Signal dynamic distribution control method based on multi-channel signal intelligent connector

Abstract

This invention discloses a dynamic signal allocation control method based on a multi-channel intelligent signal connector, belonging to the field of dynamic signal allocation technology. The method includes: obtaining a sensing confidence array based on the environmental fluctuation characteristics of the target industrial scene; compensating the sensing confidence array to obtain an environmental location confidence array; driving the dynamic allocation of physical channel resources of the multi-channel intelligent signal connector to collect high-confidence sensing data of the scene; performing real-time operational modeling and reconstruction of the target industrial scene, locating the distribution of missing data areas; triggering cross-modal signal allocation strategy updates based on the distribution, and transmitting back compensated data distribution; and obtaining a real-time digital twin model. This invention solves the technical problems of insufficient data acquisition accuracy and missing scene reconstruction data caused by environmental fluctuations and sensor positions in existing technologies, achieving real-time digital twin modeling of the target industrial scene and improving the accuracy of data acquisition and the reliability of scene reconstruction.

Description

Technical Field

[0001] This invention relates to the field of signal dynamic allocation technology, and more specifically to a signal dynamic allocation control method based on a multi-channel intelligent signal connector. Background Technology

[0002] In the field of intelligent monitoring and modeling in industrial scenarios, traditional signal acquisition and processing methods often face numerous challenges. On the one hand, complex environmental fluctuations, such as temperature changes and vibration interference, exist in industrial environments, leading to unstable sensor confidence levels and affecting the accuracy of data acquisition. On the other hand, sensor deployment locations exhibit spatial attenuation characteristics, resulting in significant differences in signal transmission quality at different locations, which can easily cause deviations in the perceived data. Furthermore, traditional signal connectors often employ fixed channel allocation, making it difficult to dynamically adjust resources based on real-time perception confidence levels. This leads to low efficiency in acquiring high-confidence data and often affects reconstruction accuracy during scene modeling due to data gaps, failing to meet the high requirements for data reliability and integrity in real-time digital twin modeling of industrial scenarios.

[0003] Existing technologies suffer from technical problems such as insufficient data acquisition accuracy and missing scene reconstruction data due to environmental fluctuations and sensor locations. Summary of the Invention

[0004] This application provides a signal dynamic allocation control method based on a multi-channel intelligent signal connector, which is used to address the technical problems of insufficient data acquisition accuracy and missing scene restoration data caused by environmental fluctuations and sensor positions in the prior art.

[0005] In view of the above problems, this application provides a signal dynamic allocation control method based on a multi-channel intelligent signal connector, the method comprising:

[0006] Based on the environmental fluctuation characteristics of the target industrial scenario, the sensing confidence of the multimodal sensing array in the target industrial scenario is iteratively updated online to obtain a sensing confidence array. Based on the spatial attenuation characteristics of the multimodal sensing array's deployment location, the sensing confidence array is compensated to obtain an environmental location confidence array. The environmental location confidence array drives the dynamic allocation of physical channel resources of the multi-channel signal intelligent connector to collect high-confidence sensing data of the scenario. Based on the high-confidence sensing data of the scenario, real-time operational modeling and reconstruction of the target industrial scenario is performed to locate the distribution of missing data areas. Based on the distribution of missing data areas, the cross-modal signal allocation strategy of the multi-channel signal intelligent connector is updated, and the compensation data distribution is transmitted back. Based on the compensation data distribution, spatiotemporal alignment compensation for the real-time operational modeling is performed to obtain a real-time digital twin model.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] Based on the environmental fluctuation characteristics of the target industrial scenario, a sensing confidence array is obtained. Based on the spatial attenuation characteristics of the multimodal sensing array's deployment location, the sensing confidence array is compensated to obtain an environmental location confidence array. The physical channel resources of the multi-channel signal intelligent connector are dynamically allocated to collect high-confidence sensing data of the scenario. Real-time operational modeling and reconstruction of the target industrial scenario are performed to locate the distribution of missing data areas. Based on the distribution of missing data areas, the cross-modal signal allocation strategy of the multi-channel signal intelligent connector is updated, and the compensated data distribution is transmitted back. Based on the compensated data distribution, spatiotemporal alignment compensation is performed for the real-time operational modeling to obtain a real-time digital twin model. This achieves the technical effect of realizing real-time digital twin modeling of the target industrial scenario, improving the accuracy of data acquisition and the reliability of scenario reconstruction. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating the signal dynamic allocation control method based on a multi-channel intelligent signal connector provided in this application embodiment;

[0011] Figure 2 This is a schematic diagram of the process for obtaining the sensing confidence array in the signal dynamic allocation control method based on a multi-channel signal intelligent connector provided in the embodiments of this application. Detailed Implementation

[0012] This application provides a signal dynamic allocation control method based on a multi-channel intelligent signal connector, which is used to address the technical problems of insufficient data acquisition accuracy and missing scene reconstruction data caused by environmental fluctuations and sensor positions in the prior art.

[0013] 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 a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] Examples, such as Figure 1As shown, this application provides a signal dynamic allocation control method based on a multi-channel signal intelligent connector, the method comprising:

[0015] Step S100: Based on the environmental fluctuation characteristics of the target industrial scene, perform online iterative updates of the sensing confidence of the multimodal sensing array in the target industrial scene to obtain the sensing confidence array.

