Data fusion intelligent equipment linkage management and control system based on large model

Abstract

The invention relates to the technical field of data processing, in particular to a data fusion intelligent equipment linkage management and control system based on a large model, and the system comprises a data collection module which outputs standardized data carrying privacy grading identifiers; the data processing module outputs an optimization algorithm identifier and a collaborative strategy identifier which are synchronously bound; the privacy algorithm fusion scheduling module generates atomization calculation graph data and strategy audit logs fusing privacy operation; the multiple optimization resource module executes calculation graph data and drives a parallel simulation environment rehearsal resource scheme, and the double-closed-loop feedback module is activated when an arbiter continuously detects that deviation exceeds the standard for three times; the safety instruction generation module fuses the output result and the audit log to generate a digital signature equipment control instruction; and the double-closed-loop feedback module updates knowledge base weight parameters through a data feedback loop, and controls the feedback loop to adjust a resource allocation threshold in a grading manner according to simulation confidence, so that dynamic balance between resource efficiency and privacy intensity is realized.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data fusion intelligent device linkage management and control system based on a large model. Background Art

[0002] The intelligent device linkage control system is a comprehensive management platform based on Internet of Things technology and artificial intelligence algorithms, designed to achieve coordinated control and optimized operation among multiple intelligent devices. The system collects environmental data from various sensors in real time through the central processing unit, combines preset logical rules or machine learning models for dynamic analysis, and outputs precise control commands, enabling lighting, security, temperature control and other equipment to cooperate with each other in response to changes in external conditions. The automation process significantly reduces manpower requirements and improves safety protection levels and energy utilization efficiency in smart home or industrial scenarios.

[0003] In the cross-scenario data fusion, the intelligent device linkage management and control system causes technical problems in the algorithm adaptive scheduling and privacy security collaborative optimization, resulting in the inefficient integration of the algorithm dynamic scheduling process and the privacy protection mechanism, thereby causing conflicts in resource allocation and data processing; specifically, in the smart community scenario, the system needs to schedule the face recognition algorithm in real time based on environmental factors to optimize access control, but the privacy security layer is not dynamically coupled with the scheduling decision. For example, when integrating access control and video stream data, algorithm adjustments require recalculation of feature vectors, while privacy encryption calculations are performed independently, which increases computing overhead and data transmission exposure risks, causes response delays and privacy leakage risks, and limits the system's robust deployment under changing conditions. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a data fusion intelligent device linkage management and control system based on a large model to solve the technical problems of algorithm adaptive scheduling and privacy security collaborative optimization in cross-scenario data fusion.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: The present invention provides a data fusion intelligent device linkage management and control system based on a large model, comprising: The data collection module collects the original data stream of the terminal device and outputs standardized data with privacy classification identifiers; The data processing module receives the environmental parameters in the standardized data, queries the dynamic knowledge base and outputs the synchronously bound optimization algorithm identifier and collaborative strategy identifier; The privacy algorithm fusion scheduling module parses the synchronously bound optimization algorithm identifier and collaborative strategy identifier, compiles the privacy operations corresponding to the collaborative strategy identifier into subgraph nodes of the pre-trained algorithm calculation graph, generates atomic calculation graph data of the fused privacy operations, and outputs the atomic calculation graph data and associated policy audit logs. Multiple resource optimization modules allocate computing resources to execute atomic computational graph data while driving a parallel simulation environment to preview resource plans. When the arbitrator detects that the deviation between simulation indicators and actual execution indicators exceeds a dynamic threshold three times in a row, the dual closed-loop feedback module is activated. The security instruction generation module receives the output results of the atomic computation graph data and the policy audit log output by the privacy algorithm fusion scheduling module, appends the policy audit log to the output results to generate data to be signed, and executes the digital signature on the signed data to generate device control instructions; The dual closed-loop feedback module receives the delay indicators generated by the device executing the control instructions, updates the dynamic knowledge base weight parameters through the data feedback loop, and adjusts the resource allocation threshold according to the simulation confidence level through the control feedback loop; The data acquisition module outputs standardized data to the data processing module, the data processing module outputs a binding identifier to the privacy algorithm fusion scheduling module, the privacy algorithm fusion scheduling module outputs atomic computation graph data and policy audit logs to the multiple optimization resource module and the security instruction generation module, the multiple optimization resource module triggers the arbitrator threshold judgment and activates the dual closed-loop feedback module, the dual closed-loop feedback module outputs the updated dynamic knowledge base weight parameters to the data processing module, outputs the resource allocation threshold to the multiple optimization resource module, the security instruction generation module outputs the device control instructions to the terminal device, and feeds back the execution delay indicator to the dual closed-loop feedback module.

[0006] Furthermore, in the data fusion intelligent device linkage management and control system based on a large model described in the present invention, the data acquisition module includes a protocol adaptation layer and an identifier injection unit; The protocol adaptation layer converts the terminal device communication protocol into a unified data format and outputs pre-processed data carrying the device type code; The identifier injection unit receives the pre-processed data, indexes the privacy classification rule base based on the device type encoding, and generates a matching privacy classification identifier to be embedded in the data header; The privacy classification rule base stores the mapping between device types and privacy levels and dynamically responds to indexing requests; The identifier injection unit outputs the standardized data integrating the privacy graded identifier to the data processing module; The protocol adaptation layer processes the original data stream to generate pre-processed data and outputs it to the identifier injection unit; The identifier injection unit extracts the device type code from the pre-processed data and calls the privacy classification rule library; The identifier injection unit embeds the generated privacy-graded identifier into the preprocessed data, generates standardized data and outputs it to the data processing module.

[0007] Furthermore, the data fusion intelligent device linkage management and control system based on a large model of the present invention further includes: The data processing module is equipped with a hierarchical feature extraction engine; The first processing layer performs lighting environment analysis: it inputs a video frame in normalized data, extracts the brightness histogram, calculates the 95th percentile brightness value, and outputs a brightness scalar; The second processing layer performs dynamic target density analysis: it inputs normalized data, applies background modeling and morphological filtering, counts the number of moving objects per square meter, and outputs a density scalar. The feature fusion unit receives the brightness scalar and the density scalar, queries the dynamic weighting coefficient according to the privacy classification identifier, and calculates the composite feature vector; The dynamic knowledge base matches the composite feature vector with the pre-stored scenario strategy mapping relationship, and outputs the bound optimization algorithm identifier and collaborative strategy identifier to the privacy algorithm fusion scheduling module; Standardized data is input into the first processing layer and the second processing layer; The first processing layer outputs the brightness scalar to the feature fusion unit; The second processing layer outputs the density scalar to the feature fusion unit; The feature fusion unit receives the privacy classification identifier, dynamically calculates the weighted value, and fuses the brightness scalar and the density scalar to generate a composite feature vector.

[0008] Furthermore, the data fusion intelligent device linkage management and control system based on a large model described in the present invention includes a dynamic knowledge base including a strategy index layer, a weight storage layer and a strategy update unit; The strategy index layer stores the mapping relationship between the composite feature vector and the bound optimization algorithm identifier and the collaborative strategy identifier; The weight storage layer records the historical performance indicators of each feature vector in the mapping relationship, including response delay and recognition accuracy; The strategy update unit receives the execution delay index transmitted by the data feedback loop and converts the delay index into a weight correction coefficient; The policy update unit receives the confidence level signal transmitted by the control feedback loop and dynamically adjusts the update frequency according to three levels: high / medium / low; The strategy update unit integrates the weight correction coefficient and the update frequency parameter, and outputs the update instruction to the strategy index layer to adjust the mapping relationship; The policy update unit synchronously outputs update instructions to the weight storage layer to calibrate historical performance indicators. The weight storage layer outputs the latest historical performance indicators to the policy index layer to optimize real-time matching accuracy. The data feedback loop transmits the execution delay indicator input to the policy update unit. The control feedback loop transmits the confidence grading signal input to the policy update unit. The policy update unit outputs update instructions to the policy index layer. The policy update unit outputs update instructions to the weight storage layer. The weight storage layer transmits the latest historical performance indicators to the policy index layer. The policy index layer outputs the optimized bound optimization algorithm identifier and collaborative strategy identifier to the privacy algorithm fusion scheduling module.

[0009] Furthermore, in the data fusion intelligent device linkage management and control system based on a large model of the present invention, the privacy algorithm fusion scheduling module is also used to: Receive the synchronization binding optimization algorithm identifier and the collaboration strategy identifier output by the data processing module; Parse the optimization algorithm identifier to locate the output node position of the feature extraction layer in the pre-trained model calculation graph; The privacy rule library corresponding to the index collaborative strategy identifier obtains the noise amplitude parameter, inserts the noise injection node configured with the noise amplitude parameter after the located node position, associates the collaborative strategy identifier to dynamically link the privacy encryption dynamic library to the noise injection node, compiles the computational graph embedded with the noise node and the encryption library to generate atomic computational graph data, and simultaneously generates a policy audit log that records the node positioning, noise parameter configuration, and encryption library link operation details, outputs the atomic computational graph data to the multiple optimization resource module, and outputs the policy audit log to the security instruction generation module.

[0010] Furthermore, the data fusion intelligent device linkage management and control system based on a large model of the present invention further includes: Atomized computation graph data encapsulation in independent containers; The container integrates a policy consistency engine to perform the following collaborative operations: Load the algorithm executable file that integrates privacy operations; Dynamically parse the configuration file corresponding to the collaborative strategy identifier and generate the noise amplitude and encryption library link parameters in real time; Synchronously start the logging service to continuously capture the running status of the noise injection node and the encryption library; Automatically verify the matching between the configuration file digital signature and the registration signature of the collaborative policy identifier when the container is started; Synchronize the collaborative strategy identifier changes to the noise injection node and encryption library link in real time during runtime.

[0011] Furthermore, the data fusion intelligent device linkage management and control system based on a large model of the present invention and the container internal deployment policy driving engine are also used to: Load the algorithm executable file that integrates the noise node and the encryption library, parse the configuration file corresponding to the collaborative strategy identifier, and generate the noise amplitude parameters and encryption library link instructions; Start the logging service to capture the noise node output data and the cryptographic library operation status; During the container initialization phase, the hash value deviation between the configuration file digital signature and the registration signature of the collaborative policy identifier is verified. If the deviation is greater than 5%, a circuit breaker alarm is triggered. During the container operation phase, monitor the collaborative policy identifier change event and reconfigure the noise amplitude parameters and the encryption library link within ≤50 milliseconds; The integrity of the configuration file is periodically verified. When tampering is detected, the container is frozen and a backup instance is started. The logging service outputs noise indicators and encrypted audit records to the security instruction generation module in real time.

