, it calculates and outputs the equivalent resource demand prediction curve within the future set time window through the equivalent resource demand prediction function. Joint Decision Optimization Module: Using the equivalent resource demand prediction curve as the core constraint, and combining the resource utilization rate of the current detection service slice and the interference information of adjacent slices in the multi-dimensional state vector, it performs rapid optimization calculation with the resource utilization rate within the current detection service slice, the interference to adjacent slices, and the energy consumption of the point cloud data acquisition terminal as optimization objectives. It generates and outputs a joint scheduling strategy that includes a time-frequency resource block pre-allocation scheme, a transmit power adjustment value, and an adaptive modulation and coding level suggestion. The execution and model evolution module distributes the joint scheduling strategy to the base station and the point cloud data acquisition terminal for execution. After the strategy is executed, the actual performance parameters of the point cloud data transmission and the communication resource data consumed are collected and fed back to the data fusion and synchronization module to update the multi-dimensional state vector. At the same time, the deviation between the equivalent resource demand prediction curve and the guaranteed demand calculated based on the actual results is calculated. This deviation is used to drive the channel quality spatiotemporal variation model used for service flow decomposition and channel prediction in the cross-domain prediction module to adjust the parameters.

[0020] In this embodiment, it is specifically necessary to explain the following operations in the data fusion and synchronization module: aligning and fusion heterogeneous data with timestamps. A timeline based on a global clock is established, recording the arrival timestamp for each received heterogeneous data set. For the wireless channel state parameters and real-time bit error rate from the point cloud data acquisition terminal, the point cloud service flow sequence within the current detection service slice from the data plane gateway, and the load and interference information (including interference levels) of adjacent slices from the base station, a delay compensation value is calculated and assigned to each heterogeneous data set based on the data type's transmission and processing path. For example, for uplink data from the point cloud data acquisition terminal, the delay compensation value mainly includes estimates of the wireless air interface transmission delay and terminal processing delay; for interference information from the base station, the delay compensation value mainly includes the base station's internal processing and reporting delay, which can be dynamically estimated based on the moving average of the delays from the most recent reported data. The aligned timestamp for each data point is obtained by subtracting the delay compensation value from the arrival timestamp. Based on the aligned timestamps, heterogeneous data from different sources are categorized into unified and continuous time slots, completing the time dimension classification. Alignment is achieved through a time slot. The length of the time slot can be set according to the system scheduling cycle, such as 1 millisecond or a transmission time interval. Each type of data within each time slot is assigned a delay confidence weight. The calculation of this delay confidence weight is based on an exponential function with a base of the natural constant. The exponent of this exponential function is negative. The numerator is the absolute value of the difference between the average transmission processing delay of this type of data in a preset historical period and the delay compensation value assigned this time. The denominator is a preset time sensitivity constant. The preset time sensitivity constant (e.g., it can be set to 100 milliseconds) is used to control the sensitivity of the confidence weight to delay deviation. The larger the deviation, the faster the weight decays, thereby reducing the impact of unreliable data. The preset historical period can be set to the past 100 data reporting periods. A delay confidence weight close to 1 indicates high data timeliness, while a weight close to 0 indicates that the data may be outdated or the delay estimate is inaccurate. Finally, various heterogeneous data belonging to the same time slot and with corresponding delay confidence weights are combined into a preliminary aligned data unit. Based on the initially aligned data units, cross-domain logical correlation fusion is performed. For data from the same point cloud data acquisition terminal, the wireless channel state parameters, real-time bit error rate, and point cloud service flow sequences within the current detection service slice generated by that point cloud data acquisition terminal are logically bound to form a terminal channel service triplet at the point cloud data acquisition terminal granularity. Simultaneously, the adjacent slice loads from the base station and interference information, including interference levels, are mapped to the radio resource blocks used by the current detection service slice based on frequency band information, generating a slice interference mapping relationship. A spatial-logical correlation factor is calculated to quantify the degree of consistency between changes in wireless channel quality and changes in external network interference levels. The calculation process is as follows: Normalize the rate of change of bit error rate and the interference level in the slice interference mapping relationship respectively. Then calculate the absolute value of the difference between the two normalization results, and subtract the absolute value from 1. The result is used as the spatial-logical correlation factor. The normalization process can adopt the minimum-maximum normalization method to map the rate of change and the interference level to the interval [0,1]. The closer the spatial-logical correlation factor is to 1, the more synchronous the channel quality deterioration and the external interference enhancement are in time and space, and the higher the credibility of this state. The closer it is to 0, the more likely the two are unrelated. The state at this moment should be accepted with caution. A correlation threshold (e.g., 0.8) can be set. The correlation factor below this threshold will trigger special marking or weight reduction processing of the corresponding time slot data. The specific operation for outputting the multidimensional state vector is as follows: A structured multidimensional state vector is constructed for each time slot. The multidimensional state vector includes the following fields: a time identifier field, recording the aligned time; a terminal state set field, containing the wireless channel state parameters, real-time bit error rate, and delay confidence weights calculated for each point cloud data acquisition terminal within the time slot; where the wireless channel state parameters may specifically include one or more of the following: reference signal received power, signal-to-interference-plus-noise ratio, and channel quality indicator; a service state set field, containing the aggregated service flow characteristics of this detection service slice within the time slot and delay confidence weights calculated for these aggregated service flow characteristics; aggregated service flow characteristics include, but are not limited to: the arrival rate of point cloud data packets within the time slot, average packet size, packet size variance, and an indicator of whether data generation is in a burst period; and a network environment set field, containing the resource utilization rate of this detection service slice, the interference level in the slice interference mapping relationship, and delay confidence weights calculated for the resource utilization rate and interference level; resource utilization rate can refer to physical... Resource block utilization or bandwidth occupancy rate; the comprehensive quality factor field, whose value is obtained through the following calculation: calculate the arithmetic mean of the latency confidence weights in the terminal state set field and the network environment set field; multiply the spatial-logical correlation factor by the arithmetic mean to obtain the first product; multiply the first product by the latency confidence weights in the service state set field, and the result is the value of the comprehensive quality factor. This calculation method ensures that the comprehensive quality factor reflects the timeliness and reliability of data on the terminal side and the network side, the consistency of cross-domain data, and the timeliness of the service data itself. The comprehensive quality factor is a scalar value between 0 and 1, and a quality qualification threshold (e.g., 0.7) can be set. Only when the comprehensive quality factor of the multidimensional state vector is higher than this threshold will it be preferentially adopted by the subsequent cross-domain prediction module for model input and analysis, thereby effectively filtering out low-quality or contradictory noise data and improving the robustness and accuracy of the entire system's decision-making. The comprehensive quality factor is used to characterize the overall reliability and consistency level of the multidimensional state vector.

