A local closed-loop dynamic decoupling data processing method

Abstract

This invention discloses a local closed-loop dynamic decoupled data processing method, belonging to the field of local intelligent data processing technology. The invention adopts a layered decoupled architecture, setting up independent precise data processing units and fuzzy data processing units. It relies on a global data unit to achieve autonomous data distribution, rule storage, and weight control. Through result aggregation, iterative optimization, and rule writing back, a local operating closed loop is formed, enabling dynamic checks and balances and autonomous evolution between the two processing paths. All core system operations, data storage, and iteration processes are completed locally, without uploading data to the cloud. Network outages do not affect core functions. Simultaneously, the highest level of human access is set, limiting system iteration to a controllable range. This invention solves the problems of high coupling, cloud dependence, and inability to adaptively evolve in traditional intelligent system architectures. The architecture has strong versatility and scalability, and can be applied to scenarios such as smart homes, embedded intelligence, and offline data processing.

Description

Technical Field

[0001] This invention relates to the fields of local intelligent data processing, embedded intelligent architecture, and offline autonomous decision-making technology, specifically to a local closed-loop dynamic decoupling data processing method. Background Technology

[0002] Current mainstream intelligent processing systems and smart home architectures generally suffer from two major flaws: First, the overall system adopts a unified, coupled architecture, where deterministic instruction processing and non-deterministic semantic reasoning share the same core resources and computational links. This results in rigid module division of labor, high coupling, and an inability to adaptively adjust processing weights based on operational status, leading to poor system flexibility and adaptability. Second, existing intelligent systems heavily rely on cloud servers for data processing, rule iteration, and model optimization, with local systems only handling simple data collection and instruction forwarding. In the event of network interruptions, cloud service anomalies, or network latency fluctuations, the system's core decision-making, data iteration, and intelligent control functions will directly fail, preventing fully localized and autonomous closed-loop operation.

[0003] Meanwhile, the system resource scheduling and algorithm priority adjustment in the existing technology are all parameter fine-tuning, thread ratio and resource allocation adjustment within the integrated kernel. They are only surface-level resource optimization methods and do not achieve functional layering and decoupling at the architectural level. There is no underlying architectural mechanism of autonomous evolution and dynamic balance. The system has no self-iterative optimization capability, and its intelligence level cannot be improved autonomously in the long run, which has strong technical limitations. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a local closed-loop dynamic decoupled data processing method. This method solves the technical problems of high coupling, inability to dynamically balance and iterate, reliance on the cloud, and weak local autonomous operation capabilities in existing intelligent system architectures. The core innovation of this invention lies in constructing a layered and decoupled dual-path processing architecture. It relies on global data units to achieve local closed-loop self-iteration and dynamic weight balancing, without relying on cloud-based core computing throughout the process, thus achieving architecture-level autonomous evolution and stable operation.

[0005] The core technical solution of this invention is as follows: A locally closed-loop dynamically decoupled data processing method, the core architecture of which consists of functional units and corresponding data links, is a purely locally closed-loop, layered decoupled, and dynamically self-evolving architecture, specifically including: The data input unit is used to receive and summarize external data to be processed. The global data unit is used to store the system's baseline operating rules, data processing logic, historical iteration data, and the capability weight benchmarks of each processing unit, providing a unified benchmark for system traffic distribution, processing, and iteration. The precise data processing unit and the fuzzy data processing unit are set up independently and decoupled from each other, with no fixed computing resource allocation and no architectural binding relationship. They can independently complete the processing of deterministic data instructions and the processing of non-deterministic semantic and complex scene data. The data iteration unit is used to summarize the running results of each processing unit, combine them with global historical data to generate iterative optimization logic, and realize the autonomous updating of system rules. The global data reading link is used to intelligently and autonomously distribute input data according to the baseline rules of global data unit storage, and accurately allocate different types of data to the corresponding processing units. The data processing result aggregation link of the sub-unit is used to collect all the output results of the precise data processing unit and the fuzzy data processing unit and feed them back to the data iteration unit in a unified manner. Iterative data is written to the global link to write back the optimized and updated running rules and weight ratio parameters of the data iteration unit to the global data unit, forming a continuous self-updating closed-loop running mechanism. The on-demand output of processed data units is used to adapt the output format according to the baseline rules of the global data units and complete the final data output.