[0012] Furthermore, the data fusion intelligent device linkage management and control system based on a large model of the present invention further includes: The arbitrator calculates the Euclidean distance between the simulation indicator and the actual execution indicator. When the distance exceeds the dynamic threshold three times in a row, the dual closed-loop feedback module is activated and a fuse trigger signal is sent to the container. After receiving the circuit breaker signal, the container freezes the currently running instance, starts a clean backup container to load the atomic computation graph data, records the container switching delay time, and feeds the switching delay indicator back to the dual closed-loop feedback module. The control feedback loop adjusts the dynamic threshold based on the simulation confidence level signal. The threshold for high-confidence scenarios is set to 80% of the dynamic threshold baseline value, and the threshold for low-confidence scenarios is set to 120% of the dynamic threshold baseline value. The dual closed-loop feedback module calibrates the generation logic of the confidence grading signal based on the container switching delay index. When the switching delay is greater than 50 milliseconds, the simulation confidence grading signal is automatically downgraded by one level.

[0013] Furthermore, in the data fusion intelligent device linkage management and control system based on a large model described in the present invention, the security instruction generation module includes a reverse verification unit and an audit encapsulation unit; The reverse verification unit receives the feature vector output by the atomic computation graph data, performs feature reconstruction to generate reconstructed data, and calculates the deviation between the feature vector and the reconstructed data; When the deviation value is less than 5%, the irreversibility verification pass signal is output; When the deviation value is ≥5%, a fuse trigger signal is output to the container instance of claim 7; The audit encapsulation unit receives the irreversibility verification pass signal output by the reverse verification unit, receives the policy audit log output by the privacy algorithm fusion scheduling module, binds the irreversibility verification pass signal with the policy audit log to generate a digitally signed audit package, encapsulates the device control instruction and the digitally signed audit package to generate the final control instruction, and the final control instruction carries the irreversibility verification pass signal and the operation audit traceability chain.

[0014] Furthermore, the data fusion intelligent device linkage management and control system based on a large model of the present invention further includes: The data feedback loop receives the delay indicator generated by the device executing the control instruction, converts the delay indicator into the dynamic knowledge base weight correction parameter, and transmits the weight correction parameter to the policy update unit of the dynamic knowledge base and the container policy driving engine; A control feedback loop receives the confidence level signal output by the parallel simulation environment, dynamically adjusts the arbitrator sensitivity parameters according to the three levels of high confidence, medium confidence, and low confidence, and transmits the sensitivity parameters to the multiple optimization resource modules and the arbitrator; In high-confidence scenarios, the weight correction parameter is tightened to the current value of the dynamic knowledge base weight parameter × 80%, and in low-confidence scenarios, the sensitivity parameter is relaxed to the current value of the arbitrator sensitivity parameter × 120%; The policy update unit updates the knowledge base mapping relationship based on the weight correction parameters. The container policy driving engine calibrates the noise injection amplitude based on the weight correction parameters. The multiple optimization resource module adjusts the resource allocation strategy based on the sensitivity parameters. The arbitrator dynamically adjusts the deviation detection response speed based on the sensitivity parameters.

[0015] Beneficial effects of the present invention: The present invention uses a binding identifier transmission mechanism to force algorithm scheduling instructions to form an atomic association with privacy policy parameters, eliminating the timing misalignment problem caused by independent decision-making in traditional solutions; the atomic computation graph compilation technology deeply embeds privacy operation nodes into the pre-trained model computation graph structure to generate an indivisible execution unit, avoiding resource competition and additional overhead caused by independent scheduling of privacy components; the dual closed-loop feedback mechanism quantifies the actual execution effect through the data feedback loop to optimize the knowledge base policy weight, and the control feedback loop dynamically adjusts the resource allocation threshold according to the simulation confidence grading signal to achieve an adaptive balance between resource allocation efficiency and privacy protection strength; the audit encapsulation unit binds the irreversibility proof signal generated by reverse verification with the full-link operation log to generate a digitally signed audit package, so that the device control instructions carry a verifiable traceability chain output, meeting the data desensitization irreversibility and legal compliance requirements, and ultimately solving the core technical contradiction of the difficulty in efficient coordination between algorithm dynamic scheduling and privacy security mechanisms in cross-scenario data fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0017] Figure 1 This is a system architecture diagram of a data fusion intelligent device linkage management and control system based on a large model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0019] See also Figure 1 The present invention provides a data fusion intelligent device linkage management and control system based on a large model, comprising: The data collection module collects the original data stream of the terminal device and outputs standardized data with privacy classification identifiers; The data acquisition module includes a protocol adaptation layer and an identifier injection unit. The protocol adaptation layer deploys a multi-protocol parsing engine. For camera terminals, the ONVIF protocol driver is loaded to convert video streams into structured frame sequences. For access controllers, a Modbus-TCP parser is enabled to generate status event records. For environmental sensors, an MQTT converter is activated to output time-series data packets. All converted data is encapsulated as pre-processed data in a unified JSON format, with the device type encoding metadata field embedded in the data header.

[0020] The identifier injection unit integrates a metadata processor and a rule indexer. The metadata processor extracts the device type code from the preprocessed data header, and the rule indexer uses the code as a key to query the privacy classification rule base. The privacy classification rule base uses an in-memory database to store the mapping between device types and PII levels and returns matching privacy classification identifiers through a hash indexing mechanism. The metadata expander writes the identifier into a new field in the data header to generate standardized data.

[0021] The privacy grading rule base utilizes a distributed key-value storage architecture. It allocates storage nodes using a consistent hashing algorithm and dynamically maintains a mapping table between device type codes and privacy levels. Index requests trigger real-time node location logic, returning the corresponding privacy identifier within milliseconds. The mapping table supports a hot update mechanism to accommodate policy expansion requirements for newly added device types.

[0022] The data processing pipeline forms a logically closed loop. Raw data flows through the protocol identifier to classify the device type, then calls the corresponding protocol parser to convert the data format. The device type code is added to the output port to generate preprocessed data. The preprocessed data is transmitted to the identifier injection unit via a zero-copy channel. The metadata processor extracts the device code to trigger a rule base query, and the metadata expander embeds the privacy identifier to generate standardized data. Finally, structured data integrating device attributes and privacy levels is output to downstream modules.

[0023] The data processing module receives the environmental parameters in the standardized data, queries the dynamic knowledge base and outputs the synchronously bound optimization algorithm identifier and collaborative strategy identifier; The data processing module comprises a hierarchical feature extraction engine and a dynamic knowledge base. The hierarchical feature extraction engine deploys parallel processing channels: the first processing layer performs brightness histogram analysis on the video frame sequence in the standardized data, extracts high-percentile brightness values, and outputs a scalar of light intensity. The second processing layer uses a Gaussian mixture model to model the background, combined with a morphological filtering algorithm to count the number of moving objects per unit area and output a density scalar. This dual-channel processing enables independent quantification of environmental parameters.

[0024] The feature fusion unit integrates a weighted calculation engine. It receives brightness and density scalars and queries a dynamic weighting coefficient table based on the privacy classification identifier in the standardized data header. The weighting coefficient table stores weight configuration rules corresponding to different privacy levels, such as increasing the density scalar weight for high-privacy scenarios. The weighted calculation engine executes a linear fusion algorithm to generate a composite feature vector that comprehensively represents the correlation between light intensity and target density.

[0025] The dynamic knowledge base builds a strategic mapping and matching mechanism. The dynamic knowledge base's strategic index layer stores mappings between composite feature vectors and binding identifier pairs. Binding identifier pairs are fixedly composed of an optimization algorithm identifier and a collaborative strategy identifier. The matching engine uses a k-nearest neighbor algorithm to calculate the Euclidean distance between the input feature vector and a pre-stored vector, returning the binding identifier pair that corresponds to the closest mapping record. This mapping relationship optimizes matching priorities based on the weighting of historical performance indicators.

[0026] The policy update unit responds to dual feedback signals. It receives the execution delay indicator from the data feedback loop and generates weight correction coefficients using a linear transformation model. It also simultaneously analyzes the confidence rating signal from the control feedback loop and dynamically adjusts the update frequency parameter. The fusion engine outputs update instructions to the policy index layer, triggering adjustments to the feature vector matching range and identifier priority reordering.

[0027] The data processing process forms a closed technical loop, with standardized data diverted to a dual processing layer to generate environmental feature scalars. The feature fusion unit dynamically weights and outputs composite feature vectors based on privacy identifiers. The dynamic knowledge base returns binding identifier pairs through similarity matching. The entire process transforms environmental parameter analysis into policy decision-making, and the binding output mechanism ensures atomic synchronization of algorithm scheduling instructions and privacy control parameters.

[0028] The privacy algorithm fusion scheduling module parses the synchronously bound optimization algorithm identifier and collaborative strategy identifier, compiles the privacy operations corresponding to the collaborative strategy identifier into subgraph nodes of the pre-trained algorithm calculation graph, generates atomic calculation graph data of the fused privacy operations, and outputs the atomic calculation graph data and associated policy audit logs. The privacy algorithm fusion scheduling module includes an identifier resolution unit, a node injection engine, and a compilation engine. The identifier resolution unit receives the synchronously bound optimization algorithm identifier and collaboration strategy identifier, queries the computational graph structure corresponding to the optimization algorithm identifier through the pre-trained model registry, and locates the coordinates of the feature extraction layer output nodes. The collaboration strategy identifier indexes the privacy rule base to obtain the noise amplitude parameter set in real time. The rule base stores the mapping between strategy identifiers and perturbation intensities.

[0029] The node injection engine performs a computational graph rewrite. After the node coordinates are output from the localized feature layer, the node injection engine inserts a noise injection node and loads the noise amplitude parameters to configure the node properties. The dynamic linker associates the collaborative strategy identifier with the privacy encryption dynamic library and establishes a runtime call path from the encryption library to the noise injection node. This operation preserves the original computational graph logical structure and only expands the privacy operation nodes.

[0030] The compilation engine embeds noise injection nodes and associated cryptographic library call paths into the algorithm computation graph, generating a computational subgraph that incorporates privacy-enhancing operations. The compilation outputs indivisible, atomic computational graph data, eliminating the resource overhead of independently scheduling privacy modules. Simultaneously, a policy audit log is created, containing a structured record of node location coordinates, noise parameter values, and cryptographic library link addresses. The log includes timestamps and operation hashes.

[0031] Atomized computation graph data is transmitted to multiple optimization resource modules for execution, and policy audit logs are output to the secure instruction generation module. The computation graph data contains complete algorithm instructions and a chain of privacy-preserving operation nodes, and the audit logs provide a foundation for traceability of the operational process. The entire process achieves a deep integration of algorithmic logic and privacy technologies. Binding identifier inputs ensures scheduling consistency, and atomic compilation eliminates execution timing conflicts.