[0021] In this embodiment, it is specifically necessary to explain the specific operations of the cross-domain prediction module in analyzing the point cloud service flow sequence and wireless channel state parameters as follows: First, the multidimensional state vectors are filtered based on the comprehensive quality factor in the received multidimensional state vectors. A quality filtering threshold is set, and only multidimensional state vectors with a comprehensive quality factor value greater than or equal to the quality filtering threshold are selected from the multidimensional state vector sequence. These selected multidimensional state vectors are arranged in their original chronological order to form a high-quality state sequence. The quality filtering threshold can be set in the range of 0.6 to 0.9, for example, 0.7. This step effectively eliminates low-quality historical states caused by large time delay deviations or cross-domain data contradictions, providing a clean data source for subsequent accurate modeling. Next, the arrival rate of point cloud data packets in the recorded aggregated business flow features is extracted from the business state set field of each multidimensional state vector in the high-quality state sequence as the instantaneous rate of the point cloud business flow, forming a historical business flow sequence. The historical business flow sequence is then decomposed into a trend component reflecting the progress of long-term tasks, a periodic component reflecting the fixed scanning cycle of the equipment, and a residual component reflecting changes in scene details using time-series decomposition technology. The residual component is identified as a data burst component, and the peak amplitude and duration of the data burst component are quantified to form a list of burst events described by burst intensity, burst start time, and burst duration. The time-series decomposition technology can use methods such as singular spectrum analysis or variational mode decomposition, which can effectively separate components of different time scales in the business flow. Simultaneously, for each multi-dimensional state vector in the high-quality state sequence, the wireless channel state parameters contained in its terminal state set field are analyzed. Combined with point cloud data to collect terminal location information, a spatiotemporal channel quality variation model describing the channel quality changes with location and time is constructed using a spatiotemporal channel quality interpolation algorithm or a graph neural network-based propagation model. This spatiotemporal channel quality variation model is used to predict the channel quality indication at a future specified time and predicted location, and the confidence interval of the predicted channel quality indication is output simultaneously to characterize the prediction uncertainty. The spatiotemporal channel quality interpolation algorithm can make full use of the spatial correlation and temporal continuity of channel quality to more accurately predict the signal fading area caused by building obstruction. The confidence interval, such as the 90% confidence level interval, quantifies the credibility range of the prediction result. The specific steps for performing cross-domain correlation analysis and calculating the equivalent resource demand forecast curve are as follows: Perform cross-domain correlation analysis, and perform spatiotemporal alignment and correlation analysis between the list of burst events obtained based on the burst component quantization of data and the channel quality prediction information including the predicted channel quality indicator and its confidence interval obtained by using the spatiotemporal variation model of channel quality; calculate a cross-domain risk coefficient for each future moment within each predicted burst period; Based on the results of cross-domain correlation analysis, the equivalent resource demand forecast curve is calculated using the equivalent resource demand forecast function. The calculation process is as follows: First, the predicted value of the future basic service data rate is obtained by extrapolating the trend component and periodic component in the historical service flow sequence. Then, the predicted value of the future basic service data rate is calculated as the product of the cross-domain risk coefficient adjusted by the value 1 plus a preset first weighting factor to obtain the first intermediate value. The first weighting factor is used to adjust the degree of influence of cross-domain risk on resource demand, and can be taken between 0.1 and 0.5. Next, the predicted channel quality indicator is converted into the corresponding spectral efficiency estimate through a preset mapping relationship that maps the channel quality indicator to spectral efficiency. This mapping relationship can be defined by the 3GPP standard or obtained by fitting measured data. Then, the first intermediate value is divided by the spectral efficiency estimate to obtain the second intermediate value. Subsequently, the adjustment term is calculated, which is equal to the value 1 plus the preset first weighting factor. The latency sensitivity coefficient adjusted by the second weighting factor is then subtracted from the bit error tolerance coefficient adjusted by the preset third weighting factor. The latency sensitivity coefficient and bit error tolerance coefficient are determined by the requirements of the point cloud online detection application; for example, the latency sensitivity coefficient is set to a positive value, and the bit error tolerance coefficient is set to a negative value. The second and third weighting factors are used to balance the contribution of the two to the resource requirements; for example, both can be set to 0.1. Finally, the second intermediate value is multiplied by the adjustment term, and the result is the equivalent resource requirement at that future moment. This calculation process is repeated for each future moment within the set time window to generate an equivalent resource requirement prediction curve. The equivalent resource requirement prediction curve represents the equivalent amount of wireless air interface resources required within the set time window to ensure the deterministic transmission of the point cloud data stream of this detection service slice. The final generated prediction curve is a continuous or discrete sequence that changes over time and can be directly used as the input basis for the resource scheduler to make forward-looking resource reservations. The calculation process for the cross-domain risk coefficient is as follows: A1. Normalize the burst intensity obtained from the quantization of the burst components at that moment to obtain the normalized burst intensity; A2. Calculate the value of an exponential function with the natural constant as the base, where the exponent of the exponential function is negative. The numerator is the difference between the predicted channel quality indicator and a preset channel quality threshold, and the denominator is a preset scaling factor. The preset channel quality threshold can be set based on historical channel quality statistics, such as the median or lower quartile of historical CQI data. The preset scaling factor is used to control the sensitivity of risk to channel quality degradation, and can be set to 1.0 for example. A3. Calculate the difference between the value 1 and a ratio, which is the difference between the upper and lower bounds of the confidence interval of the predicted channel quality indicator divided by the entire range of values ​​of the channel quality indicator. Finally, multiply the normalized burst intensity obtained from A1, the exponential function value obtained from A2, and the difference obtained from A3. The product is the cross-domain risk coefficient for that future moment. This formula is designed to ensure that the cross-domain risk coefficient reaches its maximum value when the service burst intensity is high, the predicted channel quality is significantly lower than the threshold, and the prediction result has high certainty, thus accurately capturing the high-risk transmission moment when "service peak meets channel trough".

