An information intelligent collection method and system based on scene design feedback

Abstract

The application discloses a kind of information intelligent acquisition method and system based on scene design feedback, it is related to scene intelligent information acquisition technical field, including establishing scene three-dimensional model obtains scene domain, based on element division rule to element division to obtain element set Scene domain;Based on event generation mechanism generates dynamic event, sets the dynamic interface of element set, for receiving dynamic event and feedback dynamic feedback data;Based on dynamic feedback data generation interface heat map, for reflecting the activity of dynamic interface;Based on activity dynamic adjustment of dynamic feedback data acquisition frequency, dynamic feedback data is input into feedback database, and multiple data groups are generated by analysis;Based on the connection between element set Generation element graph, based on element graph analysis data group generates scene optimization plan;The application realizes real-time accurate regulation and control to data acquisition and transmission process by the synergistic effect of dynamic valve control, interface heat map and element graph.

Description

Technical Field

[0001] This invention relates to the field of scene intelligent information acquisition technology, specifically to an intelligent information acquisition method and system based on scene design feedback. Background Technology

[0002] In the field of intelligent scene information acquisition, existing information acquisition and optimization methods mainly rely on static data acquisition and rule-driven decision-making mechanisms, but these technologies have significant limitations in dynamic and complex environments. With the development of IoT, digital twins and artificial intelligence technologies, the demand for real-time, adaptive and precise optimization capabilities is growing, and traditional methods can hardly meet the needs of practical applications.

[0003] Traditional methods employ fixed-frequency data acquisition and threshold-triggered mechanisms, such as periodic sensor readings or simple conditional judgments, which are ill-suited to handling sudden load changes in dynamic environments. Furthermore, optimization strategies based on isolated data analysis lack consideration for the complex relationships between scene elements, leading to fragmented decision-making and hindering holistic optimization. Based on the above analysis, the shortcomings of existing technologies include: 1. Rigid resource control, unable to adaptively adjust acquisition frequency and transmission bandwidth according to real-time needs; 2. One-sided optimization analysis, with knowledge graphs disconnected from real-time data, failing to capture deep semantic relationships across multiple data sources; 3. Weak verification process, with direct deployment of optimization solutions risking cascading failures.

[0004] Therefore, developing an intelligent information collection method based on scenario design feedback is of great significance. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent information collection method and system based on scenario design feedback, so as to solve the problems in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent information acquisition method based on scenario design feedback, comprising:

[0007] A 3D model of the scene is established to obtain the scene domain. Based on the element partitioning rules, the scene domain is partitioned into elements to obtain the element set.

[0008] Dynamic events are generated based on the event generation mechanism, and a dynamic interface for the feature set is set to receive dynamic events and provide dynamic feedback data.

[0009] An interface heatmap is generated based on dynamic feedback data to reflect the activity level of the dynamic interface.

[0010] The collection frequency of dynamic feedback data is dynamically adjusted based on activity level, and the dynamic feedback data is input into the feedback database and generated into multiple data clusters through analysis.

[0011] A feature map is generated based on the connection relationships between feature sets, and a scenario optimization plan is generated based on the analysis of data clusters from the feature map.

[0012] In a preferred embodiment, the step of establishing a 3D scene model to obtain a scene domain, and dividing the scene domain into elements based on element partitioning rules to obtain an element set, is as follows:

[0013] A 3D model of the scene is constructed using 3D modeling technology, and the 3D model is mapped to a computable scene domain, including the geometric layer, physical layer, business layer, and rule layer.

[0014] The rules for dividing elements include functional division, physical attributes, interaction levels, and data relationships;

[0015] The scene domain is divided into multiple elements based on the element partitioning rules, and the result is output as a set of elements.

[0016] In a preferred embodiment, the steps of generating dynamic events based on the event generation mechanism and setting a dynamic interface for the feature set to receive dynamic events and provide dynamic feedback data are as follows:

[0017] The event generation mechanism includes the trigger source classification and the corresponding event parameters;

[0018] Trigger sources are categorized into user interaction events, environmental events, time-series events, and logical events;

[0019] The event parameters are metadata including event ID, triggering conditions, priority, and validity period;

[0020] Set up dynamic interfaces for multiple features in a feature set. The dynamic interfaces are used to receive dynamic events. The features process the dynamic events to generate dynamic data, which is then output by the dynamic interfaces.

[0021] In a preferred embodiment, the step of generating an interface heatmap based on dynamic feedback data to reflect the activity level of the dynamic interface is as follows:

[0022] A data access center is constructed, which is connected to multiple elements through dynamic transmission channels;

[0023] The data call center collects call information from dynamic interfaces, including call frequency, data throughput, and response latency;

[0024] Map each dynamic interface in the feature set to coordinates in the scene domain to generate an interface spatial distribution matrix;

[0025] The call information and the interface spatial distribution matrix are combined to generate an interface heatmap;

[0026] Interface heatmaps are used to display the activity level of dynamic interfaces. Interface activity is determined by normalizing and weighting the call information of dynamic interfaces.

[0027] In a preferred embodiment, the step of dynamically adjusting the collection frequency of dynamic feedback data based on activity level is as follows:

[0028] A dynamic valve is set in the dynamic transmission channel, and the control factor is stored in the dynamic valve.