[0032] Multiple resource optimization modules allocate computing resources to execute atomic computational graph data while driving a parallel simulation environment to preview resource plans. When the arbitrator detects that the deviation between simulation indicators and actual execution indicators exceeds a dynamic threshold three times in a row, the dual closed-loop feedback module is activated. The multi-optimization resource module includes a resource allocation unit, a parallel simulation unit, and an arbitration analysis unit. The resource allocation unit deploys a multi-threaded scheduling architecture, allocating GPU / CPU resources to execute atomic computational graph data based on the real-time load status of the computing resource pool. The resource allocation strategy uses a minimum wait time priority algorithm to dynamically adjust the task queues of computing nodes. The parallel simulation unit replicates real-world input data to an isolated environment. The solution generator uses a genetic algorithm to iteratively generate multiple resource allocation solutions. A time series prediction model evaluates the execution efficiency of each solution and outputs the optimal solution number.

[0033] The arbitration analysis unit implements a dual-channel monitoring mechanism. A metric collector is deployed within the arbitration analysis unit to simultaneously capture simulation metrics output by the parallel simulation environment and operational metrics from the actual execution pipeline. The distance calculation engine uses the Euclidean distance algorithm to quantify the difference between the dual-channel metrics. A sliding window mechanism maintains the calculation results for three consecutive detection cycles. A dynamic threshold comparator compares the distance value with a dynamically updated threshold. If the threshold is exceeded three times in a row, a circuit breaker trigger is generated.

[0034] The circuit breaker trigger activates the dual closed-loop feedback module, synchronously sending a circuit breaker signal to the container management unit. The arbitration analysis unit outputs a dynamic threshold adjustment request to the control feedback loop, triggering a reconfiguration of the resource allocation policy. This entire process ensures real-time verification of resource allocation and actual execution status. A rehearsal mechanism enhances decision reliability, and deviation detection ensures the system's ability to respond to exceptions.

[0035] The security instruction generation module receives the output results of the atomic computation graph data and the policy audit log output by the privacy algorithm fusion scheduling module, appends the policy audit log to the output results to generate data to be signed, and executes the digital signature on the signed data to generate device control instructions; The security instruction generation module includes a reverse verification unit and an audit encapsulation unit. The reverse verification unit receives the feature vectors output by the atomized computation graph data and deploys a lightweight variational autoencoder model to attempt feature reconstruction. The reconstruction model generates reconstructed samples that mimic the original data structure through a decoding layer. A deviation analyzer calculates the statistical distribution difference between the feature vectors and the reconstructed data to quantify the degree of data irreversibility. Based on a preset critical threshold, the module outputs either a pass signal for irreversibility verification or a fuse trigger signal.

[0036] The audit encapsulation unit integrates a signal binding engine and a digital signature engine. The signal binding engine receives the verification signal and policy audit log output by the reverse verification unit and uses a timestamp alignment algorithm to bind the irreversibility verification signal with the node operation record to generate a structured data packet. The digital signature engine uses an asymmetric encryption algorithm to sign the data packet, generating an audit packet with a digital fingerprint. The instruction encapsulator uses the device control instruction as the primary data body and appends the digitally signed audit packet to generate the final control instruction with a layered encapsulation structure.

[0037] The reverse verification unit provides mathematical verification of the strength of privacy protection, and the audit encapsulation unit builds a verifiable operation traceability chain. The audit attachment area of the final control instruction carries a digitally signed audit package, including irreversibility verification proof and full-link operation node records.

[0038] The dual closed-loop feedback module receives the delay indicators generated by the device executing the control instructions, updates the dynamic knowledge base weight parameters through the data feedback loop, and adjusts the resource allocation threshold according to the simulation confidence level through the control feedback loop; The dual closed-loop feedback module consists of a data feedback unit and a control feedback unit. The data feedback unit receives the latency metrics generated by the device executing control commands and deploys a metric converter to linearly map the latency data into dynamic knowledge base weight correction parameters. The weight correction parameters represent the effectiveness of policy execution and are transmitted via a message queue to the dynamic knowledge base's policy update unit and the container policy driver engine. Timestamps are added to the parameter transmission process to ensure data temporal integrity.

[0039] The control feedback unit receives confidence-level signals from the parallel simulation environment. The signal classifier classifies the confidence levels into high, medium, and low based on historical accuracy data. The sensitivity adjuster dynamically calculates the arbitrator's sensitivity parameters: high-confidence scenarios increase detection sensitivity, while low-confidence scenarios enhance fault tolerance. These sensitivity parameters are synchronously transmitted to the multiple optimization resource modules and the arbitrator via control channels.

[0040] A parameter coupling mechanism enables dynamic adjustment. The weight correction parameters output by the data feedback unit work in tandem with the sensitivity parameters output by the control feedback unit. High-confidence scenarios tighten the weight correction parameters to a baseline ratio, while low-confidence scenarios relax the sensitivity parameters to a baseline ratio. This parameter coupling mechanism balances the system's dynamic needs between resource efficiency and privacy.

[0041] The execution unit responds to the logic, and the policy update unit re-prioritizes the mapping relationships in the dynamic knowledge base based on weight correction parameters. The container policy driver engine synchronously calibrates the disturbance intensity of noise injection nodes. The multi-optimization resource module optimizes the computing resource allocation strategy based on sensitivity parameters. The arbitrator dynamically adjusts the deviation detection response frequency. The execution unit response implements closed-loop control from the decision-making layer to the operation layer.

[0042] Data feedback quantifies actual execution results, while control feedback predicts system reliability. This dual-loop collaboratively drives strategic decision optimization and operational parameter adjustment, enabling resource allocation efficiency and privacy protection to continuously adapt to dynamic scenario changes, achieving a stable balance in overall system performance.

[0043] The data acquisition module outputs standardized data to the data processing module, the data processing module outputs a binding identifier to the privacy algorithm fusion scheduling module, the privacy algorithm fusion scheduling module outputs atomic computation graph data and policy audit logs to the multiple optimization resource module and the security instruction generation module, the multiple optimization resource module triggers the arbitrator threshold judgment and activates the dual closed-loop feedback module, the dual closed-loop feedback module outputs the updated dynamic knowledge base weight parameters to the data processing module, outputs the resource allocation threshold to the multiple optimization resource module, the security instruction generation module outputs the device control instructions to the terminal device, and feeds back the execution delay indicator to the dual closed-loop feedback module.

[0044] The protocol adaptation layer converts the raw data stream into preprocessed data containing the device type code. The identifier injection unit indexes the privacy classification rule base based on the device type code and embeds the matching privacy classification identifier into the data header to form standardized data. This standardized data is transmitted to the data processing module via a zero-copy channel, providing structured input that includes device attributes and privacy levels.

[0045] The hierarchical feature extraction engine parses environmental parameters from standardized data: the first processing layer outputs a scalar for light intensity, and the second layer outputs a scalar for target density. The feature fusion unit dynamically weights the privacy-level identifiers to generate a composite feature vector. The dynamic knowledge base uses a k-nearest neighbor algorithm to match the strategy mapping table and outputs a bound optimization algorithm identifier and a collaborative strategy identifier. This bound identifier pair is transmitted to the privacy algorithm fusion scheduling module via a message queue.

[0046] The identifier resolution unit locates the coordinates of nodes in the pre-trained model's computational graph. The node injection engine inserts perturbation nodes with configured noise parameters and links to the encryption library. The compilation engine generates atomic computational graph data that incorporates privacy operations and outputs it to the multi-optimization resource module. Simultaneously, a policy audit log is created to record node parameter configurations, and the log file is output to the security instruction generation module.

[0047] The resource allocation unit schedules computing resources to execute atomic computational graph data, rehearsing resource plans in a parallel simulation environment. The arbitration analysis unit calculates the Euclidean distance between simulation metrics and actual execution metrics. When three consecutive detection results exceed the dynamic threshold, a circuit breaker trigger is generated to activate the dual closed-loop feedback module. The arbitrator simultaneously sends a circuit breaker signal to the container management unit.

[0048] The data feedback unit converts device execution delay metrics into weighted correction parameters, which are then transmitted to the dynamic knowledge base policy update unit and the container policy driver engine. The control feedback unit outputs sensitivity parameters based on the confidence level grading signal to the multi-optimization resource module and arbitrator. The weight parameters update the knowledge base policy priorities and calibrate the noise amplitude, while the sensitivity parameters adjust the resource allocation strategy and deviation detection response.

[0049] The reverse verification unit quantifies the irreversibility of the feature vector, and the audit encapsulation unit binds the verification signal with the policy audit log to generate a digitally signed audit package. Finally, the control instruction encapsulates the device operation command and the audit package, outputting them to the terminal device for execution. The execution delay indicator is fed back to the dual closed-loop feedback module through a feedback channel, forming a closed-loop control system.

[0050] The data acquisition module implements communication protocol conversion for multi-source heterogeneous devices through a protocol adaptation layer. This protocol adaptation layer, with its built-in multi-protocol parsing engine, converts the raw data streams from terminal devices into pre-processed data in a unified data format. The identifier injection unit, based on the device type encoding index of the privacy classification rule base, embeds the privacy classification identifier in the pre-processed data header through a metadata extension mechanism, generating standardized data output for the data processing module. This process converts raw data into structured data, providing a unified input with privacy attributes for subsequent modules.

[0051] The data processing module's hierarchical feature extraction engine processes environmental parameters in parallel. The first processing layer performs brightness histogram analysis on video frame data, extracting high-percentile brightness values as a quantitative indicator of light intensity. The second processing layer calculates the density of moving objects per unit area using background modeling and morphological filtering techniques. The feature fusion unit dynamically loads weighting coefficients based on privacy classification identifiers, fusing light intensity and object density features into a composite feature vector. The dynamic knowledge base matches feature vectors to a pre-set strategy mapping table and outputs a synchronously bound optimization algorithm identifier and collaborative strategy identifier to the privacy algorithm fusion scheduling module, forming an atomic association between algorithm selection and privacy policy.

[0052] The privacy algorithm fusion scheduling module parses the optimization algorithm identifier to locate the node position in the pre-trained model computation graph. It also indexes the privacy rule base associated with the collaborative policy identifier to obtain the noise parameters. A noise injection node with configured noise parameters is inserted after the output node of the computation graph feature extraction layer, and a dynamic link is made to the privacy encryption dynamic library. The computation graph rewriting engine compiles the privacy operation nodes into a subgraph structure of the algorithm computation graph, generating atomic computation graph data for the fused privacy operations and simultaneously outputting a policy audit log containing node location coordinates and parameter configuration details.

[0053] Multiple optimization resource modules deploy a dual channel of optimization algorithm executors and parallel simulation environments: The executors allocate computing resources to run the atomic computation graph data, while the simulation environment replicates real-world input data to rehearse resource allocation plans. Multiple detection arbitrators compare the Euclidean distance between simulation metrics and actual execution metrics in real time. A dual closed-loop feedback module is activated when dynamic thresholds are exceeded three consecutive times. This mechanism improves the reliability of resource decisions through parallel rehearsals.