[0022] In this embodiment, it is necessary to specifically explain the process of performing fast optimization calculation in the joint decision optimization module as follows: First, the received equivalent resource demand prediction curve is fused with the latest multi-dimensional state vector. Based on these two, a dynamic multi-objective constrained optimization problem with a future-oriented time window is constructed. The equivalent resource demand prediction curve serves as the core constraint condition that must be satisfied in the dynamic multi-objective constrained optimization problem. The optimization variables specifically include three categories: The first category of optimization variables consists of each point cloud data acquisition terminal, each scheduling cycle within the future optimization window, and a pre-allocated set of time-frequency two-dimensional resource blocks. This set of time-frequency two-dimensional resource blocks represents the resource occupancy in the time and frequency dimensions. The set of time-frequency two-dimensional resource blocks can be represented by a three-dimensional binary matrix, with its dimensions corresponding to the terminal identifier, scheduling cycle index, and physical resource block index, respectively. The second category of optimization variables consists of each point cloud data acquisition terminal... The terminal plans a transmit power adjustment sequence covering each scheduling period within the future optimization window. Each value in the transmit power adjustment sequence represents the adjustment amount of the transmit power of the point cloud data acquisition terminal relative to its current power level within the corresponding scheduling period. The transmit power adjustment amount is usually measured in decibels and milliwatts, and its value range is limited by the terminal hardware capabilities and communication protocols. The third type of optimization variable is an adaptive modulation and coding level combination recommended for each point cloud data acquisition terminal, covering each scheduling period within the future optimization window. Each level corresponds to a specified data transmission efficiency and reliability scheme. The adaptive modulation and coding level combination is defined according to the 3GPP standard, such as different levels from QPSK1 / 2 to 256QAM3 / 4. Different levels correspond to different spectral efficiency and block error rate curves.Next, a unified performance function is constructed as the objective function for the entire dynamic multi-objective constrained optimization problem. The unified performance function is composed of three weighted sub-objectives: resource utilization efficiency within the current detection service slice, interference performance loss caused to adjacent slices, and total energy consumption of all point cloud data acquisition terminals. The calculation process for the unified performance function value is as follows: multiply the first weight coefficient by the resource utilization efficiency, subtract the second weight coefficient multiplied by the interference performance loss, and then subtract the third weight coefficient multiplied by the total energy consumption. The calculation process for resource utilization efficiency is as follows: first, calculate the capacity within the future optimization window, based on the number of resource blocks determined by the first type of optimization variable, the transmit power determined by the second type of optimization variable, and the modulation and coding scheme determined by the third type of optimization variable. The ratio of the total number of point cloud data bits after removing invalid data affected by the bit error rate to the theoretical maximum number of data bits that the allocated time-frequency resource block can carry is multiplied by a correction factor. This correction factor is determined by how well the actual allocated resource quantity fits the equivalent resource demand prediction curve in the time dimension. The closer the actual allocated resource quantity is to the predicted demand value, the higher the correction factor value, and vice versa. The calculation of the correction factor can be specifically represented as a Gaussian kernel function, with its center located on the equivalent resource demand prediction curve. The closer the actual allocated quantity is to the center, the closer the factor value is to 1; the farther it deviates, the smaller the value. For example, when the allocated quantity exceeds or falls below the predicted value by 20%, the correction factor may decrease. Reduced to below 0.5 to penalize oversupply or severe shortage of resources; the calculation process for interference effectiveness loss is as follows: for each frequency point used by this detection service slice within the future optimization window, calculate the predicted interference caused by this slice to adjacent slices at that frequency point, divide by the sum of the background noise power and the interference level extracted from the latest multidimensional state vector in the slice interference mapping relationship, and then sum the calculation results of all frequency points; the predicted interference is the expected interference power of this slice to adjacent slices calculated based on the first and second type of optimization variables; the calculation of predicted interference can be estimated based on the path loss model and the transmit power adjustment sequence, focusing on the out-of-band leakage power generated in overlapping or adjacent frequency bands; the calculation of total energy consumption is related to a The total energy is directly proportional to the total energy, which equals the sum of the transmit power adjustments of all point cloud data acquisition terminals in each scheduling cycle within the optimization window, plus the sum of the basic power consumption required by all point cloud data acquisition terminals to maintain communication within the optimization window. The basic power consumption is the inherent power consumption of the terminal when maintaining a wireless connection but not transmitting data, and can be obtained from the equipment specifications. The first, second, and third weight coefficients are dynamic weights, and the dynamic adjustment logic is as follows: when the slope of the average rate of change of the equivalent resource demand prediction curve within the optimization window is positive and the value increases, the value of the first weight coefficient is increased. For example, when the average slope exceeds a preset slope threshold (such as 0.1), the first weight coefficient is increased from the baseline value of 0.5 to 0.7. To strengthen the emphasis on resource protection; when the uncertainty assessment value of the prediction results from the cross-domain prediction module increases, the value of the second weighting coefficient is increased; the uncertainty assessment value of the prediction results can be quantified by the width of the prediction channel quality confidence interval output by the cross-domain prediction module. When the average confidence interval width exceeds the preset width threshold, the second weighting coefficient is increased, for example, from 0.3 to 0.5, to enhance the robustness of the scheduling scheme and actively avoid the high interference risk that may be caused by inaccurate prediction; the third weighting coefficient is adjusted in the opposite direction after the other two weighting coefficients are adjusted to maintain the balance of the optimization objectives; usually, the sum of the three weighting coefficients is kept at 1, for example, initially set to [0.5, 0.3, 0.2], and after adjustment may become [0.7, 0.25, 0.05], while prioritizing resource protection and reducing interference, appropriately reducing the absolute requirements for energy consumption; The specific steps for generating and outputting a joint scheduling strategy that includes a time-frequency resource block pre-allocation scheme, transmit power adjustment values, and adaptive modulation and coding level recommendations are as follows: An improved particle swarm optimization algorithm suitable for constrained multi-objective optimization is adopted to quickly solve dynamic multi-objective constrained optimization problems. The improved particle swarm optimization algorithm includes three mechanisms executed sequentially: prediction-guided particle initialization, feasibility rule-driven particle updating, and Pareto optimal solution selection based on historical performance evaluation. Prediction-guided particle initialization refers to the allocation of initial resource blocks to candidate solutions in the time-frequency two-dimensional resource space based on the future resource demand indicated by the equivalent resource demand prediction curve, prioritizing the allocation of initial resource blocks to candidate solutions near future time periods with higher predicted demand. Specifically, at the initial particle position (i.e., candidate...) In the scheduling scheme, for each particle, within the time period corresponding to the peak of the equivalent resource demand prediction curve, a certain number of resource blocks are pre-allocated with a high probability (e.g., 80%). This makes the initial population more likely to contain feasible solutions that satisfy the constraints of high-demand periods, significantly accelerating the convergence speed. Feasibility rule-driven particle update refers to calculating the estimated available resources that each candidate scheduling scheme can provide at future times during the particle swarm optimization algorithm iteration process, and comparing it with the value required by the equivalent resource demand prediction curve at that time. If at any time the estimated available resources are less than the value of the equivalent resource demand prediction curve, then the fitness evaluation of that candidate scheme is... A penalty term is added to the value, and the value of the penalty term is proportional to the size of the resource gap. For example, the penalty term = penalty coefficient × the sum of the squares of the resource gaps at all times, where the penalty coefficient is a constant much greater than 1 (such as 1000). This causes the fitness of any solution that violates the core constraints to deteriorate drastically, thus leading to its rapid elimination in the evolution. At the same time, during the process of updating the position of the candidate solution in the particle swarm optimization algorithm, its search direction is guided towards the solution space region that can increase the resource supply at that time or better conform to the shape of the equivalent resource demand prediction curve. In the particle velocity and position update formula, an additional guiding force is added to the particle that violates the constraints, and the direction of this force is directed towards its own historical best position. Or, in the historical optimal position of the swarm optimization algorithm, those dimensions that can satisfy the constraints of the equivalent resource demand prediction curve can be intelligently guided to fly towards the feasible region; Pareto optimal solution selection based on historical performance evaluation refers to selecting the most robust solution in a network environment with similar historical channel prediction accuracy from the Pareto optimal solution set after the Particle Swarm Optimization Algorithm finishes running and generates a Pareto optimal solution set, based on the recent channel prediction accuracy historical data reflected in the latest multidimensional state vector, as the final optimization result; the specific selection method can be: to establish a historical case library, recording the actual resource utilization and interference level after the execution of different scheduling schemes in different channel prediction accuracy intervals in the past;Starting from each solution at the current Pareto frontier, matching scenarios with similar prediction accuracy in the historical case library, calculating the variance or worst case of various performance indicators (such as resource utilization and interference level) of the solution if it were executed in those historical scenarios, and selecting the solution with the smallest variance or the best worst case as the final output, this ensures that the selected solution can still maintain relatively stable and reliable performance when facing prediction errors similar to those in the past. The final selected solution, which includes the optimal time-frequency resource block allocation, the optimal transmit power adjustment sequence, and the optimal adaptive modulation and coding level combination, is transformed into a structured joint scheduling strategy that can be directly issued to the base station and the point cloud data acquisition terminal for execution. The joint scheduling strategy specifically includes the following three types of instructions: a time-frequency resource block pre-allocation scheme, which explicitly specifies in a list which physical resource blocks, determined by both time-domain and frequency-domain indices, will be allocated to which point cloud data acquisition terminals in which future scheduling cycles; the time-domain index is typically the system frame number and subframe number, and the frequency-domain index is the resource block number; a transmit power adjustment value, which provides each point cloud data acquisition terminal with a specific increase or decrease in transmit power relative to the current power level in future scheduling cycles; this instruction can be encapsulated as a standard radio resource control signaling message; and an adaptive modulation and coding level recommendation, which recommends a specific combination of modulation scheme and channel coding rate level for each point cloud data acquisition terminal in future scheduling cycles; this adaptive modulation and coding level recommendation can be sent to the point cloud data acquisition terminal as part of the downlink control information, and the point cloud data acquisition terminal can decide whether to adopt it based on its own circumstances.