[0029] The control factors include multiple bandwidth expansion factors, multiple equalization factors, and multiple throttling factors;

[0030] Generate an interface activity table based on the activity level of the dynamic interface;

[0031] The release of the control factor in the dynamic valve is adjusted based on the interface activity table;

[0032] A factor-rate mapping model is constructed, and the expected transmission rate is obtained by predicting the nonlinear effect of the control factor on the transmission rate of the dynamic transmission channel through a pre-trained neural network model.

[0033] A feedback compensation mechanism is set up in the data call center. When the deviation between the call information and the expected transmission rate exceeds a preset threshold, the dynamic valve automatically switches to the redundant channel or triggers data retransmission.

[0034] In a preferred embodiment, the step of adjusting the release of the regulating factor in the dynamic valve based on the interface activity table is as follows:

[0035] In the data call center, an interface activity-adjustment factor correlation matrix is ​​constructed based on the interface activity table. The elements in the correlation matrix represent the sensitivity of dynamic interfaces to adjustment factors.

[0036] An activity-factor mapping table is constructed based on sensitivity. The activity-factor mapping table contains a combination of factors pre-assigned to each interface's activity range.

[0037] The data call center sends the activity-factor mapping table to the dynamic valve based on the association matrix, and the dynamic valve executes the factor release strategy based on the activity-factor mapping table;

[0038] The factor release strategy includes a two-level triggering strategy, namely primary triggering and advanced triggering;

[0039] The initial trigger is a dynamic valve that directly queries the activity-factor mapping table to release the pre-allocated factor combination;

[0040] Advanced triggering occurs when a sudden change in traffic occurs in the dynamic transmission channel. Traffic changes include sudden traffic, interface anomalies, and activity changes.

[0041] Advanced triggering employs a dynamic feedback optimization algorithm to adjust the pre-assigned factor combination in real time;

[0042] Simultaneously, cluster analysis algorithms are used to perform cluster analysis on the interface activity table, and similar dynamic interfaces are clustered to obtain similar interface clusters. Dynamic interfaces in similar interface clusters share the factor release strategy.

[0043] In a preferred embodiment, the step of inputting dynamic feedback data into a feedback database and generating multiple data clusters through analysis is as follows:

[0044] The data retrieval center and the feedback database are connected via a data transmission channel;

[0045] Once the feature set completes the dynamic event, it transmits the dynamic feedback data to the data call center. The data call center then transmits the interface activity table and the dynamic feedback data to the feedback database through the data transmission channel.

[0046] The feedback database sets up a sub-database for each feature in the feature set, and sets up a data distribution station in the sub-database;

[0047] The data distribution station is used to encapsulate dynamic feedback data from multiple sub-databases into multiple data clusters, and assign data cluster weights to multiple data clusters based on the interface activity table.

[0048] In a preferred embodiment, the steps of generating a feature map based on the connection relationships between feature sets and generating a scenario optimization plan based on the feature map data clusters are as follows:

[0049] Constructing an element map based on the functional dependencies and physical couplings among multiple elements;

[0050] The feature map includes feature nodes and edges connecting the feature nodes. The edge weights are quantified functional dependencies and physical coupling. Preset functional attributes are set for each node in the feature map.

[0051] The semantic feature engine is used to extract data features from data clusters, and the weights of the data clusters are embedded into the data features to generate weighted data features.

[0052] The semantic annotation engine maps the weighted data features back to the corresponding nodes in the feature map.

[0053] The dynamic spectral clustering algorithm is used to identify anomalous communities in the feature graph, including anomalous nodes and anomalous edges.

[0054] The feature map is reconstructed based on preset functional attributes to make the data features close to the preset functional attributes. The reconstructed feature map is then transformed into a set of reconstructed features as a scene optimization plan.

[0055] This invention also provides an intelligent information acquisition system based on scenario design feedback, comprising:

[0056] Model building module: Builds a 3D model of the scene to obtain the scene domain, and divides the scene domain into elements based on the element division rules to obtain the element set;

[0057] Event Triggering Module: Connects to the model building module, generates dynamic events based on the event generation mechanism, sets up the dynamic interface of the feature set, and is used to receive dynamic events and provide dynamic feedback data.

[0058] Data Feedback Module: Connected to the event triggering module, it generates an interface heatmap based on dynamic feedback data to reflect the activity level of the dynamic interface; it dynamically adjusts the collection frequency of dynamic feedback data based on the activity level, inputs the dynamic feedback data into the feedback database, and generates multiple data clusters through analysis;

[0059] Scene optimization module: Connected to the data feedback module, it generates a feature map based on the connection relationships between feature sets, and generates a scene optimization plan based on the analysis of data clusters from the feature map.

[0060] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0061] 1. This invention achieves real-time and precise control of the data acquisition and transmission process through the synergistic effect of dynamic valve regulation and interface heatmap. Traditional methods usually use fixed frequency acquisition or static threshold triggering mechanisms, which are difficult to cope with sudden high load demand or inefficient transmission problems in complex scenarios. This solution innovatively introduces a dynamic release mechanism of activity factors. Combined with a pre-trained factor-rate mapping model, it can automatically adjust the channel transmission rate according to the spatiotemporal variation characteristics of interface activity (such as call frequency, throughput, response delay). For example, when the heatmap detects a sudden increase in interface activity in a certain area, the dynamic valve will immediately release the bandwidth expansion factor to temporarily increase the priority and resource quota of the channel. At the same time, the feedback compensation mechanism avoids the risk of data packet loss.