[0016] Specifically, multi-source environmental sensing is performed on the target industrial scene to obtain real-time multi-source environmental data. Using the data acquisition timestamp as a starting point, historical related environmental data is traced back to obtain spatiotemporally related environmental slices. Spatiotemporal feature decoupling is performed on these slices, extracting historical time-cycle feature vectors and historical spatial grid feature matrices. The real-time multi-source environmental data is then mapped and aligned to the spatiotemporal reference of the spatiotemporally related environmental slices to obtain real-time time feature vectors and real-time spatial feature matrices. Furthermore, temporal and spatial fluctuation components are calculated. Finally, environmental sensitivity attributes of the target industrial scene are interactively obtained, and weights are configured to obtain environmental sensitivity weights. The array is weighted and fused with temporal and spatial fluctuation components to output the real-time environmental fluctuation scale. If the scale exceeds the preset fluctuation limit, multiple sample acquisition attenuation coefficient sets of multiple sample sensors in multiple sample environmental scene sets are interactively obtained. These are used to train a spatiotemporal convolutional network to obtain multiple sensor attenuation characteristic models. Real-time multi-source environmental data is loaded into the model to perform multi-threaded parallel prediction to obtain multiple theoretical signal quality benchmarks. Then, multiple confidence offsets of KL divergence quantification are calculated by combining multiple acquisition signal quality indicators of multiple sample sensors. The sensing confidence is iterated online by traversing the multimodal sensing array to finally obtain the sensing confidence array.

[0017] Step S200: Based on the spatial attenuation characteristics of the deployment location of the multimodal sensing array, compensate the sensing confidence array to obtain the environmental location confidence array.

[0018] Specifically, a three-dimensional attenuated potential field model is constructed based on the building layout and equipment layout information of the target industrial scene. Multiple visibility indices and multipath distortion factors of multiple sensing nodes in the multimodal sensing array are extracted based on this model. Multiple location compensation factors are calculated based on the extracted visibility indices and multipath distortion factors. According to the topology of the multimodal sensing array, the multiple location compensation factors are mapped into a location compensation factor array in the order of the sensing node indices. The location compensation factor array is used as the observation input to spatially fuse and reconstruct the sensing confidence array, outputting an environmental location confidence array.

[0019] Step S300: Based on the environmental location confidence array, dynamically allocate the physical channel resources of the multi-channel signal smart connector to collect high-confidence perception data of the scene.

[0020] Specifically, the physical connection topology between the multi-channel signal intelligent connector and the multi-modal sensing array is analyzed to generate a communication link array; the environmental location confidence array is converted into a discrete priority array; the channel type is assigned to the discrete priority array according to the priority interval to obtain a channel allocation array; after confidence-weighted collaborative time slot bandwidth scheduling is performed on the communication link array according to the channel allocation array, the high-confidence sensing data of the scenario is transferred back through the multi-channel signal intelligent connector.

[0021] Step S400: Based on the high-confidence perception data of the scenario, perform real-time operation modeling and reconstruction of the target industrial scenario, and locate the distribution of data missing areas.

[0022] Specifically, based on the depth distribution data of the target industrial scene, a three-dimensional mesh of the scene is constructed by integrating high-confidence perception data of the scene; 26-neighborhood connectivity analysis is performed on the three-dimensional mesh of the scene to locate multiple three-dimensional void regions; by binding the grid vertex confidence of these multiple three-dimensional void regions, the distribution of missing data regions with labeled probability distribution is output.

[0023] Step S500: Based on the distribution of the missing data regions, trigger the cross-modal signal allocation strategy update of the multi-channel signal smart connector and send back the compensation data distribution.

[0024] Specifically, based on the distribution of missing data areas obtained from positioning, the multi-channel signal intelligent connector is triggered to start the cross-modal signal allocation strategy update mechanism. By adjusting the allocation weight of different modal sensing signals in physical channel resources, transmission path and time slot bandwidth, the sensing capabilities of other modal sensors in the multi-modal sensing array are used to fill the missing data areas, so that the system can transmit compensation data distribution that covers the missing areas, ensuring the integrity of the sensing data and the accuracy of scene restoration.

[0025] Step S600: Perform spatiotemporal alignment compensation for the real-time operation modeling based on the compensation data distribution to obtain a real-time digital twin model.

[0026] Specifically, based on the distribution of compensation data, real-time operational modeling of the target industrial scenario is performed. In the time dimension, the collection timestamps of the compensation data and the original modeling data are aligned to ensure temporal consistency. In the spatial dimension, the spatial coordinate information of the compensation data is accurately mapped to the corresponding data missing areas in the real-time operational modeling, and the empty areas in the three-dimensional mesh are filled and corrected. Through the alignment and compensation in both the time and space dimensions, the deviation caused by data missing in the real-time operational modeling is corrected, and finally a real-time digital twin model that can accurately reflect the real-time state of the target industrial scenario is generated.

[0027] In one possible implementation, such as Figure 2 As shown, step S100 further includes:

[0028] Step S110: Perform multi-source environmental perception on the target industrial scene to obtain real-time multi-source environmental data.