[0023] In this embodiment, it is specifically necessary to explain the following operations in the execution and model evolution module: collecting the actual performance parameters of point cloud data transmission and the communication resource data consumed, and calculating the deviation: First, the received joint scheduling strategy is broken down and distributed separately to the base station and point cloud data acquisition terminal to monitor its execution within a future set time window. During execution, real data is collected synchronously from three levels: The first level is the performance layer, which collects actual performance parameters of point cloud data transmission from the data plane gateway, including the actual transmission delay experienced by point cloud data packets from sending to receiving, the actual throughput of successful transmission per unit time, and the actual block error rate during transmission; the actual transmission delay can be calculated based on the data packet timestamp, the actual throughput is the ratio of the successfully received payload data volume to the time, and the actual block error rate is determined by cyclic redundancy check or hybrid automatic retransmission request feedback; the second level is the resource layer, which collects data from the base station scheduler... The communication resource data consumed in executing the joint scheduling strategy includes the number of physical resource blocks actually used and the actual average transmit power of the point cloud data acquisition terminal when transmitting data. The number of physical resource blocks is obtained from the media access control layer scheduling log of the base station, and the actual average transmit power is obtained from the terminal's power margin report or base station measurement. The third layer is the environment layer, which extracts the actual channel quality indication of the wireless network during the strategy execution period and the actual network interference level from real-time network data, as a post-event verification of the channel and environment predictions previously made by the cross-domain prediction module. The actual channel quality indication can be obtained by demodulation reference signal measurement, and the actual interference level can be estimated by subtracting the serving cell signal strength from the received signal strength indication. Next, based on the collected actual performance parameters and the communication resource data consumed, the deviation of the guarantee requirement satisfaction is calculated. The deviation calculation process is as follows: E1, based on the actual throughput achieved, the actual block error rate, and the actual number of physical resource blocks used, the equivalent resource requirement per unit time required to support this actual transmission effect is derived in reverse. This derived resource quantity is defined as the actual guarantee requirement. This inverse derivation can be based on an approximate implementation of Shannon's formula. The actual throughput is divided by (1 - actual block error rate) to obtain the equivalent rate under error-free conditions, and then combined with the actual number of physical resource blocks used. E1. Calculate the unit resource block spectral efficiency required to support the rate, and then deduce the equivalent resource requirement; E2. Retrieve the predicted value corresponding to the equivalent resource requirement prediction curve output by the cross-domain prediction module at the same time from the historical records of the storage equivalent resource requirement prediction curve. This predicted value represents the previously predicted guarantee requirement; E3. Calculate the absolute difference between the predicted guarantee requirement and the actual guarantee requirement, and divide the absolute difference by the actual guarantee requirement to obtain a basic deviation ratio; E4. Calculate the adjustment factor, which is a value of 1 plus the sum of the latency out-of-bounds penalty and the block error rate over-limit penalty; The latency exceeding the limit penalty takes effect under the first condition. Its value is equal to the first penalty weight coefficient multiplied by an indicator function. This indicator function takes a value of 1 when the actual transmission latency exceeds a preset latency tolerance threshold, and 0 otherwise. The latency tolerance threshold is set according to the requirements of the point cloud online detection application, for example, it can be 10 milliseconds. The first penalty weight coefficient can take a value between 0.2 and 1.0, for example, 0.5. The block error rate exceeding the limit penalty takes effect under the second condition. Its value is equal to the second penalty weight coefficient multiplied by an indicator function. This indicator function takes a value of 1 when the actual block error rate exceeds a preset block error rate tolerance threshold, and 0 otherwise. 0; The block error rate tolerance threshold is set according to the forward error correction capability, for example, it can be 1e-3; The second penalty weight coefficient can be the same as or different from the first penalty weight coefficient; E5, Multiply the basic deviation ratio by the adjustment factor, and the result is the guarantee requirement satisfaction deviation at that moment; This deviation calculation method not only reflects the numerical error of resource prediction, but also amplifies the consequences of the degradation of business performance (latency, bit error) caused by prediction error through the penalty term, so that the system can more sensitively capture and learn key errors; Perform steps E1 to E5 for each moment within the set time window to obtain the deviation sequence composed of the deviations at each moment; The specific operation for parameter adjustment using the channel quality spatiotemporal variation model for traffic flow decomposition and channel prediction in the bias-driven cross-domain prediction module is as follows: First, a source tracing analysis is performed on the calculated deviation sequence, linking the historical multidimensional state vector features corresponding to each high deviation moment with high deviation values ​​to the cause of the deviation. Specifically, the source tracing analysis involves checking whether the data burst component features recorded in the service state set field of the historical multidimensional state vector used to generate the predicted value at each high deviation moment are significant. If significant, the deviation is determined to be mainly due to the insufficient predictive ability of the model used for service flow decomposition for burst patterns. The threshold for "significance" can be set based on the statistical distribution of historical burst intensity, for example, exceeding the historical average burst intensity by one standard deviation. Simultaneously, the difference between the actual channel quality indicator collected at that moment and the predicted channel quality indicator used during prediction is checked. If the difference exceeds the preset channel prediction error tolerance, the deviation is determined to be mainly due to the failure of the channel quality spatiotemporal variation model. The channel prediction error tolerance can be set based on the quantization level of the channel quality indicator, for example, a difference of two or more CQI levels. Then, based on the source analysis results, a learning confidence score is calculated for each deviation data point in the deviation sequence. The learning confidence score is calculated as follows: First, the exponential function value with the natural constant as the base is calculated, with the exponent of the exponential function being negative. The numerator is the value of the deviation data point, and the denominator is the moving average of all recent deviation data points. Then, the calculated exponential function value is multiplied by the comprehensive quality factor of the multidimensional state vector on which the historical prediction corresponding to the deviation is based. The product is the learning confidence score of the deviation data point. This design ensures that: when the deviation is too large (which may be an outlier), the confidence score decreases; and the higher the quality (comprehensive quality factor) of the input data used for the original prediction, the greater the reference value of the deviation for model correction, and the higher the confidence score, thus achieving robust incremental learning. The higher the learning confidence score, the stronger the guiding significance of the deviation data for model correction, and the higher its weight in subsequent parameter updates. Finally, using the calculated learning confidence as weights, a loss function is constructed using the deviation sequence data. The internal parameters of models deemed ineffective are then incrementally updated using an online gradient descent algorithm. The learning rate for online gradient descent should be set relatively small (e.g., 0.001 or below) to ensure smooth evolution of model parameters and avoid drastic oscillations. If the deviation is traced back to the model used for traffic flow decomposition, the internal parameters in that model used to identify trends, periods, and burst components are updated. For example, the construction dimension of the trajectory matrix in singular spectral analysis or the center frequency constraint parameters of the intrinsic mode functions in variational mode decomposition are updated. If the deviation is traced back to the spatiotemporal variation model of channel quality, the interpolation or neural network in that model is updated. Network weight parameters; for example, updating the mutation function parameters in the spatiotemporal kriging algorithm, or the edge weights in the graph neural network; simultaneously, the collected actual performance parameters, consumed communication resource data, actual channel quality indicators and network interference levels are aligned and fused according to the heterogeneous data timestamp alignment and fusion format and rules defined in the data fusion and synchronization module to generate new state records, which are then fed back to the data fusion and synchronization module to generate new multidimensional state vectors, thereby ensuring the continuous accuracy of the system's perceived state representation, providing more accurate recent history for the next round of prediction, and completing the complete data closed loop from decision execution to state perception; this feedback loop is the key mechanism for the system to achieve continuous self-optimization and adapt to environmental and business changes.