[0054] The reverse verification unit of the security instruction generation module performs reconstruction and inversion on the output feature vector, verifying data irreversibility through deviation analysis. The audit encapsulation unit combines the irreversibility verification results with the policy audit log to generate a digitally signed audit package, which is attached to the device control instruction to form the final instruction. This process builds a complete chain of evidence for privacy operations.

[0055] The dual closed-loop feedback module converts device execution delay metrics into weight correction parameters through a data feedback loop, updating the mapping weights in the dynamic knowledge base. A control feedback loop adjusts resource allocation threshold parameters based on simulation confidence grading signals. The updated weight parameters are fed back to the data processing module for optimized strategy matching, and the adjusted threshold parameters are fed into the multi-resource optimization module to achieve a dynamic balance between resource efficiency and privacy strength.

[0056] Specifically, the data fusion intelligent device linkage management and control system based on the large model described in the present invention has a data acquisition module including a protocol adaptation layer and an identifier injection unit; The protocol adaptation layer converts the terminal device communication protocol into a unified data format and outputs pre-processed data carrying the device type code; The identifier injection unit receives the pre-processed data, indexes the privacy classification rule base based on the device type encoding, and generates a matching privacy classification identifier to be embedded in the data header; The privacy classification rule base stores the mapping between device types and privacy levels and dynamically responds to indexing requests; The identifier injection unit outputs the standardized data integrating the privacy graded identifier to the data processing module; The protocol adaptation layer processes the original data stream to generate pre-processed data and outputs it to the identifier injection unit; The identifier injection unit extracts the device type code from the pre-processed data and calls the privacy classification rule library; The identifier injection unit embeds the generated privacy-graded identifier into the preprocessed data, generates standardized data and outputs it to the data processing module.

[0057] The protocol adaptation layer has a built-in multi-protocol parsing engine, loading corresponding communication driver components for different terminal device types. Cameras activate the video stream parsing protocol to convert raw video data into structured frame sequences. Environmental sensors activate the lightweight communication protocol conversion module to generate time-series data packets. The access controller invokes the device state converter to generate structured state records. All processed data is encapsulated into pre-processed data in a unified format, and a unique device type code is embedded in the data header as metadata identifier.

[0058] After receiving the preprocessed data, the identifier injection unit extracts the device type code as an index key to access the privacy classification rule base. The privacy classification rule base uses an in-memory database to store the mapping between device types and privacy levels, achieving millisecond-level response times through a hash indexing mechanism. After the rule base returns a matching privacy classification identifier, the identifier injection unit inserts the privacy classification identifier field into the preprocessed data header using a metadata extension mechanism.

[0059] The privacy grading rule base is deployed in a distributed architecture and stores a table mapping device type codes to privacy levels. Upon receiving an index request from the identifier injection unit, the rule base locates the storage node using a consistent hashing algorithm and returns the corresponding privacy level identifier in real time. This process enables dynamic mapping of device types to privacy attributes.

[0060] The protocol adaptation layer uses a pipeline architecture to process raw data streams. The raw data is fed into the protocol identifier to determine the device type. The corresponding protocol parser is then called to convert the data format. Finally, the device type code is added to generate preprocessed data. This preprocessed data is transferred to the identifier injection unit via a zero-copy memory channel, reducing data transmission overhead.

[0061] The identifier injection unit extracts the device type code using a metadata parser. The metadata parser separates the preprocessed data header information, reads the device type code, and then initiates an index request to the privacy classification rule library. After the rule library returns the privacy classification identifier, the metadata expander writes it into the data header extension field, generating standardized data output for downstream modules.

[0062] The entire data collection process forms a closed-loop processing chain: the protocol adaptation layer unifies the data format and adds device type metadata, and the identifier injection unit dynamically embeds privacy attribute identifiers based on this metadata. The structured output of standardized data provides a complete data foundation, including device type and privacy level, for subsequent processing modules, supporting the implementation of differentiated privacy policies.

[0063] Specifically, the data fusion intelligent device linkage management and control system based on a large model of the present invention also includes: The data processing module is equipped with a hierarchical feature extraction engine; The first processing layer performs lighting environment analysis: it inputs a video frame in normalized data, extracts the brightness histogram, calculates the 95th percentile brightness value, and outputs a brightness scalar; The second processing layer performs dynamic target density analysis: it inputs normalized data, applies background modeling and morphological filtering, counts the number of moving objects per square meter, and outputs a density scalar. The feature fusion unit receives the brightness scalar and the density scalar, queries the dynamic weighting coefficient according to the privacy classification identifier, and calculates the composite feature vector; The dynamic knowledge base matches the composite feature vector with the pre-stored scenario strategy mapping relationship, and outputs the bound optimization algorithm identifier and collaborative strategy identifier to the privacy algorithm fusion scheduling module; Standardized data is input into the first processing layer and the second processing layer; The first processing layer outputs the brightness scalar to the feature fusion unit; The second processing layer outputs the density scalar to the feature fusion unit; The feature fusion unit receives the privacy classification identifier, dynamically calculates the weighted value, and fuses the brightness scalar and the density scalar to generate a composite feature vector.

[0064] The hierarchical feature extraction engine utilizes a parallel processing architecture. The first processing layer performs illumination analysis on the video frame sequence from the standardized data. This layer extracts the histogram distribution features of the video frame's luminance channel and quantifies the ambient light intensity by counting the high percentile luminance values. The illumination analysis output is a scalar luminance index that indicates the degree to which the current scene's lighting conditions affect device interaction.

[0065] The second processing layer performs dynamic target density analysis on the standardized data, using a Gaussian mixture model to construct the scene background. Foreground target segmentation and morphological filtering eliminate noise interference and accurately identify moving targets. The number of valid targets per unit area is counted to generate a density scalar, which reflects the distribution density characteristics of the target objects in the scene.

[0066] The feature fusion unit receives the brightness and density scalars output by the dual processing layer and retrieves a dynamic weighting coefficient table based on the privacy level identifier in the standardized data header. The weighting coefficient table stores the feature weight configuration rules corresponding to different privacy levels. The feature fusion unit then executes a weighted fusion algorithm to generate a composite feature vector. This vector comprehensively represents the correlation between ambient lighting conditions and target distribution density.

[0067] The dynamic knowledge base stores a pre-configured scenario strategy mapping table, using a feature vector similarity matching mechanism. This mapping table records the binding relationships between composite feature vector ranges, optimization algorithm identifiers, and collaborative strategy identifiers. The matching engine retrieves the vector range closest to the input feature vector and outputs the corresponding binding identifier pair to the privacy algorithm fusion scheduling module.

[0068] Normalized data is processed in parallel within the hierarchical feature extraction engine: video frame data is fed into the first processing layer to generate a brightness scalar, while device status and environmental data are fed into the second processing layer to generate a density scalar. The dual scalar outputs are transmitted via a data bus to the feature fusion unit, which dynamically loads weight configuration rules based on the privacy classification identifier.

[0069] The feature fusion unit performs privacy-weighted fusion of environmental features: It applies linear weighting to the brightness and density scalars, applying weighting coefficients corresponding to the numeric level of the privacy classification identifier. The resulting vector is a dimensionally compressed composite feature vector, which serves as a comprehensive representation of scene features and is fed into the dynamic knowledge base for policy matching.

[0070] The entire processing flow forms a closed loop from environmental feature extraction to policy decision-making: layered processing quantifies environmental parameters, privacy identifiers drive feature fusion and weighting, and composite feature vectors match pre-set policies to generate binding identifiers. This mechanism dynamically couples environmental features with privacy policies, providing a decision-making foundation for subsequent privacy-enhancing computations.

[0071] Specifically, the data fusion intelligent device linkage management and control system based on the big model described in the present invention includes a dynamic knowledge base including a strategy index layer, a weight storage layer and a strategy update unit; The strategy index layer stores the mapping relationship between the composite feature vector and the bound optimization algorithm identifier and the collaborative strategy identifier; The weight storage layer records the historical performance indicators of each feature vector in the mapping relationship, including response delay and recognition accuracy; The strategy update unit receives the execution delay index transmitted by the data feedback loop and converts the delay index into a weight correction coefficient; The policy update unit receives the confidence level signal transmitted by the control feedback loop and dynamically adjusts the update frequency according to three levels: high / medium / low; The strategy update unit integrates the weight correction coefficient and the update frequency parameter, and outputs the update instruction to the strategy index layer to adjust the mapping relationship; The policy update unit synchronously outputs update instructions to the weight storage layer to calibrate historical performance indicators. The weight storage layer outputs the latest historical performance indicators to the policy index layer to optimize real-time matching accuracy. The data feedback loop transmits the execution delay indicator input to the policy update unit. The control feedback loop transmits the confidence grading signal input to the policy update unit. The policy update unit outputs update instructions to the policy index layer. The policy update unit outputs update instructions to the weight storage layer. The weight storage layer transmits the latest historical performance indicators to the policy index layer. The policy index layer outputs the optimized bound optimization algorithm identifier and collaborative strategy identifier to the privacy algorithm fusion scheduling module.

[0072] The policy indexing layer uses a distributed key-value storage architecture to store mappings between composite feature vector hash values and binding identifier pairs. These binding identifier pairs consist of an optimization algorithm identifier and a collaborative strategy identifier, which maintain a fixed association. The indexing engine uses a feature vector similarity matching algorithm to retrieve the closest mapping record and output the corresponding binding identifier pair to downstream modules.

[0073] The weight storage layer constructs a time series database that continuously records the historical execution metrics of each feature vector-associated strategy. This metric data includes response latency distribution statistics and recognition accuracy curves. The storage unit uses a sliding window mechanism to retain the latest historical data. The weight calculator dynamically calculates the combined weight of the feature vectors based on a time decay model, providing a priority reference for strategy matching.

[0074] The policy update unit consists of a signal converter and an update scheduler. The signal converter linearly maps the execution delay indicator of the data feedback loop into a weight correction coefficient. The update scheduler analyzes the confidence level signal of the control feedback loop and generates an update frequency parameter based on the rule of high-confidence frequent updates and low-confidence infrequent updates. The fusion engine combines the weight correction coefficient and the update frequency parameter into an update instruction.

[0075] The update command is transmitted to the policy indexing layer, triggering mapping adjustments: the mapping adjuster reorders the association priorities between feature vectors and identifiers based on the weight correction coefficients and simultaneously updates the feature vector matching range thresholds. Upon receiving the update command, the weight storage layer activates the metric calibrator, recalculates historical performance metrics based on the latest weight parameters, and outputs the calibrated metric data to the policy indexing layer to optimize the real-time matching algorithm.