[0029] Step S120: Starting from the collection timestamp of the real-time multi-source environmental data, perform historical correlation environmental data backtracking to obtain spatiotemporal correlation environmental slices.

[0030] Step S130: By comparing the real-time multi-source environmental data with the spatiotemporally correlated environmental slices, output the real-time environmental fluctuation scale.

[0031] Step S140: If the real-time environmental fluctuation scale exceeds the preset fluctuation limit, the sensing confidence of the multimodal sensing array is updated online based on the real-time multi-source environmental data to obtain the sensing confidence array.

[0032] Specifically, for the target industrial scenario, various sensors (such as temperature sensors, humidity sensors, vibration sensors, electromagnetic sensors, etc.) deployed in the multimodal sensing array within the scenario are used to comprehensively perceive the scenario environment from multiple sources and dimensions. Real-time data reflecting the environmental state, including but not limited to temperature, humidity, vibration frequency, electromagnetic intensity, and air pressure, are collected and integrated to form real-time multi-source environmental data covering multiple characteristics of the scenario environment.

[0033] Starting with the acquisition timestamp of real-time multi-source environmental data, the system calls the environmental data storage system interface of the target industrial scenario to retrieve historical environmental data from the same monitoring dimensions (such as temperature, humidity, vibration, etc.) and spatial monitoring points as the real-time data according to a preset time backtracking period (such as the last 24 hours, the last 7 days, etc.). The retrieved historical data is timestamped to ensure that it is consistent with the time reference of the real-time data, and then structured and integrated according to time series and spatial distribution location to form a dataset containing historical time series data and corresponding spatial monitoring point coordinate information. The dataset is then associated and mapped with the spatiotemporal attributes of the real-time multi-source environmental data to generate spatiotemporally associated environmental slices that can reflect the changing trend of environmental data in the time dimension and the distribution characteristics in the spatial dimension.

[0034] First, the spatiotemporal features of the spatiotemporally correlated environmental slices are decoupled. Historical time-cycle feature vectors (such as periodic fluctuation patterns and trend changes) are extracted through time series decomposition, and historical spatial grid feature matrices (such as the distribution differences and correlations of environmental parameters in different regions) are extracted through spatial grid division. Then, real-time multi-source environmental data are mapped and aligned according to the temporal granularity and spatial division criteria of the spatiotemporally correlated environmental slices to obtain the corresponding real-time temporal feature vectors and real-time spatial feature matrices. Subsequently, the difference between the real-time temporal feature vectors and historical time-cycle feature vectors is calculated to obtain the temporal fluctuation component, and the difference between the real-time spatial feature matrix and the historical spatial grid feature matrix is ​​calculated to obtain the spatial fluctuation component. Finally, environmental sensitivity weight arrays are obtained by configuring weights based on the environmental sensitivity attributes of the target industrial scenario. The temporal and spatial fluctuation components are then weighted and fused using this array to output a real-time environmental fluctuation scale that can quantitatively reflect the degree of difference between the current environment and the historical correlated environment.

[0035] When the real-time environmental fluctuation scale exceeds the preset fluctuation limit, multiple sample acquisition attenuation coefficient sets of various sample sensors in multiple sample environmental scene sets are obtained interactively. These sample environmental scene sets and sample acquisition attenuation coefficient sets are used as training data to input spatiotemporal convolutional networks for training, resulting in multiple sensor attenuation characteristic models. Subsequently, real-time multi-source environmental data is loaded into these sensor attenuation characteristic models to perform multi-threaded parallel prediction, resulting in multiple theoretical signal quality benchmarks. Then, multiple acquisition signal quality indicators of the above-mentioned multiple sample sensors are obtained interactively. These indicators are used to perform KL divergence quantification on the theoretical signal quality benchmarks to obtain multiple confidence offsets. Finally, these confidence offsets are used to traverse the multimodal sensing array, and the confidence of each sensing node is updated online iteratively, ultimately obtaining the sensing confidence array.

[0036] In one possible implementation, step S130 further includes:

[0037] Step S131: Decouple the spatiotemporal features of the spatiotemporal associated environment slices and extract the historical time period feature vector and the historical spatial grid feature matrix.

[0038] Step S132: Map and align the real-time multi-source environmental data to the spatiotemporal reference of the spatiotemporal correlated environmental slice to obtain the real-time temporal feature vector and the real-time spatial feature matrix.

[0039] Step S133: Calculate the time-domain fluctuation component based on the real-time time feature vector and the historical time period feature vector.

[0040] Step S134: Calculate the spatial fluctuation component based on the real-time spatial feature matrix and the historical spatial grid feature matrix.

[0041] Step S135: Output the real-time environmental fluctuation scale by synthesizing the temporal fluctuation component and the spatial fluctuation component.

[0042] Specifically, for spatiotemporally correlated environmental slices, time series decomposition algorithms (such as seasonal-trend decomposition) are used to process the time dimension data contained therein, separating periodic components, trend components, and residual components. The periodic components are quantized into vector form according to time intervals to obtain historical time period feature vectors, which cover the fluctuation patterns of environmental parameters within different time periods. At the same time, the spatial dimension data of spatiotemporally correlated environmental slices are divided into grids according to the spatial coordinate system of the target industrial scene, dividing the scene into several regular spatial grid units. The mean, extreme values, and correlation degree of environmental parameters with adjacent grids within each grid unit are extracted and arranged in grid index order to form a matrix, thereby obtaining the historical spatial grid feature matrix, thus achieving decoupled extraction of spatiotemporal features.