[0024] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0025] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0026] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0027] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0028] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0029] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0030] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A 5G slicing online communication system for digital twin point cloud detection of antique-style buildings, characterized in that, Specifically, it includes: The data fusion and synchronization module, cross-domain prediction module, joint decision optimization module, and execution and model evolution module are connected sequentially. Data fusion and synchronization module: Within the slice controller, a logical model is established to support online detection of point clouds of ancient building components, verification of construction deviations, and dynamic updates of digital twin models. It continuously receives heterogeneous data, including wireless channel state parameters and real-time bit error rate from point cloud data acquisition terminals, point cloud service flow sequences within the current detection service slice from the data plane gateway, and load and interference information of adjacent slices from base stations. The heterogeneous data is timestamped and fused to output a multi-dimensional state vector. Cross-domain prediction module: Receives multi-dimensional state vectors, decomposes the point cloud service flow sequence within the detection service slice to identify its data generation period, trend, and burst components, and analyzes wireless channel state parameters to establish a channel quality spatiotemporal variation model. Then, it performs cross-domain correlation analysis on the identified burst components and the channel quality prediction information obtained based on the channel quality spatiotemporal variation model. Based on the preset delay sensitivity coefficient and bit error tolerance coefficient, it calculates and outputs the equivalent resource demand prediction curve within the future set time window through the equivalent resource demand prediction function. Joint Decision Optimization Module: Using the equivalent resource demand prediction curve as the core constraint, and combining the resource utilization rate of the current detection service slice and the interference information of adjacent slices in the multi-dimensional state vector, it performs rapid optimization calculation with the resource utilization rate within the current detection service slice, the interference to adjacent slices, and the energy consumption of the point cloud data acquisition terminal as optimization objectives. It generates and outputs a joint scheduling strategy that includes a time-frequency resource block pre-allocation scheme, a transmit power adjustment value, and an adaptive modulation and coding level suggestion. The execution and model evolution module distributes the joint scheduling strategy to the base station and the point cloud data acquisition terminal for execution. After the strategy is executed, the actual performance parameters of the point cloud data transmission and the communication resource data consumed are collected and fed back to the data fusion and synchronization module to update the multi-dimensional state vector. At the same time, the deviation between the equivalent resource demand prediction curve and the guaranteed demand calculated based on the actual results is calculated. This deviation is used to drive the channel quality spatiotemporal variation model used for service flow decomposition and channel prediction in the cross-domain prediction module to adjust the parameters.