[0062] 2. This invention achieves an intelligent leap from raw data to optimization strategies through deep semantic fusion of element maps and data clusters. Traditional optimization methods often rely on manual rules or isolated data analysis, making it difficult to capture the implicit relationships between elements in complex systems. However, the element map constructed by this solution not only includes multi-dimensional connections such as functional dependencies and physical couplings, but also automatically identifies abnormal communities through dynamic spectral clustering algorithms and performs root cause localization by combining the weight features of data clusters. For example, in the operation and maintenance scenario of industrial equipment, the system can discover through map analysis that there is a strong correlation between the vibration data cluster of a certain motor node and the energy consumption data cluster of the lubrication system, thereby generating a composite optimization scheme of "adjusting the oil supply frequency + replacing worn bearings". Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0064] Figure 1 This is a flowchart of the method of the present invention.

[0065] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Example 1, please refer to Figure 1 As shown in this embodiment, an intelligent information collection method based on scenario design feedback includes:

[0068] S1. Establish a 3D model of the scene to obtain the scene domain, and divide the scene domain into elements based on the element division rules to obtain the element set;

[0069] S2. Generate dynamic events based on the event generation mechanism, and set up a dynamic interface for the element set to receive dynamic events and provide dynamic feedback data.

[0070] S3. Generate an interface heatmap based on dynamic feedback data to reflect the activity level of the dynamic interface;

[0071] S4. Based on activity level, dynamically adjust the collection frequency of dynamic feedback data, input the dynamic feedback data into the feedback database, and generate multiple data clusters through analysis;

[0072] S5. Generate an element map based on the connection relationships between element sets, and generate a scenario optimization plan based on the analysis of data clusters from the element map.

[0073] As described in steps S1-S5 above, in the field of intelligent scene information collection, existing information collection and optimization methods mainly rely on static data collection and rule-driven decision-making mechanisms. However, these technologies have significant limitations in dynamic and complex environments. With the development of IoT, digital twins and artificial intelligence technologies, the demand for real-time, adaptive and precise optimization capabilities is increasing, and traditional methods are no longer able to meet the needs of practical applications.

[0074] Traditional methods employ fixed-frequency data acquisition and threshold-triggered mechanisms, such as periodic sensor readings or simple conditional judgments, which are ill-suited to handling sudden load changes in dynamic environments. Furthermore, optimization strategies based on isolated data analysis lack consideration for the complex relationships between scene elements, leading to fragmented decision-making and hindering holistic optimization. Based on the above analysis, the shortcomings of existing technologies include: 1. Rigid resource control, unable to adaptively adjust acquisition frequency and transmission bandwidth according to real-time needs; 2. One-sided optimization analysis, with knowledge graphs disconnected from real-time data, failing to capture deep semantic relationships across multiple data sources; 3. Weak verification process, with direct deployment of optimization solutions risking cascading failures.

[0075] This invention achieves real-time and precise control of the data acquisition and transmission process through the synergistic effect of dynamic valve regulation and interface heatmap. Traditional methods usually use fixed frequency acquisition or static threshold triggering mechanisms, which are difficult to cope with sudden high load demands or inefficient transmission problems in complex scenarios. This solution innovatively introduces a dynamic release mechanism of activity factors. Combined with a pre-trained factor-rate mapping model, it can automatically adjust the channel transmission rate according to the spatiotemporal variation characteristics of interface activity (such as call frequency, throughput, and response latency). For example, when the heatmap detects a sudden increase in interface activity in a certain area, the dynamic valve will immediately release the bandwidth expansion factor to temporarily increase the priority and resource quota of the channel. At the same time, the feedback compensation mechanism avoids the risk of data packet loss.

[0076] Simultaneously, through deep semantic fusion of element maps and data clusters, an intelligent leap from raw data to optimization strategies is achieved. Traditional optimization methods often rely on manual rules or isolated data analysis, making it difficult to capture the implicit relationships between elements in complex systems. However, the element map constructed by this solution not only includes multi-dimensional connections such as functional dependencies and physical couplings, but also automatically identifies abnormal communities through dynamic spectral clustering algorithms and performs root cause localization by combining the weight features of data clusters. For example, in industrial equipment operation and maintenance scenarios, the system can discover through map analysis that there is a strong correlation between the vibration data cluster of a certain motor node and the energy consumption data cluster of the lubrication system, thereby generating a composite optimization solution of "adjusting the oil supply frequency + replacing worn bearings".

[0077] In one embodiment, step S1, which involves establishing a 3D model of the scene to obtain a scene domain and then dividing the scene domain into feature sets based on feature partitioning rules, includes:

[0078] S11. Construct a 3D model of the scene using 3D modeling technology, and map the 3D model into a computable scene domain including the geometric layer, physical layer, business layer and rule layer;

[0079] S12. The rules for dividing elements include functional division, physical attributes, interaction levels, and data correlation.