[0043] First, the temporal reference (such as sampling frequency and time period division rules) and spatial reference (such as grid cell size and coordinate mapping rules) of the spatiotemporally correlated environmental slices are analyzed to establish a unified spatiotemporal mapping standard. Then, the timestamps of real-time multi-source environmental data are calibrated to ensure consistency with the temporal granularity of the spatiotemporally correlated environmental slices. Real-time time series data are matched to the corresponding time period intervals through interpolation or resampling methods to generate real-time time feature vectors. Simultaneously, the acquisition location coordinates of real-time multi-source environmental data are mapped to the corresponding spatial grid cells according to the spatial reference. The statistical values ​​of real-time environmental parameters (such as mean and peak values) within each grid cell are calculated and arranged in grid index order to form a real-time spatial feature matrix, thereby completing the spatiotemporal reference mapping and alignment between real-time multi-source environmental data and spatiotemporally correlated environmental slices.

[0044] While ensuring consistency between the real-time time feature vector and the historical time period feature vector in terms of time period division and parameter dimensions, the temporal fluctuation component is obtained by calculating the difference between the two. Using vector interpolation, the parameter values ​​of corresponding dimensions in the two vectors are subtracted point by point to obtain the difference in each dimension. Then, by taking the square root of the sum of squares of these differences (i.e., Euclidean distance calculation), the differences in each dimension are integrated into a comprehensive quantitative value, which serves as the temporal fluctuation component. This component can intuitively reflect the degree of difference in the changes of real-time time features and historical time period features in the time dimension.

[0045] While ensuring consistency between the real-time spatial feature matrix and the historical spatial grid feature matrix in terms of grid cell division and parameter index dimensions, the spatial fluctuation component is obtained by calculating the difference between the two. A matrix difference quantization method is employed, calculating the point-by-point difference of the parameter values ​​of corresponding grid cells in the two matrices to obtain the difference for each grid cell. Then, by calculating the F-norm of the matrix (i.e., the square root of the sum of squares of all elements) or averaging the absolute values ​​of the differences between each grid cell, the spatial dimensional differences are integrated into a comprehensive quantified value, which serves as the spatial fluctuation component. This component can intuitively reflect the degree of difference in the spatial distribution between real-time spatial features and historical spatial grid features.

[0046] The environmental sensitivity attributes of the target industrial scenario are obtained through interactive methods. These attributes reflect the sensitivity of the scenario to different environmental factors (such as temperature fluctuations and vibration intensity). Based on these environmental sensitivity attributes, weights are configured to assign corresponding weight values ​​to the temporal and spatial fluctuation components, forming an environmental sensitivity weight array. This environmental sensitivity weight array is then used to perform weighted fusion calculations on the temporal and spatial fluctuation components. Specifically, by adding the products of the temporal fluctuation component and its corresponding weight, and the products of the spatial fluctuation component and its corresponding weight, a comprehensive quantitative value is obtained. This value is used as a real-time environmental fluctuation scale and output. This scale can comprehensively reflect the degree of environmental fluctuation in the target industrial scenario in both time and space dimensions.

[0047] In one possible implementation, step S135 further includes:

[0048] Step S1351: Interact to obtain the environmental sensitivity attributes of the target industrial scene.

[0049] Step S1352: Configure weights according to the environmental sensitivity attributes to obtain an environmental sensitivity weight array.

[0050] Step S1353: Perform weighted fusion of the temporal fluctuation component and the spatial fluctuation component based on the environmentally sensitive weight array, and output the real-time environmental fluctuation scale.

[0051] Specifically, the environmental sensitivity attributes of the scenario are obtained through system interaction modules (such as human-machine interface, preset parameter input interface or data interaction link with the target industrial scenario management system). These attributes cover the sensitivity of the scenario to fluctuations in different environmental parameters such as temperature, humidity, vibration, and electromagnetic fields. For example, which changes in environmental parameters have a greater impact on the operation of equipment and the stability of production processes within the scenario, thereby clarifying the key focus dimensions of the scenario in environmental perception.

[0052] Based on the environmental sensitivity attributes of the target industrial scenario, corresponding weight values ​​are assigned to the temporal and spatial fluctuation components. Higher weights are given to fluctuation components associated with highly sensitive environmental factors in the scenario (such as temperature-related temporal or spatial fluctuations in high-temperature sensitive scenarios), while lower weights are given to fluctuation components associated with less sensitive environmental factors. These weight values ​​for the temporal and spatial domains are arranged in a preset parameter dimension order to form an environmental sensitivity weight array containing temporal and spatial weight coefficients, thereby quantifying the impact priority of different fluctuation components in the scenario.