[0076] The data feedback loop feeds the device execution latency metric into the policy update unit's signal converter, while the control feedback loop feeds the confidence rating signal into the update scheduler. The update instructions generated by the policy update unit are synchronously sent to the policy indexing layer and the weight storage layer. The weight storage layer outputs the calibrated historical performance metric to the policy indexing layer, driving the indexing layer to output optimized binding identifier pairs.

[0077] The entire dynamic update mechanism forms a closed-loop control loop: delayed feedback quantifies the actual effectiveness of the policy, while confidence signals characterize the reliability of simulation rehearsals. Update instructions drive the coordinated optimization of mapping relationships and weight indicators, ensuring that the policy output continuously adapts to the system's operating state. The optimized binding identifier ensures real-time coordination between algorithm scheduling and privacy policies.

[0078] Specifically, the privacy algorithm fusion scheduling module of the data fusion intelligent device linkage management and control system based on the big model of the present invention is also used to: Receive the synchronization binding optimization algorithm identifier and the collaboration strategy identifier output by the data processing module; Parse the optimization algorithm identifier to locate the output node position of the feature extraction layer in the pre-trained model calculation graph; The privacy rule library corresponding to the index collaborative strategy identifier obtains the noise amplitude parameter, inserts the noise injection node configured with the noise amplitude parameter after the located node position, associates the collaborative strategy identifier to dynamically link the privacy encryption dynamic library to the noise injection node, compiles the computational graph embedded with the noise node and the encryption library to generate atomic computational graph data, and simultaneously generates a policy audit log that records the node positioning, noise parameter configuration, and encryption library link operation details, outputs the atomic computational graph data to the multiple optimization resource module, and outputs the policy audit log to the security instruction generation module.

[0079] The privacy algorithm fusion scheduling module receives the synchronization binding identifier pair output by the data processing module. This identifier pair includes the optimization algorithm identifier and the coordination strategy identifier. The parsing engine parses the optimization algorithm identifier, locates the corresponding algorithm computation graph structure through the pre-trained model registry, and accurately identifies the output node coordinates of the feature extraction layer. These coordinates serve as the insertion anchor point for the privacy operation node.

[0080] The collaborative strategy identifier is used as an index key to access the privacy rule base, which returns a set of noise amplitude parameter configurations. The node injection engine inserts a noise injection node after the output node of the located feature extraction layer and loads the acquired noise amplitude parameters into the node attribute configuration. The dynamic linker associates the collaborative strategy identifier with the privacy encryption dynamic library and establishes a runtime encryption library call path to the noise injection node.

[0081] The computational graph compilation engine performs an atomic rewrite operation: the noise injection node and the associated cryptographic library call path are embedded into the original algorithm computation graph, generating a computational subgraph that incorporates privacy operations. The compilation process preserves the original computational logic, forming an indivisible execution unit. A policy audit log is also generated simultaneously, recording the location coordinates of the noise injection node, the noise amplitude parameter value, and the dynamic link address of the cryptographic library.

[0082] Atomized computation graph data is output to the Multi-Optimization Resource Module for execution, and policy audit logs are transmitted to the Security Instruction Generation Module. The computation graph data contains complete algorithm instructions and privacy-preserving operation nodes, eliminating the overhead of independently scheduling privacy-preserving components. Audit logs are stored in a structured format, including timestamps, node identifiers, and parameter hash values, providing a data foundation for operation tracing.

[0083] The entire scheduling process achieves deep integration of algorithms and privacy technologies: binding identifiers ensure scheduling consistency, node positioning enables precise operation insertion, and dynamic linking supports hot reloading of privacy components. Atomic compilation ensures the integrity of computational logic, and audit logs provide full-link traceability.

[0084] Specifically, the data fusion intelligent device linkage management and control system based on a large model of the present invention also includes: Atomized computation graph data encapsulation in independent containers; The container integrates a policy consistency engine to perform the following collaborative operations: Load the algorithm executable file that integrates privacy operations; Dynamically parse the configuration file corresponding to the collaborative strategy identifier and generate the noise amplitude and encryption library link parameters in real time; Synchronously start the logging service to continuously capture the running status of the noise injection node and the encryption library; Automatically verify the matching between the configuration file digital signature and the registration signature of the collaborative policy identifier when the container is started; Synchronize the collaborative strategy identifier changes to the noise injection node and encryption library link in real time during runtime.

[0085] Atomized computational graph data is deployed in an independent runtime environment via a containerized packaging engine. The container image generator packages the algorithm executable files and dependent libraries that incorporate privacy-integrated operations into a lightweight virtualized image, ensuring the security of the execution environment through resource isolation. This image is loaded upon container startup to initialize the running instance.

[0086] The policy consistency engine is integrated within the container and dynamically parses the configuration file associated with the collaborative policy identifier at runtime. The configuration file parser extracts the noise amplitude parameters and the encryption library link address, updating them in real time to the in-memory parameter pool. The parameter pool is synchronized to the noise injection node and the encryption library call interface via a shared memory channel.

[0087] The logging service is automatically activated when the container starts, using bypass monitoring mode to capture the output data distribution characteristics of the noise injection node. It also monitors the call frequency and response status of the privacy-encrypted dynamic library, generating timestamps for the operation status and writing them to the circular log buffer.

[0088] During the container initialization phase, a security verification process is performed: a digital signature verifier compares the digital signature of the configuration file to be loaded with the registration signature associated with the collaborative policy identifier. The signature verification module calculates the hash value deviation. If the deviation exceeds a predetermined threshold, a circuit breaker is triggered to prevent the container from starting.

[0089] Containers respond to policy changes in real time during runtime: Event listeners capture collaborative policy identifier update messages and configure the reload engine to parse the parameter set corresponding to the new identifier. The parameter updater hot-updates the noise amplitude parameter and encryption library link address to the running instance while maintaining computational continuity.

[0090] The entire container operation and management mechanism enables dynamic adaptation of computing units: a policy consistency engine maintains real-time operational parameters, a log service provides operational status visibility, and a security verification mechanism protects against configuration tampering risks. Hot update capabilities support dynamic adjustment of privacy policies, ensuring system robustness in changing environments.

[0091] Specifically, the data fusion intelligent device linkage management and control system based on a large model of the present invention and the container internal deployment policy driving engine are also used to: Load the algorithm executable file that integrates the noise node and the encryption library, parse the configuration file corresponding to the collaborative strategy identifier, and generate the noise amplitude parameters and encryption library link instructions; Start the logging service to capture the noise node output data and the cryptographic library operation status; During the container initialization phase, the hash value deviation between the configuration file digital signature and the registration signature of the collaborative policy identifier is verified. If the deviation is greater than 5%, a circuit breaker alarm is triggered. During the container operation phase, monitor the collaborative policy identifier change event and reconfigure the noise amplitude parameters and the encryption library link within ≤50 milliseconds; The integrity of the configuration file is periodically verified. When tampering is detected, the container is frozen and a backup instance is started. The logging service outputs noise indicators and encrypted audit records to the security instruction generation module in real time.

[0092] The policy-driven engine loads the algorithm executable file for the noise injection node and encryption library. It then uses the configuration parser to read the configuration file associated with the collaborative policy identifier. The configuration file contains noise amplitude configuration items and the encryption library link address. The parser generates the corresponding parameter instruction set and transmits it to the runtime parameter register. The parameter register distributes the instruction set to the noise injection node execution unit and the encryption library call interface through a memory mapping mechanism.

[0093] The logging service utilizes a non-intrusive monitoring architecture, using data probes to capture the output data distribution characteristics of noise-injected nodes. Furthermore, the interface monitoring module collects call status data from privacy-encrypted dynamic libraries, generating audit records containing operation timestamps and status codes. The log transmitter transmits structured logs to downstream modules in real time via dedicated channels.

[0094] During the container initialization phase, a security verification process is performed: the signature verification module calculates the digital signature hash value of the configuration file and compares it with the registered signature reference value bound to the collaborative policy identifier. The hash deviation analyzer detects the degree of signature discrepancy. When the deviation exceeds a predetermined threshold, the circuit breaker generator is triggered, blocking the container startup process and activating an alarm notification.

[0095] During the container runtime phase, an event-driven mechanism is deployed: a change listener captures collaborative strategy identifier update events in real time and configures the reload engine to parse the parameter set corresponding to the new identifier. A parameter updater atomically replaces the noise amplitude value and encryption library link address in the memory parameter register, completing the hot update operation while maintaining computational continuity.

[0096] The integrity check module periodically scans the configuration file storage area, using a hash checker to compare the file's current hash value with an initial baseline value. If a hash mismatch is detected, the container freezer suspends the currently running instance, while the container orchestrator launches a clean backup container and loads the original image. The log relay continuously transmits noise metric statistics and cryptographic operation audit records to the security instruction generation module.

[0097] Specifically, the data fusion intelligent device linkage management and control system based on a large model of the present invention also includes: The arbitrator calculates the Euclidean distance between the simulation indicator and the actual execution indicator. When the distance exceeds the dynamic threshold three times in a row, the dual closed-loop feedback module is activated and a fuse trigger signal is sent to the container. After receiving the circuit breaker signal, the container freezes the currently running instance, starts a clean backup container to load the atomic computation graph data, records the container switching delay time, and feeds the switching delay indicator back to the dual closed-loop feedback module. The control feedback loop adjusts the dynamic threshold based on the simulation confidence level signal. The threshold for high-confidence scenarios is set to 80% of the dynamic threshold baseline value, and the threshold for low-confidence scenarios is set to 120% of the dynamic threshold baseline value. The dual closed-loop feedback module calibrates the generation logic of the confidence grading signal based on the container switching delay index. When the switching delay is greater than 50 milliseconds, the simulation confidence grading signal is automatically downgraded by one level.

[0098] The arbitrator deploys a multi-dimensional metric collector to capture, in real time, simulation metrics output by the parallel simulation environment and operational metrics from the actual execution pipeline. The distance calculation engine uses a Euclidean distance algorithm to quantify the difference between the dual-channel metrics and maintains the calculation results for three consecutive detection cycles through a sliding window mechanism. If the distance value exceeds the dynamic threshold three times in a row, the arbitrator generates a circuit breaker trigger instruction, activating the dual closed-loop feedback module and simultaneously sending a circuit breaker signal to the container instance.

[0099] Upon receiving the circuit breaker signal, the container management unit initiates emergency response: the instance freezer immediately suspends the memory state and processes of the currently running container, preventing potential error propagation. The container orchestrator simultaneously launches a clean backup instance, loading the original image of the atomic computation graph data from persistent storage. The switchover process records the time difference between the initial container freeze and the backup container startup completion time, generating an accurate switchover latency metric that is transmitted via a feedback channel to the dual closed-loop feedback module.