2. The 5G slicing online communication system for digital twin point cloud detection of antique buildings according to claim 1, characterized in that: In the data fusion and synchronization module, the specific operations for aligning and fusion heterogeneous data with timestamps are as follows: Establish a timeline based on a global clock, record the arrival timestamp for each set of heterogeneous data received, and assign a delay compensation value to each set of heterogeneous data; The alignment timestamp is obtained by subtracting the delay compensation value from the arrival timestamp, and heterogeneous data is classified into a unified time slot based on the alignment timestamp; Each type of data within each time slot is assigned a delay confidence weight. The delay confidence weight is calculated based on an exponential function with the natural constant as the base. The exponent is negative. The numerator is the absolute value of the difference between the historical average transmission and processing delay of this type of data and the current delay compensation value. The denominator is a preset time sensitivity constant. The various heterogeneous data with delay confidence weights within the same time slot are combined into a preliminary aligned data unit. Based on this, cross-domain logical association fusion is performed: for the same point cloud data acquisition terminal, its wireless channel status parameters, real-time bit error rate and point cloud service flow sequence are logically bound to form a terminal channel service triplet; at the same time, the load and interference information of adjacent slices are mapped to the wireless resource blocks used by the current detection service slice to generate slice interference mapping relationship. A spatial-logical correlation factor is calculated to quantify the degree of consistency between changes in wireless channel quality and changes in interference levels.