[0053] Based on the obtained environmentally sensitive weight array, the weight coefficients of the corresponding temporal fluctuation components and the corresponding spatial fluctuation components are extracted. The temporal fluctuation component is multiplied by its real-time weight coefficient to obtain the temporal weighted fluctuation value, and the spatial fluctuation component is multiplied by its corresponding weight coefficient to obtain the spatial weighted fluctuation value. Then, the temporal weighted fluctuation value and the spatial weighted fluctuation value are summed, and the result is used as a real-time environmental fluctuation scale that can comprehensively reflect the degree of environmental fluctuation in the target industrial scene in the time and space dimensions and is output. Because this scale combines the environmentally sensitive attribute weight configuration of the scene, it can accurately reflect the environmental fluctuation characteristics that have a significant impact on the scene.

[0054] In one possible implementation, step S140 further includes:

[0055] Step S141: Interact to obtain multiple sample acquisition attenuation coefficient sets of multiple sample sensors in multiple sample environment scene sets.

[0056] Step S142: Use the multiple sample environment scene sets and multiple sample acquisition attenuation coefficient sets as training data to train a spatiotemporal convolutional network and obtain multiple sensor attenuation characteristic models.

[0057] Step S143: When the real-time environmental fluctuation scale exceeds the preset fluctuation limit, the real-time multi-source environmental data is loaded into the multiple sensor attenuation characteristic models to perform multi-threaded parallel prediction and obtain multiple theoretical signal quality benchmarks.

[0058] Step S144: Interact to obtain multiple acquisition signal quality indicators from the various sample sensors, and use the multiple acquisition signal quality indicators to perform KL divergence quantification of the multiple theoretical signal quality benchmarks to obtain multiple confidence offsets.

[0059] Step S145: Use the multiple confidence offsets to traverse the multimodal sensing array and perform online iteration of sensing confidence to obtain the sensing confidence array.

[0060] Specifically, through system interaction interfaces (such as human-machine input interfaces, sensor parameter database calls, or communication links with the sample data management system), multiple sets of sample acquisition attenuation coefficients corresponding to various types of sample sensors (such as temperature sensors, humidity sensors, vibration sensors, etc.) under multiple preset sample environment scenario sets (covering simulated environmental conditions such as different temperatures, humidity, vibration intensity, and electromagnetic interference) are obtained. These attenuation coefficient sets reflect the attenuation parameters of various sensors when acquiring signals in different environmental scenarios, including key indicators such as signal strength attenuation value and data accuracy attenuation rate.

[0061] Multiple sample environmental scene sets (containing parameter data under different environmental conditions) and corresponding multiple sample acquisition attenuation coefficient sets (reflecting the signal attenuation characteristics of the sensor in the corresponding scene) are input as training data into a spatiotemporal convolutional network. The sample environmental scene sets serve as the network's input features, and the sample acquisition attenuation coefficient sets serve as the expected output labels. The network's temporal convolutional layers capture the characteristics of environmental parameters changing over time, and the spatial convolutional layers extract the spatial distribution features under different scenes. After multiple rounds of forward propagation, the error between the predicted value and the actual label is calculated, and the network weight parameters are adjusted through a backpropagation algorithm to minimize the error. When the network training reaches a convergent state (i.e., the prediction error is below a preset threshold), the trained network parameters are saved, resulting in multiple sensor attenuation characteristic models corresponding to each type of sensor sample. These models can characterize the signal attenuation law of the sensor under different environmental scenes.

[0062] When the real-time environmental fluctuation scale exceeds the preset fluctuation limit, real-time multi-source environmental data (including multi-dimensional environmental parameters such as temperature, humidity, and vibration collected in the target industrial scene) are input into multiple sensor attenuation characteristic models trained respectively. The models are driven to run simultaneously through multi-threaded parallel processing. By using the learning results of the models on the signal transmission attenuation law of different sensors in the current environment, the theoretical quality parameters of each sensor when collecting signals in the current real-time environment are predicted. These parameters are integrated to form multiple theoretical signal quality benchmarks, which are used to evaluate the credibility of the actual data collected by the sensors.

[0063] Multiple signal quality indicators collected by various sample sensors during actual operation are obtained through interactive methods. These indicators include parameters that reflect the actual signal quality, such as signal-to-noise ratio, transmission rate stability, and data error rate. These collected signal quality indicators are used as actual reference data and compared with multiple theoretical signal quality benchmarks. The KL divergence algorithm is used to calculate the probability distribution difference between the two, quantifying the deviation between the theoretical prediction value and the actual collected value. Each theoretical signal quality benchmark corresponds to a quantization result, which are multiple confidence offsets used to adjust the sensor's sensing confidence level in the future.

[0064] The obtained confidence offsets are matched with sensor nodes in the multimodal sensing array according to sensor type. Then, each sensor node in the array is traversed in a preset order. The initial sensing confidence of each node is adjusted according to the matched confidence offset. If the offset is positive, it means that the theoretical signal quality benchmark is higher than the actual acquisition index, and the confidence of the corresponding node needs to be reduced. If the offset is negative, the confidence is increased. Through multiple rounds of iterative updates, until the change in the confidence of each node is less than the preset threshold, a sensing confidence array that can accurately reflect the reliability of sensor data in the current environment is finally formed.

[0065] In one possible implementation, step S200 further includes:

[0066] Step S210: Construct a three-dimensional attenuation potential field model based on the building layout information and equipment layout information of the target industrial scene.

[0067] Step S220: Extract multiple visibility indices and multipath distortion factors of multiple sensing nodes based on the three-dimensional attenuation potential field model.