[0080] S13. Divide the scene domain into multiple elements based on the element partitioning rules and output them as an element set;

[0081] As described in steps S11-S13 above, in the stage of constructing the 3D scene model, the system acquires original scene data through high-precision LiDAR scanning combined with multi-view 4K imagery, and uses an improved ICP point cloud registration algorithm to achieve multi-source data alignment. The constructed computable scene domain adopts a four-layer architecture: the geometry layer uses a hybrid representation method of NURBS surfaces and adaptive voxels, supporting LOD multi-level detail rendering; the physics layer defines the material parameter system based on the MaterialX 2.0 standard, and integrates the BulletPhysics engine to realize rigid / flexible body dynamics simulation; the business layer adopts the BPMN2.0 specification modeling workflow, and each business node is encapsulated as an independent FaaS microservice; the rules layer uses the Drools 8.0 rule engine to realize dynamic management of constraints and supports real-time rule loading. In the element partitioning stage, the system first completes scene semantic segmentation through recurrent neural networks, and then performs intelligent partitioning based on the four-dimensional rule system: the functional dimension uses the spectral clustering algorithm to analyze service call relationships; the physical dimension uses DBSCAN clustering to process material and dynamic properties; the interaction dimension uses the PageRank algorithm to calculate node importance classification; and the data dimension is based on Apache... Kafka constructs a real-time data stream graph; finally, it performs multi-rule fusion through a weighted decision matrix, outputting a set of elements conforming to the JSON-LD specification, with each element containing complete geometric features, physical parameters, business interfaces, and data binding information.

[0082] In one embodiment, step S2, which generates dynamic events based on an event generation mechanism and sets a dynamic interface for the feature set to receive dynamic events and provide dynamic feedback data, includes:

[0083] S21. The event generation mechanism includes trigger source classification and corresponding event parameters;

[0084] S22. Trigger sources are categorized into user interaction events, environmental events, timing events, and logical events.

[0085] S23. Event parameters are metadata including event ID, triggering conditions, priority, and validity period;

[0086] S24. Set up dynamic interfaces for multiple elements in the element set. The dynamic interfaces are used to receive dynamic events. The elements process the dynamic events to generate dynamic data and output it through the dynamic interfaces.

[0087] As described in steps S21-S24 above, regarding the event generation mechanism, the system adopts a multi-source heterogeneous event triggering architecture. User interaction events capture raw input through the human-machine interface device protocol and are converted into standard event streams by the event normalization processor. Environmental events are collected in real time by IoT sensor nodes deployed in the scene, and data cleaning and feature extraction are performed through the edge computing gateway. Time-series events are driven by a high-precision timer and use a time wheel algorithm to achieve millisecond-level event scheduling. Logical events are dynamically generated based on the changes in the state of elements through a rule reasoning engine and can use the Rete algorithm. All events are encapsulated in a unified metadata format containing event ID, triggering conditions, priority, and effective duration. In terms of dynamic interface implementation, an asynchronous non-blocking design is adopted. Each interface contains a triple buffering mechanism: an input buffer layer using a circular buffer with configurable capacity, a processing layer using a sandboxed processing unit based on Wasm, and an output buffer layer using a priority queue with QoS policy. Moreover, the dynamic interface can achieve multi-protocol adaptive switching, which can meet the requirements of receiving different dynamic events. After receiving a dynamic event, the dynamic interface transmits it to the element. After the element completes the response to the dynamic event, it outputs it to the data call center through the dynamic interface.

[0088] In one embodiment, step S3, which generates an interface heatmap based on dynamic feedback data to reflect the activity level of the dynamic interface, includes:

[0089] S31. Construct a data access center, which is connected to multiple elements through a dynamic transmission channel;

[0090] S32. The data call center collects call information from dynamic interfaces, including call frequency, data throughput, and response latency.

[0091] S33. Map each dynamic interface in the element set to the coordinates of the scene domain to generate an interface spatial distribution matrix;

[0092] S34. Combine the call information and the interface spatial distribution matrix to generate an interface heatmap;

[0093] S35. The interface heatmap is used to display the activity level of dynamic interfaces. The interface activity level is obtained by normalizing and weighting the call information of dynamic interfaces.

[0094] As described in steps S31-S5 above, the interface heatmap generation method adopts a distributed data call center architecture to achieve real-time visual monitoring of dynamic interface status. The data call center adopts a microservice cluster design, including three types of specialized nodes: collection nodes, computing nodes, and visualization nodes. The collection nodes poll the call information of each interface at an adjustable period through the transmission channel. The computing nodes execute a three-layer processing pipeline: first, they standardize the raw indicators, including but not limited to call frequency, throughput, and latency, using a sliding window; then, they calculate the weights of each indicator using the entropy weight method; finally, they calculate the activity score according to the formula. The visualization nodes fuse the weighted activity data with the interface spatial distribution matrix to generate a three-dimensional heatmap. The precise calculation process of interface activity includes... The process includes: smoothing the call frequency using the natural logarithm, normalizing the throughput according to channel capacity, and transforming the latency metric using an inverse proportional function. Finally, a weighted summation is used to generate an activity score of 0-100. For example, within a 1-minute window, an interface has a call frequency of 120 times, a throughput of 8 Mbps, and an average latency of 45 ms. The baseline call frequency is 200 times, the maximum throughput is 10 Mbps, and the maximum latency is 50 ms. After normalization, the call frequency score (based on logarithm) is 0.92, the throughput score is 0.8, and the latency score is 0.1 (lower latency results in a lower latency score). With weights set at 40% for call frequency, 35% for throughput, and 25% for latency, the interface activity score is 0.92 × 0.4. + 0.80×0.35+ 0.10×0.25 = 0.673, where the weight adjustment adopts a dynamic weight strategy driven by business needs, including basic weights, which are initialized by domain knowledge and verified for rationality through the AHP (Analytic Hierarchy Process); real-time adjustment based on the coefficient of variation of statistical indicators using a sliding window, automatically increasing the weight of indicators with large fluctuations; and scenario-adaptive pre-set multiple weight templates that are automatically switched based on business type identification.