3. The 5G slicing online communication system for point cloud detection of digital twins of ancient buildings according to claim 2, characterized in that: The specific operation for outputting the multidimensional state vector is as follows: For each time slot, a structured multidimensional state vector is constructed, which includes a time identifier field, a terminal state set field, a service state set field, a network environment set field, and a comprehensive quality factor field. The terminal status set field includes the wireless channel status parameters, real-time bit error rate, and delay confidence weights calculated for the wireless channel status parameters and real-time bit error rate of each point cloud data acquisition terminal. The business status set field contains the aggregated business flow characteristics and their latency confidence weights for this detection business slice within the time slot; The network environment set field includes the resource utilization rate of the slice in this detection service, the interference level in the slice interference mapping relationship, and its delay confidence weight; The value of the comprehensive quality factor field is calculated as follows: the arithmetic mean of the latency confidence weights in the terminal state set field and the network environment set field is multiplied by the spatial-logical correlation factor, and then multiplied by the latency confidence weights in the service state set field. The result is the comprehensive quality factor.

4. The 5G slicing online communication system for point cloud detection of digital twins of ancient buildings according to claim 3, characterized in that: The specific operations for analyzing point cloud service flow sequences and wireless channel state parameters in the cross-domain prediction module are as follows: First, based on the comprehensive quality factor in the multidimensional state vector, only multidimensional state vectors with a comprehensive quality factor not lower than the quality filtering threshold are selected to form a high-quality state sequence. Next, the arrival rate of point cloud data packets in the recorded aggregated business flow features is extracted from the business state set field of each multidimensional state vector in the high-quality state sequence as the instantaneous rate of the point cloud business flow, forming a historical business flow sequence. The historical business flow sequence is then decomposed into a trend component reflecting the progress of long-term tasks, a periodic component reflecting the fixed scanning cycle of the equipment, and a residual component reflecting changes in scene details using time-series decomposition technology. The residual component is identified as a data burst component, and the peak amplitude and duration of the data burst component are then quantified to form a list of burst events described by burst intensity, burst start time, and burst duration. Meanwhile, for each multi-dimensional state vector in the high-quality state sequence, the wireless channel state parameters contained in its terminal state set field are analyzed. Combined with point cloud data to collect terminal location information, a spatiotemporal channel quality variation model describing the channel quality changes with location and time is constructed using a spatiotemporal kriging interpolation algorithm or a graph neural network-based propagation model. This spatiotemporal channel quality variation model is used to predict the channel quality indication at a specified time and predicted location in the future, and the confidence interval of the predicted channel quality indication is output simultaneously to characterize the prediction uncertainty.

5. A 5G slicing online communication system for point cloud detection of digital twins of ancient buildings according to claim 4, characterized in that: The specific steps for performing cross-domain correlation analysis and calculating the equivalent resource demand forecast curve are as follows: Perform cross-domain correlation analysis, and perform spatiotemporal alignment and correlation analysis between the list of burst events obtained based on the burst component quantization of data and the channel quality prediction information including the predicted channel quality indicator and its confidence interval obtained by using the spatiotemporal variation model of channel quality; calculate a cross-domain risk coefficient for each future moment within each predicted burst period; Based on the results of cross-domain correlation analysis, the equivalent resource demand forecast curve is calculated using the equivalent resource demand forecast function. The calculation process is as follows: First, the predicted value of the future basic service data rate is obtained by extrapolating the trend component and periodic component in the historical service flow sequence; then, the product of the predicted value of the future basic service data rate and the cross-domain risk coefficient adjusted by the value 1 plus the preset first weighting factor is calculated to obtain the first intermediate value; next, the predicted channel quality indicator is converted into the corresponding spectrum efficiency estimate through the preset mapping relationship that maps the channel quality indicator to spectrum efficiency. Then, the first intermediate value is divided by the estimated spectral efficiency to obtain the second intermediate value; Subsequently, the adjustment term is calculated. The adjustment term is equal to the value 1 plus the time delay sensitivity coefficient adjusted by the preset second weighting factor, and then the error tolerance coefficient adjusted by the preset third weighting factor is subtracted. Finally, the second intermediate value is multiplied by the adjustment term, and the result is the equivalent resource demand at that future moment. This calculation process is repeated for each future moment within the set time window to generate the equivalent resource demand prediction curve.