[0068] Step S230: Calculate multiple position compensation factors based on the multiple visibility indices and multipath distortion factors.

[0069] Step S240: Based on the topology of the multimodal sensing array, map the multiple position compensation factors into a position compensation factor array according to the sensor node index order.

[0070] Step S250: Using the location compensation factor array as the observation input, perform spatial fusion reconstruction on the perception confidence array, and output the environmental location confidence array.

[0071] Specifically, the system collects building layout information of the target industrial scene, including the location coordinates, material properties (such as concrete, metal, glass, etc.) and structural dimensions of walls, columns, ceilings, and floors. At the same time, it collects equipment layout information within the scene, covering the three-dimensional spatial position, geometry, and material parameters of various production equipment, pipes, metal frames, etc. This information is then imported into a 3D modeling tool to construct a physical space model of the scene. Based on electromagnetic wave propagation theory, the system simulates the propagation path, reflection, refraction, and attenuation process of signals in this space using ray tracing or finite element analysis. Finally, it generates a three-dimensional attenuation potential field model that can quantify the degree of signal attenuation at different spatial locations. This model can intuitively reflect the variation of signal strength with spatial location in the scene.

[0072] For each sensing node in the multimodal sensing array, after determining its spatial coordinates in a three-dimensional attenuated potential field model, the signal propagation path is simulated using a ray projection algorithm to determine whether there is an unobstructed direct path between the node and the signal source, and the proportion of direct signal strength is calculated as the visibility index. At the same time, the number of multipaths formed by the signal reflection from building structures, equipment, etc. in the scene is counted. By calculating the amplitude difference and phase difference between each multipath signal and the direct signal, the mean square error formula is used to quantify the degree of distortion of the original signal by the multipath signal, and the multipath distortion factor is obtained. This completes the extraction of the corresponding parameters for each sensing node.

[0073] For each sensor node, its visibility index and multipath distortion factor are standardized to ensure they are within the same numerical range. Then, a preset weighted formula is used for calculation, such as multiplying the visibility index by a positive weighting coefficient and the multipath distortion factor by a negative weighting coefficient, and then adding the two results to obtain the position compensation factor for that node. The weighting coefficients can be adjusted according to the sensitivity of signal propagation to occlusion and multipath effects in the target industrial scene, ultimately generating a corresponding position compensation factor for each sensor node.

[0074] First, the topology information of the multimodal sensing array is obtained, and the index number and arrangement order of each sensing node are determined. Then, the correspondence between the position compensation factor and the sensing node index is established. According to the index number of each sensing node, the obtained multiple position compensation factors are sequentially filled into the preset array structure to ensure that the arrangement order of the compensation factors is completely consistent with the index order of the sensing nodes in the array, and finally a position compensation factor array that matches the topology of the multimodal sensing array is formed.

[0075] The obtained location compensation factor array is used as the observation input, and a spatial fusion algorithm (such as a Bayesian estimation or Kalman filtering method) is employed to fuse it with the sensing confidence array. During the fusion process, the confidence level of the corresponding sensing nodes in the sensing confidence array is adjusted according to the magnitude of the location compensation factor. For nodes with higher location compensation factors, their confidence weight in the array is increased, and vice versa. Through this spatial dimension reconstruction optimization, the final output is an environmental location confidence array that comprehensively reflects environmental influences and location attenuation characteristics.

[0076] In one possible implementation, step S300 further includes:

[0077] Step S310: Analyze the physical connection topology between the multi-channel signal smart connector and the multimodal sensing array to generate a communication link array.

[0078] Step S320: Convert the environmental location confidence array into a discrete priority array.

[0079] Step S330: Assign channel types to the discrete priority array according to the priority range to obtain the channel allocation array.

[0080] Step S340: After performing confidence-weighted collaborative time slot bandwidth scheduling of the communication link array according to the channel allocation array, the high-confidence perception data of the scene is transferred back through the multi-channel signal intelligent connector.

[0081] Specifically, by scanning and analyzing the physical connection relationship between the multi-channel signal intelligent connector and the multimodal sensing array, the corresponding connection relationship between each sensing node and each channel interface of the connector is clarified, including information such as the type of connection line, the properties of the transmission medium, and the direction of signal flow. This physical connection information is abstracted into structured data containing node identifiers, link identifiers, connection port numbers, and transmission parameters, and arranged in a preset format to form a communication link array. This array can clearly reflect the physical path network of signal transmission in the entire sensing system.

[0082] The confidence values ​​of each sensor node in the environmental location confidence array are divided into intervals, with multiple consecutive confidence threshold ranges preset (such as [0.8, 1.0], [0.6, 0.8], [0.4, 0.6], etc.). Each range corresponds to a discrete priority level (such as level 1, level 2, level 3, etc.). Each element in the environmental location confidence array is traversed, and its confidence value is converted into the corresponding priority level according to the threshold range to which it belongs. Finally, a discrete priority array composed of these priority levels is formed, so that the originally continuous confidence information is transformed into a discrete priority identifier that can be used for resource allocation.