[0100] The control feedback loop receives confidence-level signals from the parallel simulation environment. The signal classification logic is based on historical simulation accuracy. The threshold adjuster dynamically adjusts the arbitrator's dynamic threshold based on the confidence level: a tightening strategy reduces the threshold ratio in high-confidence scenarios, while a loosening strategy increases the threshold ratio in low-confidence scenarios. The adjusted threshold ratio is then applied to subsequent anomaly detection decisions.

[0101] The calibration engine of the dual closed-loop feedback module receives container switching delay metrics and uses a time delay analyzer to evaluate switching efficiency. When the delay exceeds a predetermined threshold, a confidence degrader automatically downgrades the simulation confidence level by one level. This downgraded confidence level is then re-inputted into the control feedback loop, forming a closed-loop control chain with adaptive detection accuracy.

[0102] The arbitrator triggers circuit breaking based on indicator distance detection. The container management unit performs hot switching and reports latency data. The confidence signal dynamically degrades based on switching efficiency, and the degradation signal readjusts the detection threshold. This entire process achieves a dynamic balance between abnormal response and detection accuracy, improving the system's fault tolerance under complex operating conditions.

[0103] Specifically, the data fusion intelligent device linkage management and control system based on the big model described in the present invention, the security instruction generation module includes a reverse verification unit and an audit encapsulation unit; The reverse verification unit receives the feature vector output by the atomic computation graph data, performs feature reconstruction to generate reconstructed data, and calculates the deviation between the feature vector and the reconstructed data; When the deviation value is less than 5%, the irreversibility verification pass signal is output; When the deviation value is ≥5%, a fuse trigger signal is output to the container instance of claim 7; The audit encapsulation unit receives the irreversibility verification pass signal output by the reverse verification unit, receives the policy audit log output by the privacy algorithm fusion scheduling module, binds the irreversibility verification pass signal with the policy audit log to generate a digitally signed audit package, encapsulates the device control instruction and the digitally signed audit package to generate the final control instruction, and the final control instruction carries the irreversibility verification pass signal and the operation audit traceability chain.

[0104] The reverse verification unit receives the feature vectors output by the atomized computation graph data and performs feature reconstruction using a lightweight decoding model. The reconstruction process uses a generative inversion algorithm to simulate the original data generation path, generating reconstructed data samples that are consistent with the original data structure. The deviation analyzer calculates the statistical distribution difference between the feature vector and the reconstructed data to quantify the strength of privacy protection.

[0105] When the deviation analysis result falls below a predetermined critical value, the signal generator outputs an irreversibility verification pass signal. This signal includes a timestamp and a hash digest of the feature vector, confirming that the data desensitization operation complies with privacy protection requirements. When the deviation value reaches or exceeds a critical threshold, the fuse signal generator outputs a fuse trigger signal, which is transmitted via a dedicated channel to the container instance described in claim 7.

[0106] The audit encapsulation unit receives the irreversibility verification pass signal and the policy audit log. The signal binding engine aligns the two input data by timestamp and generates a bound data packet containing the verification results and operation records. The digital signature module performs asymmetric cryptographic signatures on the bound data packet to generate a digitally signed audit packet. The instruction encapsulator uses the device control instruction as the primary data body and appends the digitally signed audit packet to generate the final control instruction.

[0107] The final control instruction adopts a layered encapsulation structure: the instruction header stores the device control command code, the payload area embeds the device operating parameters, and the audit attachment area carries a digitally signed audit package. The audit attachment includes an irreversible verification pass signal and a complete operation audit traceability chain, which records the entire process from feature generation to instruction output.

[0108] The security instruction generation module ensures verifiability of privacy protection and auditability of operational processes. Reverse verification quantifies the strength of desensitization, a circuit breaker mechanism blocks privacy leaks, and digital signatures ensure the integrity of audit data. Structured instructions meet the dual requirements of device control and regulatory compliance. Ultimately, control instructions are output to the terminal device for execution, completing the system's closed-loop processing.

[0109] Specifically, the data fusion intelligent device linkage management and control system based on a large model of the present invention also includes: The data feedback loop receives the delay indicator generated by the device executing the control instruction, converts the delay indicator into the dynamic knowledge base weight correction parameter, and transmits the weight correction parameter to the policy update unit of the dynamic knowledge base and the container policy driving engine; A control feedback loop receives the confidence level signal output by the parallel simulation environment, dynamically adjusts the arbitrator sensitivity parameters according to the three levels of high confidence, medium confidence, and low confidence, and transmits the sensitivity parameters to the multiple optimization resource modules and the arbitrator; In high-confidence scenarios, the weight correction parameter is tightened to the current value of the dynamic knowledge base weight parameter × 80%, and in low-confidence scenarios, the sensitivity parameter is relaxed to the current value of the arbitrator sensitivity parameter × 120%; The policy update unit updates the knowledge base mapping relationship based on the weight correction parameters. The container policy driving engine calibrates the noise injection amplitude based on the weight correction parameters. The multiple optimization resource module adjusts the resource allocation strategy based on the sensitivity parameters. The arbitrator dynamically adjusts the deviation detection response speed based on the sensitivity parameters.

[0110] The data feedback loop receives the latency metrics generated by terminal devices executing control commands. A metric converter linearly maps the latency data into a weight correction coefficient. This coefficient represents the effectiveness of policy execution and is transmitted via the data bus to the policy update unit of the dynamic knowledge base and the container policy driver engine. The transmission process uses a timestamped message queue to ensure data timing integrity.

[0111] The control feedback loop receives confidence-level signals from the parallel simulation environment. A signal classifier classifies the signals into three levels: high, medium, and low confidence, based on pre-set rules. A sensitivity adjuster dynamically calculates the arbitrator's sensitivity parameters based on the confidence levels: high-confidence scenarios increase detection sensitivity, while low-confidence scenarios enhance fault tolerance. These sensitivity parameters are synchronously transmitted to the multiple optimization resource modules and the arbitrator via control channels.

[0112] Driven by confidence signals, the weight correction parameters and sensitivity parameters are dynamically adapted: a tightening strategy is used to reduce the relative proportion of the weight correction parameters in high-confidence scenarios, while a relaxation strategy is used to increase the relative proportion of the sensitivity parameters in low-confidence scenarios. This mechanism balances the accuracy and stability requirements of the system under complex operating conditions.

[0113] The policy update unit updates the mapping relationships in the dynamic knowledge base based on the weight correction parameters and adjusts the weights associated with feature vectors and identifiers using a priority reordering algorithm. The container policy driver engine synchronously receives the weight correction parameters and dynamically adjusts the perturbation intensity of the noise injection node using the amplitude calibrator. The multi-optimization resource module optimizes the computing resource allocation strategy based on the sensitivity parameters, and the arbitrator dynamically adjusts the response frequency of deviation detection based on the sensitivity parameters.

[0114] A dual-loop system forms a collaborative optimization network: data feedback quantifies actual execution results, while control feedback predicts system reliability. Parameter transfer enables real-time linkage between the decision-making and execution units. Weight correction optimizes long-term strategy quality, while sensitivity adjustment ensures short-term execution stability. The system achieves a dynamic balance between resource efficiency and privacy through continuous parameter iteration.

[0115] This invention addresses the core contradictions in cross-scenario data fusion by establishing an integrated collaborative mechanism for algorithm scheduling and privacy protection. During the environmental feature analysis phase, the data processing module outputs a synchronously bound optimization algorithm identifier and a collaborative strategy identifier, forming an atomic association between algorithm selection and privacy strategy. This bound identifier pair is transmitted to the privacy algorithm fusion scheduling module, where computational graph rewriting technology compiles privacy operations into subgraph nodes of the algorithm computation graph, generating atomic computational graph data for the fused privacy operations. This process eliminates the scheduling delays caused by the independent operation of the algorithm module and the privacy module in traditional architectures, achieving deep integration of computational instructions.

[0116] At the resource scheduling level, multiple resource optimization modules deploy parallel simulation environments to rehearse resource plans. An arbitrator compares simulated metrics with actual execution metrics in real time. When deviations exceeding the specified threshold are continuously detected, a dual closed-loop feedback module is activated, triggering a dynamic adjustment mechanism. The data feedback loop converts device execution delays into knowledge base weight parameters to optimize policy decision accuracy. The control feedback loop adjusts resource allocation thresholds based on simulation confidence grading signals, achieving a dynamic balance between resource efficiency and privacy. The synergistic effect of these dual closed-loops enables the system to adapt to changing scenarios, maintaining an optimal balance between computing efficiency and security.

[0117] Privacy and security are strengthened through a full-link audit mechanism. The reverse verification unit of the security instruction generation module quantifies the irreversibility of feature vectors. Exceeding deviations trigger the container's fuse mechanism, blocking privacy leaks. The audit encapsulation unit binds the irreversibility verification signal to the policy audit log to generate a digitally signed audit package. The final control instruction carries this audit package and the operation traceability chain and is output to the terminal device. This design meets the requirements for data desensitization and irreversibility verification, providing legally compliant operational evidence.

[0118] Ultimately, the identifier-binding transmission mechanism solves the timing issues of policy coordination, the atomic computation graph eliminates scheduling resource competition, dual closed-loop feedback enables dynamic environment adaptation, and audit traceability ensures privacy-preserving operational compliance. These four layers of technology collaborate to overcome the limitations of traditional system architectures, achieving an efficient unification of algorithmic adaptive scheduling and privacy security through cross-scenario data fusion.

[0119] The specific implementation of the present invention is as follows: The raw input consists of the video stream captured by the camera and the access controller status data. The protocol adaptation layer activates the ONVIF protocol parser to convert the video frames into a structured sequence and enables the Modbus-TCP converter to generate access control event records. After preprocessing the data and adding the device type code, the identifier injection unit indexes the privacy classification rule library to obtain the PII-C1 privacy identifier. This identifier is then embedded into the data header through the metadata extension mechanism to generate standardized data.

[0120] Standardized data is fed into the hierarchical feature extraction engine. The first processing layer extracts the brightness histogram of the video frame, calculates the 95th percentile brightness value, and outputs a light intensity scalar. The second processing layer uses a Gaussian mixture model to segment moving objects, counting the number of valid objects per unit area and outputting a density scalar. The feature fusion unit loads dynamic weighting coefficients based on the PII-C1 identifier, fusing the two scalars to generate a composite feature vector. The dynamic knowledge base matches the feature vector with the preset policy table and outputs the bound facial recognition algorithm identifier and differential privacy policy identifier.

[0121] The privacy algorithm fusion scheduling module parses the algorithm identifier to locate the ResNet model feature layer nodes and indexes the differential privacy policy to obtain the noise parameters. After locating the node, a perturbation node is inserted to configure the noise amplitude and dynamically linked to the homomorphic encryption dynamic library. The computational graph rewriting engine generates atomic computational graph data that incorporates privacy operations and simultaneously outputs an audit log that records the node coordinates and encryption parameters.