[0095] In one embodiment, step S4, which dynamically adjusts the collection frequency of dynamic feedback data based on activity level, includes:

[0096] S41. Set a dynamic valve in the dynamic transmission channel, and store the control factor in the dynamic valve;

[0097] S42. The control factors include multiple bandwidth expansion factors, multiple equalization factors, and multiple throttling factors.

[0098] S43. Generate an interface activity table based on the activity of the dynamic interface;

[0099] S44. Adjust the release of the regulating factor in the dynamic valve based on the interface activity table;

[0100] S45. Construct a factor-rate mapping model and use a pre-trained neural network model to predict the nonlinear effect of the control factor on the transmission rate of the dynamic transmission channel to obtain the expected transmission rate.

[0101] S46. Set up a feedback compensation mechanism in the data call center. When the deviation between the call information and the expected transmission rate exceeds the preset threshold, the dynamic valve will automatically switch to the redundant channel or trigger data retransmission.

[0102] As described in steps S41-S46 above, the dynamic valve is an intelligent flow control component deployed on a dynamic transmission channel to adjust the data acquisition frequency and transmission rate in real time. Its core functions include: 1. Multi-level flow control, supporting three working modes: bandwidth expansion, equalization, and throttling, dynamically switching according to interface activity; 2. Adaptive decision-making, automatically selecting the optimal control factor based on the interface activity table and factor-rate mapping model; 3. Redundancy and fault tolerance, with a built-in fault detection mechanism that seamlessly switches to a backup redundant channel when the main channel is abnormal. The hardware architecture of the dynamic valve can use FPGA to achieve high-speed data packet processing and support line-rate forwarding. Its control logic includes input queue management, using dual buffers to hierarchically store data packets, a factor injection engine that writes the control factor into the channel control register via DMA, and a rate monitoring unit that calculates the actual throughput in real time and triggers a compensation mechanism when compared with the expected rate. The adjustment factor is a parameterized instruction used in the dynamic valve to control transmission behavior, divided into three categories: bandwidth expansion factor, equalization factor, and throttling factor. Among them, the bandwidth expansion factor temporarily increases the channel capacity, which is achieved by adjusting the physical layer coding rate, initiating multi-link aggregation to adjust the transmission form of the dynamic transmission channel, and equalization factor. The function of the adjustment factor is to maintain the current rate and optimize resource allocation. It is implemented by using weighted fair queue scheduling of data packets and dynamically adjusting the TCP window size. The function of the throttling factor is to reduce the transmission rate to reduce energy consumption or avoid congestion. It is implemented by forced frequency reduction and data compression. Each adjustment factor is stored in the FPGA on-chip memory in the form of binary instructions and supports hot updates. The standard unit of the bandwidth expansion factor is 2.4 bps / Hz, which represents the spectral efficiency improvement per hertz of one bandwidth expansion factor. The standard unit of the equalization factor is 15Q, which is dimensionless and ranges from 0 to 100. It is used to evaluate the importance of tasks in the dynamic transmission channel. The standard unit of the throttling factor is 1.5:1CR, which means that different compression ratios are applied to the transmitted data to achieve the throttling effect. For example, in the scenario of industrial robotic arm control commands, the stable bandwidth is 240 Mbps. When the activity monitoring detects that the activity of the control command interface jumps to 190%, the dynamic valve immediately releases three bandwidth expansion factors, increasing the bandwidth of the dynamic transmission channel for transmitting control commands by 9.At 6 bps / Hz, two throttling factors are released simultaneously, achieving a data compression ratio of 3:1. Based on the adjustment of the adjustment factors, a new bandwidth of 336 Mbps is obtained. After the dynamic transmission channel stabilizes, the dynamic valve releases three equalization factors to adjust the importance level of each service, ensuring balanced mixed services. New factor strategies are remotely pushed through the data call center. The factor-rate mapping model uses an LSTM neural network trained on multiple sets of "activity + factor combination - actual rate change" data to predict the nonlinear impact of the adjustment factors on the transmission rate of the dynamic transmission channel. The feedback compensation mechanism monitors the call information of the dynamic interface in real time and calculates the deviation between the deviation and the output predicted rate of the LSTM neural network. If the deviation exceeds a pre-designed threshold, different compensation methods are adopted based on the degree of deviation. Short-term compensation involves switching to a redundant channel, where the redundant channel is a pre-configured backup link; long-term compensation involves retraining the neural network model.

[0103] In one embodiment, step S44, which adjusts the release of the control factor in the dynamic valve based on the interface activity table, includes:

[0104] S441. Construct an interface activity-adjustment factor correlation matrix in the data call center based on the interface activity table. The elements in the correlation matrix represent the sensitivity of dynamic interfaces to adjustment factors.

[0105] S442. Construct an activity-factor mapping table based on sensitivity. The activity-factor mapping table contains a combination of factors pre-assigned to each interface's activity range.

[0106] S443. The data call center sends the activity-factor mapping table to the dynamic valve based on the correlation matrix. The dynamic valve executes the factor release strategy based on the activity-factor mapping table.