6. The 5G slicing online communication system for point cloud detection of digital twins of ancient buildings according to claim 5, characterized in that: The calculation process for the cross-domain risk coefficient is as follows: A1. Normalize the burst intensity obtained by quantizing the burst components of the data at this moment to obtain the normalized burst intensity; A2. Calculate the exponential function value with the natural constant as the base, where the exponent of the exponential function is negative. The numerator is the difference between the predicted channel quality indicator and a preset channel quality threshold, and the denominator is a preset scaling factor; A3. Calculate the difference between the value 1 and a ratio, where the ratio is the difference between the upper and lower bounds of the confidence interval of the predicted channel quality indicator divided by the entire range of values ​​of the channel quality indicator; Finally, multiply the normalized burst intensity obtained from A1, the exponential function value obtained from A2, and the difference obtained from A3. The product is the cross-domain risk coefficient at this future moment.

7. A 5G slicing online communication system for digital twin point cloud detection of antique buildings according to claim 6, characterized in that: The specific process of performing fast optimization calculations in the joint decision optimization module is as follows: First, the received equivalent resource demand prediction curve is fused with the latest multidimensional state vector, and a dynamic multi-objective constrained optimization problem with a future-oriented time window is constructed based on the two. The equivalent resource demand prediction curve is the core constraint, and the optimization variables include the time-frequency two-dimensional resource block set pre-allocated to each point cloud data acquisition terminal, the transmit power adjustment sequence, and the adaptive modulation and coding level combination. Next, a unified performance function is constructed as the objective function. Its value is the first weight coefficient multiplied by the resource utilization efficiency, minus the second weight coefficient multiplied by the interference performance loss, and then minus the third weight coefficient multiplied by the total energy consumption. Among them, the calculation of resource utilization efficiency introduces the evaluation correction factor of the satisfaction of the equivalent resource demand prediction curve, the calculation of interference performance loss is based on the ratio of the predicted interference of adjacent slices to the current interference level, and the total energy consumption is proportional to the sum of the power consumption of all point cloud data acquisition terminals. The three weight coefficients are dynamic weights, which are dynamically adjusted according to the changing trend of the equivalent resource demand prediction curve and the prediction uncertainty evaluation value.

8. A 5G slicing online communication system for point cloud detection of digital twins of ancient buildings according to claim 7, characterized in that: The specific operation of generating and outputting the joint scheduling strategy, which includes a time-frequency resource block pre-allocation scheme, transmit power adjustment value, and adaptive modulation and coding level suggestion, is as follows: An improved constrained multi-objective particle swarm optimization algorithm is adopted to solve dynamic multi-objective constrained optimization problems quickly. The algorithm includes three mechanisms: prediction-guided particle initialization, feasibility rule-driven particle updating, and Pareto optimal solution selection based on historical performance evaluation. Prediction-guided particle initialization prioritizes the allocation of initial resource blocks to candidate solutions during high-demand periods based on the equivalent resource demand prediction curve. Feasibility rule-driven particle updating penalizes candidate solutions that do not meet the constraints of the equivalent resource demand prediction curve and guides their search direction. Pareto optimal solution selection based on historical performance evaluation selects the most robust solution from the Pareto optimal solution set as the final result based on the historical channel prediction accuracy. The final solution is transformed into a joint scheduling strategy that includes a time-frequency resource block pre-allocation scheme, transmit power adjustment value, and adaptive modulation and coding level recommendations.

9. A 5G slicing online communication system for digital twin point cloud detection of antique buildings according to claim 8, characterized in that: In the execution and model evolution module, the specific operations for collecting actual performance parameters of point cloud data transmission and the communication resource consumption data, and calculating the deviation, are as follows: The joint scheduling strategy is issued and executed, and the actual transmission delay, actual throughput, actual block error rate, actual number of physical resource blocks used, actual average transmit power, actual channel quality indication and actual interference level are collected synchronously from the performance layer, resource layer and environment layer. Based on the collected actual performance parameters and the communication resource data consumed, the deviation of the guarantee requirement satisfaction is calculated. The calculation process is as follows: the actual guarantee requirement is deduced from the actual throughput, block error rate and the number of resource blocks; the equivalent resource requirement prediction value at the same time is retrieved. Calculate the ratio of the absolute difference between the predicted value and the actual value to the actual value to obtain the basic deviation ratio; Calculate the adjustment factor, which is 1 plus the sum of the first penalty term triggered by the actual delay exceeding the delay tolerance threshold and the second penalty term triggered by the actual block error rate exceeding the block error rate tolerance threshold; Multiply the basic deviation ratio by the adjustment factor to obtain the final deviation. Perform this calculation process at each moment within the set time window to obtain the deviation sequence composed of the deviations at each moment.

10. A 5G slicing online communication system for point cloud detection of digital twins of ancient buildings according to claim 9, characterized in that: The specific operation of adjusting the parameters using the channel quality spatiotemporal variation model for traffic flow decomposition and channel prediction in the deviation-driven cross-domain prediction module is as follows: Source analysis is performed on the deviation sequence, linking high deviation moments with historical state characteristics. If the burst component characteristics of the data in the historical state at that moment are significant, the traffic flow decomposition model is deemed insufficient in prediction. If the difference between the actual channel quality and the predicted value exceeds the tolerance, the spatiotemporal variation model of channel quality is deemed to be invalid. A learning confidence score is calculated for each deviation data point, which is the product of the exponential function value with the natural constant base and the comprehensive quality factor of the multidimensional state vector on which the deviation is based. The exponent of the exponential function is negative, the numerator is the deviation value, and the denominator is the recent deviation moving average. Using the learning confidence score as weight, the internal parameters of the model determined to be invalid are updated incrementally through an online gradient descent algorithm. At the same time, the collected actual data is fed back to the data fusion and synchronization module in a specified format to update the multidimensional state vector.

Images