[0122] Multiple resource optimization modules allocate GPU resources to execute computational graph data, while simultaneously driving the simulation environment to replicate the scenario input and preview resource plans. The arbitrator compares the Euclidean distance between the simulation results and the actual face recognition latency in real time. If the deviation exceeds a dynamic threshold three times in a row, the dual closed-loop feedback module is activated. At this point, the arbitrator sends a fuse signal to the container management unit.

[0123] The data feedback loop converts actual latency into weighted correction parameters, updating the mapping priorities in the dynamic knowledge base. The control feedback loop adjusts resource allocation thresholds based on the simulated confidence level signal: tightening the threshold to the baseline ratio in high-confidence scenarios and relaxing it in low-confidence scenarios. These updated parameters optimize resource allocation strategies and privacy operation strength in real time.

[0124] The reverse verification unit performs VAE reconstruction on the output feature vector, generating an irreversibility verification signal when the deviation falls below a critical value. The audit encapsulation unit binds this signal to the policy audit log, digitally signs it, and appends it to the access control instruction. If the deviation exceeds the limit or the arbitrator triggers a circuit breaker, the container management unit freezes the current instance and launches a clean backup container. The switch latency metric is fed back to the dual closed-loop module to calibrate the confidence model.

[0125] In the smart community access control scenario, this invention implements binding identifiers to eliminate the timing misalignment between algorithm scheduling and privacy policies. The atomic computational graph deeply integrates face recognition and differential privacy operations. Dual closed-loop feedback dynamically balances recognition efficiency and privacy strength based on latency indicators and confidence signals. The digital signature audit package provides a full-link operation traceability chain. This implementation process effectively resolves the core contradiction described in the background technology: the privacy protection lag caused by algorithm rescheduling when video streams and access control data are integrated. Through the collaborative efforts of four layers of technology, the accuracy of face recognition is maintained while the strength of differential privacy is maintained, reducing response delays to an acceptable range and eliminating the risk of privacy leaks.

[0126] The protocol adaptation algorithm processes data from multiple heterogeneous devices. The multi-protocol parsing engine identifies the terminal device type and calls the corresponding communication driver to convert the raw data stream. The camera video stream activates ONVIF protocol parsing to generate a structured frame sequence, while the access controller status data triggers the Modbus-TCP converter to generate event records. The converted data is then encapsulated into a unified format, pre-processed, and embedded with metadata identifying the device type.

[0127] A hierarchical feature extraction algorithm performs quantitative environmental analysis. The first processing layer uses a histogram statistics algorithm to process the brightness channel of the video frame, calculating the high percentile brightness value and outputting a light intensity scalar. The second processing layer runs a Gaussian mixture model to model the background, combined with a morphological filtering algorithm to segment moving objects, and counts the number of valid objects per unit area to output a density scalar. The feature fusion algorithm dynamically loads weighting coefficients based on the privacy classification identifier and linearly weights the dual scalars to generate a composite feature vector.

[0128] A dynamic policy matching algorithm optimizes decision making. The policy indexing engine uses a feature vector similarity matching algorithm to retrieve the bound algorithm identifier and privacy policy identifier from a pre-set mapping table. A weight storage algorithm records historical performance metrics and dynamically calculates feature vector weights based on a time decay model. The policy update algorithm integrates execution latency metrics and confidence signals to generate mapping adjustment instructions and weight calibration instructions.

[0129] The computational graph rewriting algorithm deeply integrates privacy operations. The node location algorithm parses the optimization algorithm identifier and determines the coordinates of the output nodes in the pre-trained model's feature layer. The noise injection algorithm inserts perturbation nodes for configuration parameters after the location nodes, associates the coordination strategy identifier, and dynamically links the encryption library. The compilation engine embeds the privacy nodes into the original computational graph, generating indivisible atomic execution units.

[0130] A dual closed-loop control algorithm dynamically balances system parameters. A data feedback algorithm converts device execution delays into weight correction coefficients, updating the knowledge base policy priorities. A control feedback algorithm adjusts resource allocation thresholds based on confidence-level signals: tightening the baseline threshold ratio in high-confidence scenarios and relaxing it in low-confidence scenarios. A sensitivity adjustment algorithm dynamically controls the arbitrator's detection response frequency.

[0131] The container hot update algorithm ensures operational continuity. The configuration parsing algorithm converts parameters corresponding to coordination policy identifiers in real time, atomically updating memory register values. The signature verification algorithm compares the digital signature of the configuration file with the registered signature hash value, triggering circuit breakers if the deviation exceeds the specified limit. The instance switching algorithm freezes abnormal containers and loads a clean backup image, recording switching delays and calibrating the confidence model.

[0132] The privacy verification algorithm builds an auditable chain of evidence. The feature reconstruction algorithm uses a lightweight decoding model to invert the original data, and the statistical distribution difference analysis algorithm quantifies the degree of data irreversibility. The audit binding algorithm aligns the verification signal with the operation log by timestamp, and the digital signature algorithm generates a tamper-proof audit package. The instruction encapsulation algorithm integrates control commands, operation parameters, and audit attachments into a hierarchical instruction structure.

[0133] Pretrained models serve as core computing units. The model library stores a variety of pretrained algorithm models. ResNet architecture models are used for face recognition tasks, while lightweight convolutional neural networks are used for environmental analysis tasks. The model computation graph structure is registered in the system registry, and the coordinates of node positions output by the feature extraction layer are predefined and stored. The privacy algorithm fusion scheduling module parses and optimizes algorithm identifiers to locate model nodes, providing anchor points for inserting privacy operations.

[0134] A Gaussian mixture model performs background modeling. The second processing layer deploys a multimodal Gaussian distribution model to process video frame sequences, learning the distribution of background features through an expectation-maximization algorithm. A foreground object segmentation algorithm identifies moving objects based on differences in pixel probability distributions, while a morphological filtering model eliminates noise interference and outputs an accurate target density scalar.

[0135] The feature reconstruction model verifies data irreversibility. The secure instruction generation module integrates a variational autoencoder model, taking as input the feature vectors output by the atomized computation graph. The decoder layer attempts to reconstruct the original data structure. The statistical distribution difference between the reconstructed data and the original feature vector quantifies the strength of privacy protection. When the deviation falls below a threshold, a signal indicating irreversibility verification is passed is generated.

[0136] A dynamic knowledge base model optimizes policy matching. The policy indexing layer constructs a composite feature vector space index tree and uses a k-nearest neighbor algorithm to match the most recent policy mapping record. The weight storage layer establishes a time series database model, using a sliding window mechanism to retain the latest historical performance indicators. The policy update model integrates execution delay and confidence signals to generate mapping relationship weight adjustment instructions.

[0137] The privacy rulebase model stores policy parameter configurations. A distributed key-value store records the mapping between collaborative policy identifiers and noise amplitudes. Differential privacy policies store Laplace noise parameters, and k-anonymity policies preserve generalization level configurations. The indexing service uses consistent hashing to quickly locate parameter sets and support real-time configuration of noise injection nodes.

[0138] The simulation environment model previews resource allocation scenarios. The parallel simulation module replicates real-world input data to build a virtualized computing resource pool model. A Monte Carlo simulation algorithm randomly generates multiple resource allocation scenarios. A time series prediction model evaluates the execution efficiency of each scenario and outputs the optimal scenario number to the multi-resource optimization module.

[0139] A confidence grading model quantifies simulation reliability. A control feedback loop integrates a historical data analysis model to calculate the correlation coefficient between simulation results and actual indicators. A hierarchical logic tree model categorizes confidence levels into high, medium, and low based on the correlation coefficient intervals, and outputs signals to adjust the arbitrator's detection sensitivity threshold.

[0140] This invention achieves deep synergy between algorithm scheduling and privacy security through a layered technical architecture. During the data collection phase, a protocol adaptation layer is deployed to convert multi-source device data into a unified format. An identifier injection unit indexes the privacy classification rule base based on device type encoding, generating a privacy classification identifier that is embedded in the data header. This process provides structured, standardized data for subsequent processing, and the privacy classification identifier drives the execution of differentiated protection strategies.

[0141] Environmental feature extraction utilizes a layered processing architecture. The first processing layer analyzes the brightness histogram of video frames to output a scalar light intensity value. The second processing layer uses background modeling to calculate a scalar moving object density. The feature fusion unit dynamically loads weighting coefficients based on the privacy classification identifier, fusing the two environmental features into a composite feature vector. The dynamic knowledge base matches the feature vectors to a pre-set strategy mapping table and outputs a bound optimization algorithm identifier and a collaborative strategy identifier. This binding mechanism atomically links algorithm selection with privacy policy, eliminating the timing misalignment caused by independent decision-making.

[0142] After receiving the binding identifier pair, the privacy algorithm fusion scheduling module parses the optimization algorithm identifier to locate the node position in the pre-trained model computation graph. It then uses the collaborative policy identifier to index the privacy rule base to obtain the noise parameters. It then inserts a noise injection node with the configured parameters after the located node and dynamically links to the privacy encryption dynamic library. The computation graph rewriting engine compiles the privacy operation nodes into algorithm subgraphs, generating atomic computation graph data for the fused privacy operations. A synchronously generated policy audit log records the node coordinates and parameter configurations, providing a foundation for operation traceability.

[0143] Multiple optimized resource modules deploy a dual-channel execution architecture: the optimization algorithm executor allocates resources to run the atomic computation graph, while the parallel simulation environment replicates input data to rehearse resource solutions. The arbitrator calculates the Euclidean distance between simulated and actual metrics, activating the dual closed-loop feedback module when dynamic thresholds are continuously exceeded. The data feedback loop converts device execution delays into weight correction parameters, updating the weights of knowledge base mapping relationships. The control feedback loop adjusts the resource allocation threshold based on simulation confidence level signals, achieving a dynamic balance between resource efficiency and privacy.

[0144] Containerized encapsulation ensures stable operation of computing units. The policy consistency engine dynamically resolves configuration parameters corresponding to collaborative policy identifiers and synchronizes them in real time to the noise injection node and encryption library. Configuration file digital signatures are verified during container initialization, and policy changes are updated dynamically during runtime. A circuit breaker mechanism freezes the container and launches a backup instance if signature deviations exceed the specified limit. The log service continuously outputs operational status to the security instruction generation module.

[0145] Security instruction generation builds a verifiable privacy protection chain. The reverse verification unit reconstructs quantified data irreversibility through feature reconstruction, triggering container circuit breakers when deviations exceed standards. The audit encapsulation unit binds the irreversibility verification signal to the policy audit log, digitally signs and generates an audit package that is attached to the device control instruction. The final instruction carries a full-link operation traceability chain, meeting legal compliance requirements.