[0107] S444, the factor release strategy includes a two-level triggering strategy, including primary triggering and advanced triggering;

[0108] S445, the primary trigger is to directly query the activity-factor mapping table for the dynamic valve and release the pre-allocated factor combination;

[0109] S446. When a traffic surge occurs in the dynamic transmission channel, an advanced trigger is activated. Traffic surges include sudden traffic, interface anomalies, and activity surges.

[0110] S447, Advanced Trigger uses a dynamic feedback optimization algorithm to adjust the pre-allocated factor combination in real time;

[0111] S448. At the same time, cluster analysis algorithm is used to perform cluster analysis on the interface activity table, and similar dynamic interfaces are clustered to obtain similar interface clusters. Dynamic interfaces in similar interface clusters share the factor release strategy.

[0112] As described in steps S441-S445 above, the core of the above scheme lies in dynamically adjusting the control factors (bandwidth expansion factor, balancing factor, and throttling factor) in the valve based on the interface activity table to achieve intelligent optimization of data acquisition frequency. The data call center first trains a sensitivity model based on historical data to calculate the response degree of each dynamic interface to different control factors. Specifically, multiple linear regression + gradient descent optimization is used to fit the nonlinear relationship between interface activity (call frequency, throughput, latency) and control factors. The rows of the correlation matrix represent interfaces, the columns represent control factors, and the matrix elements represent the sensitivity interval of the dynamic interface to the factors. Positive values ​​indicate enhancement, and negative values ​​indicate inhibition. Based on sensitivity analysis, the system divides the interface activity into several intervals (such as low, medium, and high) and constructs an activity-factor mapping table for each interval. Based on the activity-factor mapping table, the pre-allocated optimal factor combination is obtained. For example, in the low activity interval: the throttling factor is triggered to release more computing resources; in the medium activity interval: the balancing factor is used to maintain stable transmission. High-activity intervals: Release bandwidth expansion factors to improve data acquisition rates and enhance monitoring capabilities; The data call center sends the optimal factor combination to the corresponding dynamic valve. The dynamic valve executes multi-level triggering based on the activity-factor mapping table. Primary triggering (normal mode): The dynamic valve directly queries the mapping table and releases factors according to a predefined strategy, suitable for steady-state traffic; Advanced triggering (emergency mode): When a traffic mutation is detected (such as sudden traffic or interface anomalies), the system switches to the dynamic feedback optimization algorithm. The dynamic feedback optimization algorithm uses quantized PID control + reinforcement learning model to adjust the factor combination in real time. For example, if an interface experiences a sudden high traffic, the dynamic feedback optimization algorithm will dynamically increase the weight of the throttling factor while reducing the bandwidth expansion factor to smooth the transmission rate. At the same time, to improve computational efficiency, the system uses the K-means++ clustering algorithm to classify similar interfaces into interface clusters based on the interface activity mode (call frequency, data volume, response time); Interfaces within the same cluster share the factor release strategy to reduce redundant calculations. The clustering results are updated regularly to ensure dynamic adaptability.

[0113] In one embodiment, step S4, which involves inputting dynamic feedback data into a feedback database and generating multiple data clusters through analysis, includes:

[0114] S47. The data retrieval center and the feedback database are connected via a data transmission channel;

[0115] S48. After the element set completes the dynamic event, the dynamic feedback data is transmitted to the data call center. The data call center transmits the interface activity table and dynamic feedback data to the feedback database through the data transmission channel.

[0116] S49. The feedback database sets up a sub-database for each feature in the feature set, and sets up a data distribution station in the sub-database.

[0117] S410, the data distribution station is used to encapsulate dynamic feedback data from multiple sub-databases into multiple data clusters, and assign data cluster weights to multiple data clusters based on the interface activity table;

[0118] As described in steps S47-S410 above, the data distribution station is the core processing unit in the feedback database, responsible for the intelligent aggregation and weight allocation of dynamic data. It adopts a microservice-based layered architecture including an access layer, a processing layer, and a storage layer. The access layer includes a protocol adapter supporting multi-protocol access, and a data verification module using digital signatures to verify data. The processing layer includes a streaming engine, a weight calculator, and a semantic analyzer. The streaming engine is based on Apache Flink real-time window calculation, the weight calculator dynamically adjusts the weights of data clusters based on an interface activity table, and the semantic analyzer integrates a BRT model to extract data semantic types. The storage layer adopts a layered storage strategy, using different storage methods for different types of data, including hot data storage, warm data storage, and cold data storage. Hot data storage uses NVMe SSDs, while warm data uses 3D NAND. Flash storage and cold data storage use object storage. The data cluster generation process includes data sharding, feature extraction, weight calculation, and metadata encapsulation. The algorithm for assigning weights to multiple data clusters involves inputting an interface activity table and data freshness. Data freshness represents the collection timestamp. The metrics of each data cluster are standardized to the [0, 1] range through normalization. Dynamic weights are calculated using the entropy weight method. A load balancing strategy is adopted for data clusters with assigned weights. High-weight data clusters are written to high-performance storage nodes first, while low-weight data clusters are batch merged and archived. At the same time, a resource reservation module is set up to reserve a predetermined proportion of dedicated channels for key elements.

[0119] In one embodiment, step S5, which generates a feature map based on the connection relationships between feature sets and generates a scene optimization plan based on the feature map data clusters, includes:

[0120] S51. Construct an element map based on the functional dependencies and physical couplings between multiple elements;

[0121] S52. The feature map includes feature nodes and edges connecting feature nodes. The edge weights are quantified functional dependencies and physical coupling. Set the preset functional attributes of each node in the feature map.