[0006] The system relies on the dynamic update rules of global data units to create a dynamic balance between precise data processing capabilities and fuzzy data processing capabilities, which increase and decrease with each iteration. This eliminates the need for manual adjustment of routine parameters, with the human operator retaining only the highest level of system access, thus constraining the system to evolve autonomously within a controllable range. Simultaneously, all data is stored, processed, and iterated locally in a closed loop throughout the entire process, with no data uploaded to the cloud, thus autonomously constructing a secure, controllable, and self-optimizable local closed-loop data processing architecture.

[0007] The precise data processing unit and the fuzzy data processing unit are a layered, decoupled, and independent architecture with no fixed processing ratio. After continuous iterative learning, the fuzzy processing capability gradually improves, while the global data unit automatically weakens the workload of the precise processing unit, achieving a reverse balance and dynamic self-balancing of dual-path capabilities to adapt to the intelligent processing needs of different operating scenarios.

[0008] All core traffic distribution logic, data processing, iterative learning, and rule update logic of the system are completed locally in a closed loop, without relying on real-time computing and control in the cloud. External data access is an optional extension function. When the network is abnormal or external data is disconnected, it will not affect the normal operation of the core system architecture, and it has a strong offline autonomous operation capability. Beneficial effects

[0009] Compared with the prior art, the present invention has the following key beneficial effects: 1. Achieve layered decoupling at the architectural level, breaking through the limitations of traditional integrated coupling. This invention sets up independent precise and fuzzy dual processing paths, breaking the integrated kernel coupling architecture of existing intelligent systems. The two types of processing units operate independently and perform their respective functions, which can fundamentally avoid resource contention and logical interference, enabling the system to balance high reliability and high intelligence.

[0010] 2. Native dynamic checks and balances, localized autonomous evolution. This invention relies on a local closed-loop iterative mechanism to achieve dynamic adaptive adjustment of the dual-path capabilities, and autonomously completes rule optimization and weight iteration upgrades based on local operating data. It does not rely on external devices or cloud algorithm control, forming an architecture-level autonomous growth capability, which is different from the traditional static fixed scheduling architecture.

[0011] 3. Purely local closed-loop operation with extremely high offline reliability. The system's core data processing, iterative learning, and rule updates are all completed locally, eliminating dependence on cloud servers. When external network anomalies or cloud service failures occur, the core intelligent functions remain unaffected, forming a highly stable, secure, and privacy-protecting local intelligent form.

[0012] 4. Strong architectural versatility and high scalability. The core principle architecture of this invention is not tied to any specific scenario, hardware, or software, and can be adapted to various scenarios such as smart homes, local embedded intelligence, and offline data processing. The architecture is simple and universal, with ample room for iteration and expansion, and a long technology lifecycle. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the overall core principle architecture of the present invention. It is a pure principle architecture that is the core protection of the present invention, without any scenario, hardware, or device limitations. Figure 2 This is a schematic diagram of a scenario-based architecture for a smart home system implemented based on the principle architecture of this invention, and serves as an exemplary application extension of the principle architecture; Figure 3 This is a schematic diagram of the minimum closed-loop prototype and interface expansion built based on the principle architecture of this invention. All software and hardware devices are general exemplary selections and can be replaced equivalently, and do not constitute a limitation on the scope of protection of this invention.

[0014] Figure 1 Principle architecture diagram annotation 1-Data input unit; 2-Global data unit; 3-Precise data processing unit; 4-Fuzzy data processing unit; 5-Data iteration unit; 6-Global data reading link; 7-Sub-unit data processing result summary link; 8-Iterative data writing to global data unit link; 9-Processed data output unit as needed.

[0015] Figure 2 System architecture diagram annotations 1-Multimodal information unit; 2-Information translation unit; 3-Spinal reflex module; 4-Core reasoning module; 5-Cortical learning module; 6-Global data module; 7-External data interface; 8-Deterministic command direct output link; 9-Data splitting unit; 10-Interactive device unit; 11-Home appliance unit. This embodiment is a scenario-based adaptation of the principle architecture. The scenario-based terminology used is only illustrative and does not limit the scope of protection of this invention.