[0083] Different priority ranges and channel types are pre-defined. For example, high priority ranges (such as levels 1-2) correspond to dedicated channels with low latency and high bandwidth, medium priority ranges (such as levels 3-4) correspond to regular transmission channels, and low priority ranges (such as levels 5 and above) correspond to shared channels or backup channels. Each priority element in the discrete priority array is traversed, and the corresponding channel type is assigned to it according to its priority range. Finally, a channel allocation array composed of the channel types of each node is formed, realizing differentiated configuration of channel resources based on priority.

[0084] Based on the channel allocation array, and combined with the confidence values ​​of each link in the communication link array, a weighted calculation is performed to allocate appropriate time slots and bandwidth resources to links of different channel types. For links corresponding to high-priority channels, more optimal time slot positions and a larger bandwidth proportion are allocated to ensure the real-time performance and stability of their data transmission; for links corresponding to medium- and low-priority channels, the remaining time slots and bandwidth are reasonably allocated without affecting the transmission of high-priority links. A collaborative scheduling mechanism coordinates the transmission time of each link to avoid signal conflicts and ensure the orderly transmission of multi-link data. After scheduling is completed, the high-confidence perception data collected by each link is aggregated and transferred through a multi-channel intelligent signal connector, and finally transmitted back to the data processing center, achieving efficient and accurate transmission of high-confidence perception data.

[0085] In one possible implementation, step S400 further includes:

[0086] Step S410: Based on the depth distribution data of the target industrial scene, fuse the high-confidence perception data of the scene to obtain a three-dimensional mesh of the scene.

[0087] Step S420: Perform 26-neighbor connectivity analysis on the three-dimensional mesh of the scene to locate multiple three-dimensional hole regions.

[0088] Step S430: Bind the grid vertex confidence of the multiple three-dimensional hole regions, and output the distribution of the missing data regions with an identifier probability distribution.

[0089] Specifically, the process involves acquiring depth distribution data of the target industrial scene, which includes the three-dimensional spatial coordinates and depth information of various objects and structures within the scene, thus outlining the physical contours of the scene. Subsequently, high-confidence perception data of the scene (such as environmental parameter data like temperature, pressure, and vibration) is spatiotemporally aligned with the depth distribution data. A point cloud fusion algorithm is then used to associate the attribute information of the perception data with the corresponding three-dimensional spatial location. After mesh generation and data interpolation, the discrete perception data is integrated into a continuous three-dimensional mesh structure, ultimately generating a three-dimensional mesh of the scene that simultaneously reflects the physical structure of the scene and the high-confidence perception information.

[0090] For the constructed 3D scene mesh, a 26-neighborhood analysis is performed on each mesh cell, which checks the connection status of the mesh cell with 26 neighboring mesh cells along the positive and negative directions of the x, y, and z coordinate axes and the diagonal direction. By determining whether the mesh cell contains valid sensing data, mesh cells that do not contain valid data are marked as hollow cells. Then, a connected component analysis algorithm (such as the region growing method) is used to aggregate adjacent hollow cells into continuous 3D regions, thereby locating multiple 3D hollow regions in the scene 3D mesh. These regions are the parts not covered by sensing data.

[0091] For each 3D void region, the coordinates and corresponding confidence values ​​of all its grid vertices are obtained through a boundary extraction algorithm, and a vertex-confidence mapping table is established. Then, the K-Nearest Neighbors (KNN) algorithm is used to perform cluster analysis on the vertex confidence values ​​in each void region, and the proportion of vertices in different confidence intervals is statistically analyzed. Next, based on the Bayesian probability model, the overall data missing probability of the void region is calculated using the vertex confidence distribution as the prior probability. Finally, the spatial coordinates of all void regions and their corresponding missing probabilities are visualized and mapped using a heatmap generation algorithm, and the distribution of data missing regions is output as indicated by probability gradients.

[0092] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0093] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0094] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for signal dynamic allocation control based on a multi-channel signal intelligent connector, characterized in that, The method includes: Based on the environmental fluctuation characteristics of the target industrial scenario, the sensing confidence of the multimodal sensing array in the target industrial scenario is updated online to obtain the sensing confidence array. Based on the spatial attenuation characteristics of the deployment location of the multimodal sensing array, the sensing confidence array is compensated to obtain the environmental location confidence array; The physical channel resources of the multi-channel signal smart connector are dynamically allocated based on the environmental location confidence array to collect high-confidence perception data of the scene; Based on the high-confidence perception data of the scenario, the real-time operation modeling and reconstruction of the target industrial scenario is performed to locate the distribution of data missing areas; The cross-modal signal allocation strategy of the multi-channel signal intelligent connector is updated based on the distribution of the missing data regions, and the compensation data distribution is transmitted back. Based on the distribution of the compensation data, spatiotemporal alignment compensation is performed for the real-time operation modeling to obtain a real-time digital twin model.

2. The signal dynamic allocation control method based on a multi-channel intelligent signal connector as described in claim 1, characterized in that, Based on the environmental fluctuation characteristics of the target industrial scenario, the sensing confidence of the multimodal sensing array in the target industrial scenario is iteratively updated online to obtain a sensing confidence array. The method includes: Multi-source environmental perception is performed on the target industrial scene to obtain real-time multi-source environmental data; Starting from the timestamp of the real-time multi-source environmental data collection, historical associated environmental data is traced back to obtain spatiotemporal associated environmental slices; By comparing the real-time multi-source environmental data with spatiotemporally correlated environmental slices, the real-time environmental fluctuation scale is output. If the real-time environmental fluctuation scale exceeds the preset fluctuation limit, the sensing confidence of the multimodal sensing array is updated online based on the real-time multi-source environmental data to obtain the sensing confidence array.