[0122] S53. Use a semantic feature engine to extract data features from data clusters, and embed data cluster weights into data features to generate weighted data features;

[0123] S54. Map the weighted data features back to the corresponding nodes of the feature map using a semantic annotation engine;

[0124] S55. Use dynamic spectral clustering algorithm to identify anomalous communities in feature graphs, including anomalous nodes and anomalous edges.

[0125] S56. Reconstruct the element map based on the preset functional attributes to make the data features close to the preset functional attributes, and transform the reconstructed element map into a reconstructed element set as a scene optimization plan.

[0126] As described in steps S51-S56 above, firstly, an element graph is constructed through multi-dimensional relationship modeling. Nodes represent scene elements (such as devices, sensors, etc.), and edges represent functional dependencies and physical coupling relationships between elements. Edge weights are dynamically calculated based on real-time interaction frequency (functional dependency) and physical constraint strength (such as distance, energy transfer efficiency, etc.). A dynamic graph neural network updates the graph topology every 5 minutes. Feature extraction of data clusters uses a multimodal semantic engine, combining a BERT model to parse unstructured text and a temporal convolutional network to process sensor data streams, generating feature vectors with weighted labels (range 0-1). The weight values ​​are calculated by the data distribution station based on interface activity and semantic urgency. The semantic annotation engine maps data cluster features to element graph nodes using knowledge graph alignment technology. The TransR model is used to calculate the similarity between features and node attributes. For features with a matching degree below a threshold... Nodes with a value <0.6 are marked as objects to be optimized. During the anomaly detection phase, the improved dynamic spectral clustering algorithm (integrating Louvain community detection and temporal anomaly detection) identifies abnormal communities with functional deviations greater than a preset threshold. Simultaneously, it marks three types of problematic edges: overloaded edges (e.g., load > 90% capacity), idle edges (e.g., utilization < 10%), and conflicting edges (e.g., bidirectional traffic asymmetry > 30%). When generating the optimization plan, the system constructs a loss function based on preset functional attributes (e.g., device rated parameters, service SLA, etc.), adjusts the connection weights and resource allocation parameters of abnormal nodes using gradient descent, and finally outputs an optimization plan containing the following: 1. Node-level instructions (e.g., adjusting the sampling frequency of device A to 50Hz); 2. Edge-level optimization (e.g., reconstructing the service call chain E1-E2 into E1-E3-E2 to reduce latency); 3. Global strategies (e.g., triggering the pre-start of the cooling system).

[0127] Example 2, please refer to Figure 2 As shown in this embodiment, an intelligent information acquisition system based on scenario design feedback includes:

[0128] Model building module: Builds a 3D model of the scene to obtain the scene domain, and divides the scene domain into elements based on the element division rules to obtain the element set;

[0129] Event Triggering Module: Connects to the model building module, generates dynamic events based on the event generation mechanism, sets up the dynamic interface of the feature set, and is used to receive dynamic events and provide dynamic feedback data.

[0130] Data Feedback Module: Connected to the event triggering module, it generates an interface heatmap based on dynamic feedback data to reflect the activity level of the dynamic interface; it dynamically adjusts the collection frequency of dynamic feedback data based on the activity level, inputs the dynamic feedback data into the feedback database, and generates multiple data clusters through analysis;

[0131] Scene optimization module: Connected to the data feedback module, it generates a feature map based on the connection relationships between feature sets, and generates a scene optimization plan based on the analysis of data clusters from the feature map.

[0132] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent information collection based on scenario-based design feedback, characterized in that, A 3D model of the scene is established to obtain the scene domain. Based on the element partitioning rules, the scene domain is partitioned into elements to obtain the element set. Dynamic events are generated based on the event generation mechanism, and a dynamic interface for the feature set is set to receive dynamic events and provide dynamic feedback data. An interface heatmap is generated based on dynamic feedback data to reflect the activity level of the dynamic interface. The collection frequency of dynamic feedback data is dynamically adjusted based on activity level, and the dynamic feedback data is input into the feedback database and generated into multiple data clusters through analysis. A dynamic valve is set in the dynamic transmission channel, and the control factor is stored in the dynamic valve. The control factors include multiple bandwidth expansion factors, multiple equalization factors, and multiple throttling factors; Generate an interface activity table based on the activity level of the dynamic interface; The release of the control factor in the dynamic valve is adjusted based on the interface activity table; A factor-rate mapping model is constructed, and the expected transmission rate is obtained by predicting the nonlinear effect of the control factor on the transmission rate of the dynamic transmission channel through a pre-trained neural network model. A feedback compensation mechanism is set up in the data call center. When the deviation between the call information and the expected transmission rate exceeds a preset threshold, the dynamic valve automatically switches to the redundant channel or triggers data retransmission. A feature map is generated based on the connection relationships between feature sets, and a scenario optimization plan is generated based on the analysis of data clusters from the feature map.

2. The intelligent information acquisition method based on scenario design feedback according to claim 1, characterized in that, The steps of establishing a 3D model of the scene to obtain the scene domain, and dividing the scene domain into element sets based on element partitioning rules are as follows: A 3D model of the scene is constructed using 3D modeling technology, and the 3D model is mapped to a computable scene domain, including the geometric layer, physical layer, business layer, and rule layer. The rules for dividing elements include functional division, physical attributes, interaction levels, and data relationships; The scene domain is divided into multiple elements based on the element partitioning rules, and the result is output as a set of elements.