[0016] Figure 3 Explanation of Prototype Expansion Diagram 1-ESP32 voice module (prototype connected); 2-ESP32 camera module (expansion interface); 3-VOSK voice stream model (prototype connected); 4-PYTHON library (prototype connected); 5-PYTHON dictionary (prototype connected); 6-OLLAMA QWEN model (prototype connected); 7-OPEN CLAW model (prototype connected); 8-Time / Weather API (prototype connected / on-demand data collection / disconnection does not affect core functions); 9-Accurate data direct output (prototype via WIFI); 10-Output data distribution (distributes data to HomeAssistant virtual / Xiaomi gateway via internal network to distribute Bluetooth switch commands, prototype connected); 11-TTS voice model / ESP32 amplifier module (prototype connected); 12-Display board (expansion interface); 13-Bluetooth bulb (prototype connected); 14-Other home appliances (expansion interface).

[0017] It should be noted that the prototype shown in Figure 3 relies on a computer with general computing capabilities as its control platform. This platform is equipped with a processor, memory, and a graphics processor for accelerating AI model inference. The remaining hardware, software, models, and gateway devices are all existing general-purpose technology products, used only to verify the feasibility of the core principle architecture of this invention. They are exemplary implementation carriers and can be arbitrarily replaced with equivalent components. This application does not claim any rights to any hardware, software, model, or device; the scope of protection is limited to the core principle architecture. Detailed Implementation

[0018] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings. These embodiments are implemented under the premise of the technical solution of the present invention, and a detailed implementation process and architecture adaptation logic are given. However, the protection scope of the present invention is not limited to the following embodiments.

[0019] 1. Implementation of core principle architecture The core of this invention is a purely local, layered, decoupled, and dynamically balanced closed-loop data processing architecture. It has no hardware or scenario-specific constraints, making it universally adaptable to various local intelligent data processing scenarios. The specific operational logic is as follows: After receiving various types of external data to be processed, the data input unit triggers the global data reading link to retrieve the baseline diversion rules, weight configurations, and processing logic stored within the global data unit. Based on data attributes, types, and instruction characteristics, the system autonomously diverts data to either the precise data processing unit or the fuzzy data processing unit. Deterministic structured data enters the precise data processing unit for high-speed and accurate processing, while unstructured, fuzzy semantic, and complex scenario data enters the fuzzy data processing unit for deep inference processing.

[0020] The precise data processing unit and the fuzzy data processing unit operate independently and decoupled, without interfering with each other or consuming core resources, and without any fixed workload ratio. After completing data operations, the two types of processing units aggregate the results through a unified chain, and all results are fed back to the data iteration unit.

[0021] The data iteration unit combines historical iteration data and the current processing results to autonomously optimize and update the system's traffic distribution rules and dual-path processing weight ratio, generating new system operating parameters. The optimized iteration data is written to the global link through the iteration data and then written back to the global data unit to complete the autonomous update loop of the system rules.

[0022] The entire iterative update process requires no manual parameter adjustment. Human intervention is limited to the highest level of system access, restricting the system's autonomous iteration and updates within manually set controllable rules, thus preventing boundless evolution. Simultaneously, there is no cloud data upload, synchronization, or parameter interaction. All operational, iterative, and user data are stored and processed locally in a closed loop. Each data processing iteration accumulates local operational experience, continuously optimizing its processing logic. As the number of iterations increases, the fuzzy data processing unit's ability to adapt to complex scenarios and its semantic understanding capabilities continuously improve. The global data unit automatically lowers the workload weight of the precise data processing unit, creating a dynamic balance between the two pathways, allowing the overall system intelligence to continuously and autonomously upgrade.

[0023] The final processed data is then output by the on-demand output data unit according to global rules, completing a single closed-loop operation.