[0146] Dual closed-loop feedback forms a global optimization network. Weight correction parameters update the knowledge base policy priorities and calibrate the noise injection amplitude. Sensitivity parameters adjust resource allocation strategies and deviation detection responses. The parameter transfer mechanism enables real-time linkage between decision-making units and execution units, enabling the system to continuously adapt to environmental changes and maintain an optimal balance between algorithm efficiency and privacy protection in cross-scenario data fusion.

Claims

1. A data fusion intelligent equipment linkage management and control system based on a large model, characterized by: include: The data collection module collects the original data stream of the terminal device and outputs standardized data with privacy classification identifiers; The data processing module receives the environmental parameters in the standardized data, queries the dynamic knowledge base and outputs the synchronously bound optimization algorithm identifier and collaborative strategy identifier; The privacy algorithm fusion scheduling module parses the synchronously bound optimization algorithm identifier and collaborative strategy identifier, compiles the privacy operations corresponding to the collaborative strategy identifier into subgraph nodes of the pre-trained algorithm calculation graph, generates atomic calculation graph data of the fused privacy operations, and outputs the atomic calculation graph data and associated policy audit logs. Multiple resource optimization modules allocate computing resources to execute atomic computational graph data while driving a parallel simulation environment to preview resource plans. When the arbitrator detects that the deviation between simulation indicators and actual execution indicators exceeds a dynamic threshold three times in a row, the dual closed-loop feedback module is activated. The security instruction generation module receives the output results of the atomic computation graph data and the policy audit log output by the privacy algorithm fusion scheduling module, appends the policy audit log to the output results to generate data to be signed, and executes the digital signature on the signed data to generate device control instructions; The dual closed-loop feedback module receives the delay indicators generated by the device executing the control instructions, updates the dynamic knowledge base weight parameters through the data feedback loop, and adjusts the resource allocation threshold according to the simulation confidence level through the control feedback loop.

2. The data fusion intelligent device linkage management and control system based on large models according to claim 1 is characterized in that: The data acquisition module includes a protocol adaptation layer and an identifier injection unit; The protocol adaptation layer converts the terminal device communication protocol into a unified data format and outputs pre-processed data carrying the device type code; The identifier injection unit receives the pre-processed data, indexes the privacy classification rule base based on the device type encoding, and generates a matching privacy classification identifier to be embedded in the data header; The privacy classification rule base stores the mapping between device types and privacy levels and dynamically responds to indexing requests; The identifier injection unit outputs the standardized data integrating the privacy graded identifier to the data processing module; The protocol adaptation layer processes the original data stream to generate pre-processed data and outputs it to the identifier injection unit; The identifier injection unit extracts the device type code from the pre-processed data and calls the privacy classification rule library; The identifier injection unit embeds the generated privacy-graded identifier into the preprocessed data, generates standardized data and outputs it to the data processing module.

3. The data fusion intelligent device linkage management and control system based on large models according to claim 1 is characterized in that: Also includes: The data processing module is equipped with a hierarchical feature extraction engine; The first processing layer performs lighting environment analysis: it inputs a video frame in normalized data, extracts the brightness histogram, calculates the 95th percentile brightness value, and outputs a brightness scalar; The second processing layer performs dynamic target density analysis: it inputs normalized data, applies background modeling and morphological filtering, counts the number of moving objects per square meter, and outputs a density scalar. The feature fusion unit receives the brightness scalar and the density scalar, queries the dynamic weighting coefficient according to the privacy classification identifier, and calculates the composite feature vector; The dynamic knowledge base matches the composite feature vector with the pre-stored scenario strategy mapping relationship, and outputs the bound optimization algorithm identifier and collaborative strategy identifier to the privacy algorithm fusion scheduling module; Standardized data is input into the first processing layer and the second processing layer; The first processing layer outputs the brightness scalar to the feature fusion unit; The second processing layer outputs the density scalar to the feature fusion unit; The feature fusion unit receives the privacy classification identifier, dynamically calculates the weighted value, and fuses the brightness scalar and the density scalar to generate a composite feature vector.

4. The data fusion intelligent device linkage management and control system based on large models according to claim 3 is characterized in that: The dynamic knowledge base includes a policy index layer, a weight storage layer, and a policy update unit; The strategy index layer stores the mapping relationship between the composite feature vector and the bound optimization algorithm identifier and the collaborative strategy identifier; The weight storage layer records the historical performance indicators of each feature vector in the mapping relationship, including response delay and recognition accuracy; The strategy update unit receives the execution delay index transmitted by the data feedback loop and converts the delay index into a weight correction coefficient; The policy update unit receives the confidence level signal transmitted by the control feedback loop and dynamically adjusts the update frequency according to three levels: high / medium / low; The strategy update unit integrates the weight correction coefficient and the update frequency parameter, and outputs the update instruction to the strategy index layer to adjust the mapping relationship; The policy update unit synchronously outputs update instructions to the weight storage layer to calibrate historical performance indicators. The weight storage layer outputs the latest historical performance indicators to the policy index layer to optimize real-time matching accuracy. The data feedback loop transmits the execution delay indicator input to the policy update unit. The control feedback loop transmits the confidence grading signal input to the policy update unit. The policy update unit outputs update instructions to the policy index layer. The policy update unit outputs update instructions to the weight storage layer. The weight storage layer transmits the latest historical performance indicators to the policy index layer. The policy index layer outputs the optimized bound optimization algorithm identifier and collaborative strategy identifier to the privacy algorithm fusion scheduling module.

5. The data fusion intelligent device linkage management and control system based on large models according to claim 1 is characterized in that: The privacy algorithm fusion scheduling module is also used to: Receive the synchronization binding optimization algorithm identifier and the collaboration strategy identifier output by the data processing module; Parse the optimization algorithm identifier to locate the output node position of the feature extraction layer in the pre-trained model calculation graph; The privacy rule library corresponding to the index collaborative strategy identifier obtains the noise amplitude parameter, inserts the noise injection node configured with the noise amplitude parameter after the located node position, associates the collaborative strategy identifier to dynamically link the privacy encryption dynamic library to the noise injection node, compiles the computational graph embedded with the noise node and the encryption library to generate atomic computational graph data, and simultaneously generates a policy audit log that records the node positioning, noise parameter configuration, and encryption library link operation details, outputs the atomic computational graph data to the multiple optimization resource module, and outputs the policy audit log to the security instruction generation module.

6. The data fusion intelligent device linkage management and control system based on large models according to claim 5 is characterized in that: Also includes: Atomized computation graph data encapsulation in independent containers; The container integrates a policy consistency engine to perform the following collaborative operations: Load the algorithm executable file that integrates privacy operations; Dynamically parse the configuration file corresponding to the collaborative strategy identifier and generate the noise amplitude and encryption library link parameters in real time; Synchronously start the logging service to continuously capture the running status of the noise injection node and the encryption library; Automatically verify the matching between the configuration file digital signature and the registration signature of the collaborative policy identifier when the container is started; Synchronize the collaborative strategy identifier changes to the noise injection node and encryption library link in real time during runtime.

7. The data fusion intelligent device linkage management and control system based on large models according to claim 1 is characterized in that: The container internal deployment policy-driven engine is also used for: Load the algorithm executable file that integrates the noise node and the encryption library, parse the configuration file corresponding to the collaborative strategy identifier, and generate the noise amplitude parameters and encryption library link instructions; Start the logging service to capture the noise node output data and the cryptographic library operation status; During the container initialization phase, the hash value deviation between the configuration file digital signature and the registration signature of the collaborative policy identifier is verified. If the deviation is greater than 5%, a circuit breaker alarm is triggered. During the container operation phase, monitor the collaborative policy identifier change event and reconfigure the noise amplitude parameters and the encryption library link within ≤50 milliseconds; The integrity of the configuration file is periodically verified. When tampering is detected, the container is frozen and a backup instance is started. The logging service outputs noise indicators and encrypted audit records to the security instruction generation module in real time.

8. The data fusion intelligent device linkage management and control system based on large models according to claim 7 is characterized in that: Also includes: The arbitrator calculates the Euclidean distance between the simulation indicator and the actual execution indicator. When the distance exceeds the dynamic threshold three times in a row, the dual closed-loop feedback module is activated and a fuse trigger signal is sent to the container. After receiving the circuit breaker signal, the container freezes the currently running instance, starts a clean backup container to load the atomic computation graph data, records the container switching delay time, and feeds the switching delay indicator back to the dual closed-loop feedback module. The control feedback loop adjusts the dynamic threshold based on the simulation confidence level signal. The threshold for high-confidence scenarios is set to 80% of the dynamic threshold baseline value, and the threshold for low-confidence scenarios is set to 120% of the dynamic threshold baseline value. The dual closed-loop feedback module calibrates the generation logic of the confidence grading signal based on the container switching delay index. When the switching delay is greater than 50 milliseconds, the simulation confidence grading signal is automatically downgraded by one level.

9. The data fusion intelligent device linkage management and control system based on large models according to claim 1 is characterized in that: The security instruction generation module includes a reverse verification unit and an audit encapsulation unit; The reverse verification unit receives the feature vector output by the atomic computation graph data, performs feature reconstruction to generate reconstructed data, and calculates the deviation between the feature vector and the reconstructed data; When the deviation value is less than 5%, the irreversibility verification pass signal is output; When the deviation value is ≥5%, a fuse trigger signal is output to the container instance of claim 7; The audit encapsulation unit receives the irreversibility verification pass signal output by the reverse verification unit, receives the policy audit log output by the privacy algorithm fusion scheduling module, binds the irreversibility verification pass signal with the policy audit log to generate a digitally signed audit package, encapsulates the device control instruction and the digitally signed audit package to generate the final control instruction, and the final control instruction carries the irreversibility verification pass signal and the operation audit traceability chain.

10. The data fusion intelligent device linkage management and control system based on large models according to claim 1 is characterized in that: Also includes: The data feedback loop receives the delay indicator generated by the device executing the control instruction, converts the delay indicator into the dynamic knowledge base weight correction parameter, and transmits the weight correction parameter to the policy update unit of the dynamic knowledge base and the container policy driving engine; A control feedback loop receives the confidence level signal output by the parallel simulation environment, dynamically adjusts the arbitrator sensitivity parameters according to the three levels of high confidence, medium confidence, and low confidence, and transmits the sensitivity parameters to the multiple optimization resource modules and the arbitrator; In high-confidence scenarios, the weight correction parameter is tightened to the current value of the dynamic knowledge base weight parameter × 80%, and in low-confidence scenarios, the sensitivity parameter is relaxed to the current value of the arbitrator sensitivity parameter × 120%; The policy update unit updates the knowledge base mapping relationship based on the weight correction parameters. The container policy driving engine calibrates the noise injection amplitude based on the weight correction parameters. The multiple optimization resource module adjusts the resource allocation strategy based on the sensitivity parameters. The arbitrator dynamically adjusts the deviation detection response speed based on the sensitivity parameters.

Images