3. The intelligent information acquisition method based on scenario design feedback according to claim 1, characterized in that, The steps for generating dynamic events based on the event generation mechanism and setting up a dynamic interface for the feature set to receive dynamic events and provide dynamic feedback data are as follows: The event generation mechanism includes the trigger source classification and the corresponding event parameters; Trigger sources are categorized into user interaction events, environmental events, time-series events, and logical events; The event parameters are metadata including event ID, triggering conditions, priority, and validity period; Set up dynamic interfaces for multiple features in a feature set. The dynamic interfaces are used to receive dynamic events. The features process the dynamic events to generate dynamic data, which is then output by the dynamic interfaces.

4. The intelligent information acquisition method based on scenario design feedback according to claim 1, characterized in that, The step of generating an interface heatmap based on dynamic feedback data to reflect the activity level of the dynamic interface is as follows: A data access center is constructed, which is connected to multiple elements through dynamic transmission channels; The data call center collects call information from dynamic interfaces, including call frequency, data throughput, and response latency; Map each dynamic interface in the feature set to coordinates in the scene domain to generate an interface spatial distribution matrix; The call information and the interface spatial distribution matrix are combined to generate an interface heatmap; Interface heatmaps are used to display the activity level of dynamic interfaces. Interface activity is determined by normalizing and weighting the call information of dynamic interfaces.

5. The intelligent information acquisition method based on scenario design feedback according to claim 4, characterized in that, The steps for releasing the regulating factor in the dynamic valve based on the interface activity table are as follows: In the data call center, an interface activity-adjustment factor correlation matrix is ​​constructed based on the interface activity table. The elements in the correlation matrix represent the sensitivity of dynamic interfaces to adjustment factors. An activity-factor mapping table is constructed based on sensitivity. The activity-factor mapping table contains a combination of factors pre-assigned to each interface's activity range. The data call center sends the activity-factor mapping table to the dynamic valve based on the association matrix, and the dynamic valve executes the factor release strategy based on the activity-factor mapping table; The factor release strategy includes a two-level triggering strategy, namely primary triggering and advanced triggering; The initial trigger is a dynamic valve that directly queries the activity-factor mapping table to release the pre-allocated factor combination; Advanced triggering occurs when a sudden change in traffic occurs in the dynamic transmission channel. Traffic changes include sudden traffic, interface anomalies, and activity changes. Advanced triggering employs a dynamic feedback optimization algorithm to adjust the pre-assigned factor combination in real time; Simultaneously, cluster analysis algorithms are used to perform cluster analysis on the interface activity table, and similar dynamic interfaces are clustered to obtain similar interface clusters. Dynamic interfaces in similar interface clusters share the factor release strategy.

6. The intelligent information acquisition method based on scenario design feedback according to claim 5, characterized in that, The step of inputting dynamic feedback data into the feedback database and generating multiple data clusters through analysis is as follows: The data retrieval center and the feedback database are connected via a data transmission channel; Once the feature set completes the dynamic event, it transmits the dynamic feedback data to the data call center. The data call center then transmits the interface activity table and the dynamic feedback data to the feedback database through the data transmission channel. The feedback database sets up a sub-database for each feature in the feature set, and sets up a data distribution station in the sub-database; The data distribution station is used to encapsulate dynamic feedback data from multiple sub-databases into multiple data clusters, and assign data cluster weights to multiple data clusters based on the interface activity table.

7. The intelligent information acquisition method based on scenario design feedback according to claim 5, characterized in that, The steps for generating a feature map based on the connection relationships between feature sets, and generating a scenario optimization plan based on the feature map analysis of data clusters are as follows: Constructing an element map based on the functional dependencies and physical couplings among multiple elements; The feature map includes feature nodes and edges connecting the feature nodes. The edge weights are quantified functional dependencies and physical coupling. Preset functional attributes are set for each node in the feature map. The semantic feature engine is used to extract data features from data clusters, and the weights of the data clusters are embedded into the data features to generate weighted data features. The semantic annotation engine maps the weighted data features back to the corresponding nodes in the feature map. The dynamic spectral clustering algorithm is used to identify anomalous communities in the feature graph, including anomalous nodes and anomalous edges. The feature map is reconstructed based on preset functional attributes to make the data features close to the preset functional attributes. The reconstructed feature map is then transformed into a set of reconstructed features as a scene optimization plan.

8. An intelligent information acquisition system based on scenario design feedback, used to implement the intelligent information acquisition method based on scenario design feedback as described in any one of claims 1-7, characterized in that: Model building module: Builds a 3D model of the scene to obtain the scene domain, and divides the scene domain into elements based on the element division rules to obtain the element set; Event Triggering Module: Connects to the model building module, generates dynamic events based on the event generation mechanism, sets up the dynamic interface of the feature set, and is used to receive dynamic events and provide dynamic feedback data. Data feedback module: Connected to the event triggering module, it generates an interface heatmap based on dynamic feedback data to reflect the activity level of the dynamic interface; The collection frequency of dynamic feedback data is dynamically adjusted based on activity level, and the dynamic feedback data is input into the feedback database and generated into multiple data clusters through analysis. Scene optimization module: Connected to the data feedback module, it generates a feature map based on the connection relationships between feature sets, and generates a scene optimization plan based on the analysis of data clusters from the feature map.

Images