[0024] 2. Implementation of Scenario-based System Architecture Based on the core principle architecture of this invention, it can be adapted and applied to local intelligent control system scenarios in smart homes. The core principle is only adapted for specific scenarios. The multimodal information unit collects various types of home data, including voice, environment, and device status. The data is then normalized by the information translation unit. Based on the home control rules and user habit data stored in the global data module, deterministic appliance switching and numerical adjustment commands are routed to the spinal reflex module (corresponding to the precision data processing unit) to achieve high-speed, low-latency, and precise device control. User fuzzy semantics and complex scene linkage requirements are routed to the core reasoning module (corresponding to the fuzzy data processing unit) to complete semantic parsing and scene reasoning. The control commands generated by the reasoning are uniformly issued and executed.

[0025] The cortical learning module (corresponding to the data iteration unit) summarizes all home control results and user operation feedback, iteratively optimizes user habit recognition rules and command routing weights, and continuously updates the operating logic stored in the global data module.

[0026] The system supports external data interfaces for on-demand network data access, and core smart home control and iterative optimization functions operate normally even in the event of a network outage. Deterministic commands can be executed directly, while complex commands are distributed to interactive devices and home appliances via a data distribution unit to achieve scenario-based intelligent output.

[0027] 3. Prototype Verification Implementation Instructions To verify the feasibility and practical effectiveness of the core principle architecture of this invention, a minimum closed-loop verification prototype was built, and various existing general-purpose hardware and software devices were used to complete the architecture implementation verification. It should be noted that all hardware modules, software models, and gateway systems used in the prototype are existing, well-known, and general-purpose technologies, serving only as carriers for architecture verification and can be arbitrarily replaced with equivalent ones. This invention does not protect any prototype hardware, software, or combination methods, but only the top-level principle architecture and dynamic balance closed-loop mechanism.

[0028] The prototype's operation fully follows the core principles and architecture of this invention. Voice signals are acquired by the ESP32 voice module and translated by the VOSK model to complete data input. Precise processing of deterministic data is achieved through a Python dictionary and Python libraries, while fuzzy semantic reasoning is realized through the OLLAMA QWEN model. The OPEN CLAW model, as the core carrier of cortical learning, is specifically responsible for collecting and summarizing feedback data from the entire process, providing core data support for rule iteration and weight updates of global data units, enabling the system's autonomous learning and evolution. Data aggregation, iterative optimization, and rule solidification updates are completed through local programs, achieving dynamic checks and balances across dual pathways. Finally, command distribution is completed through the local network, Xiaomi gateway, and Home Assistant system to drive devices such as Bluetooth bulbs and TTS amplifiers. The system reserves expansion interfaces such as the ESP32 camera module and display dashboard, allowing for the expansion of multimodal acquisition and visualization functions as needed. The time / weather API is an optional expansion function, and disconnection does not affect the core closed-loop operation.

Claims

1. A local closed-loop dynamic decoupling data processing method, characterized in that, include: The data input unit is used to receive external data to be processed. The global data unit is used to store system baseline rules, data processing logic, and historical iteration data. The precise data processing unit and the fuzzy data processing unit are set up independently and decoupled from each other, and are used to complete deterministic data processing and non-deterministic data processing respectively; The data iteration unit is used to summarize the processing results and generate iterative optimization logic. The global data reading link is used to autonomously distribute the input data to the corresponding processing unit according to the baseline rules of the global data unit. The data processing result aggregation link for sub-units is used to collect the output results of precise processing and fuzzy processing. Iterative data is written to the global link to update the global data unit with the iteratively optimized rules, forming a closed-loop self-updating mechanism. The processed data units are output on demand to adapt the output format according to global rules. The system relies on the dynamic rules of global data units to create a dynamic balance between precise processing capabilities and fuzzy processing capabilities, which evolves and diminishes with each iteration, thus constructing an autonomous, locally closed-loop data processing architecture.

2. The local closed-loop dynamic decoupling data processing method according to claim 1, characterized in that: The precise data processing unit and the fuzzy data processing unit are a hierarchical, decoupled, and independent architecture with no fixed processing ratio. After iterative learning, the system automatically weakens the workload of precise processing as the fuzzy processing capability improves, thus achieving a reverse balance and dynamic self-balancing of dual-path capabilities.

3. The local closed-loop dynamic decoupling data processing method according to claim 1, characterized in that: All core traffic distribution, processing, iteration, and rule update logic of the system are completed locally in a closed loop, without relying on real-time cloud computing. External data access is an optional extension and does not affect the main operation of the architecture.

Images