3. The signal dynamic allocation control method based on a multi-channel intelligent signal connector as described in claim 2, characterized in that, The method includes outputting the real-time environmental fluctuation scale by comparing the real-time multi-source environmental data with spatiotemporally correlated environmental slices. Spatiotemporal feature decoupling is performed on the spatiotemporal associated environment slices to extract historical time period feature vectors and historical spatial grid feature matrices; The real-time multi-source environmental data is mapped and aligned to the spatiotemporal reference of the spatiotemporally correlated environmental slice to obtain the real-time temporal feature vector and the real-time spatial feature matrix; The time-domain fluctuation component is calculated based on the real-time time feature vector and the historical time period feature vector; The spatial fluctuation components are calculated based on the real-time spatial feature matrix and the historical spatial grid feature matrix. The real-time environmental fluctuation scale is output by synthesizing the temporal fluctuation component and the spatial fluctuation component.

4. The signal dynamic allocation control method based on a multi-channel intelligent signal connector as described in claim 3, characterized in that, The method includes synthesizing the temporal and spatial fluctuation components to output the real-time environmental fluctuation scale. Interactively obtain the environmental sensitivity attributes of the target industrial scenario; Based on the environmental sensitivity attributes, weights are configured to obtain an environmental sensitivity weight array; The time-domain fluctuation component and the spatial-domain fluctuation component are weighted and fused according to the environmentally sensitive weight array to output the real-time environmental fluctuation scale.

5. The signal dynamic allocation control method based on a multi-channel intelligent signal connector as described in claim 2, characterized in that, If the real-time environmental fluctuation scale exceeds a preset fluctuation limit, the sensing confidence of the multimodal sensing array is iteratively updated online based on the real-time multi-source environmental data to obtain the sensing confidence array. The method includes: Interactively obtain multiple sample acquisition attenuation coefficient sets from various sample sensors in multiple sample environment scenarios; The multiple sample environment scene sets and multiple sample acquisition attenuation coefficient sets are used as training data to train a spatiotemporal convolutional network, resulting in multiple sensor attenuation characteristic models. When the real-time environmental fluctuation scale exceeds the preset fluctuation limit, the real-time multi-source environmental data is loaded into the multiple sensor attenuation characteristic models to perform multi-threaded parallel prediction and obtain multiple theoretical signal quality benchmarks. Multiple signal quality indices from the various sample sensors are obtained interactively, and the KL divergence of the multiple theoretical signal quality benchmarks is quantified using the multiple signal quality indices to obtain multiple confidence offsets. The sensing confidence array is obtained by traversing the multimodal sensing array using the multiple confidence offsets and performing online iteration of the sensing confidence.

6. The signal dynamic allocation control method based on a multi-channel intelligent signal connector as described in claim 1, characterized in that, Based on the spatial attenuation characteristics of the multimodal sensing array's deployment location, the sensing confidence array is compensated to obtain an environmental location confidence array. The method includes: Based on the building layout information and equipment layout information of the target industrial scenario, a three-dimensional attenuation potential field model is constructed. Based on the three-dimensional attenuation potential field model, multiple visibility indices and multipath distortion factors of multiple sensing nodes are extracted. Multiple location compensation factors are calculated based on the multiple visibility indices and multipath distortion factors; Based on the topology of the multimodal sensing array, the multiple position compensation factors are mapped into a position compensation factor array according to the sensor node index order. Using the location compensation factor array as the observation input, the perception confidence array is spatially fused and reconstructed to output the environmental location confidence array.

7. The signal dynamic allocation control method based on a multi-channel intelligent signal connector as described in claim 1, characterized in that, The method involves dynamically allocating physical channel resources of the multi-channel signal smart connector based on the environmental location confidence array to collect high-confidence scene perception data. The physical connection topology between the multi-channel signal intelligent connector and the multimodal sensing array is analyzed to generate a communication link array; Convert the environmental location confidence array into a discrete priority array; Based on the priority range, channel types are assigned to the discrete priority array to obtain the channel allocation array; After performing confidence-weighted collaborative time slot bandwidth scheduling of the communication link array according to the channel allocation array, the high-confidence perception data of the scene is transferred back through the multi-channel signal intelligent connector.

8. The signal dynamic allocation control method based on a multi-channel intelligent signal connector as described in claim 1, characterized in that, Based on the high-confidence perception data of the aforementioned scenario, real-time operational modeling and reconstruction of the target industrial scenario are performed to locate the distribution of missing data areas. The method includes: Based on the depth distribution data of the target industrial scene, and by fusing the high-confidence perception data of the scene, a three-dimensional mesh of the scene is obtained; Perform 26-neighbor connectivity analysis on the 3D mesh of the scene to locate multiple 3D hole regions; Bind the grid vertex confidence of the multiple three-dimensional void regions and output the distribution of the missing data regions with an identifier probability distribution.

Images