An environment electrical appliance native control system, method and device for artificial intelligence agent and storage medium

Abstract

This invention relates to a native control system and method for environmental appliances oriented towards artificial intelligence agents. The system includes: an AI agent interface module for receiving control tasks from the AI ​​agent; a capability registration and management module for abstracting the various functions of the environmental appliances into capability objects; a capability parsing module for parsing the syntax and semantics of the capability objects; a tool mapping module for automatically mapping capability objects to tool interfaces callable by the AI ​​agent; an execution control module for generating execution plans; a status feedback module for receiving the device's execution status and returning it to the AI ​​agent; a dynamic capability update module for implementing incremental updates, hot updates, and rollbacks of capability objects; and a storage and logging module for storing information. This invention enables environmental appliances to possess structured capability description capabilities, intent-driven control capabilities, a safety-level execution mechanism, rollback and auditability mechanisms, and compatibility with AI agent tool call interfaces.

Description

Technical Field

[0001] This invention belongs to the field of intelligent home appliance control technology, and in particular relates to a native control system, method, device and storage medium for environmental appliances oriented towards artificial intelligence agents. Background Technology

[0002] Current environmental appliances (including air purifiers, humidifiers, dehumidifiers, and fresh air systems) are typically controlled via physical buttons, mobile applications, or rule-based automatic modes. Although some devices support network interface calls, their control methods are still based on fixed API commands or preset scene logic.

[0003] With the development of large-scale models and AI agent technology, AI agents are able to understand natural language and generate control commands. However, existing technologies have the following problems: 1. The equipment capabilities are not subjected to structured semantic abstraction, making it difficult for AI agents to understand the boundaries of the equipment capabilities.

[0004] 2. There are physical security risks associated with AI agents directly outputting low-level control commands.

[0005] 3. Lacking an intent-level semantic orchestration layer, it is impossible to achieve secure mapping from semantics to physical actions.

[0006] 4. Lack of risk classification, idempotent control, and rollback mechanism.

[0007] 5. Lack of auditable execution chain records.

[0008] Therefore, there is an urgent need for an environmental electrical control architecture designed natively for artificial intelligence agents. Summary of the Invention

[0009] The purpose of this invention is to provide a native control system and method for environmental appliances oriented towards artificial intelligence agents, enabling environmental appliances to have the ability to describe capabilities in a structured manner, the ability to drive control by intent, a safety-level execution mechanism, a rollback and auditability mechanism, and compatibility with the tool calling interface of artificial intelligence agents. It is applicable to various environmental electrical devices such as air purifiers, humidifiers, dehumidifiers, and air quality monitors.

[0010] This invention provides a native environmental electrical control system for artificial intelligence agents, comprising: The AI ​​agent interface module is used to receive control tasks from the AI ​​agent and return the execution status to the AI ​​agent; The capability registration and management module is used to abstract the various functions of environmental electrical appliances into capability objects, and supports dynamic registration, updating and version management; The capability parsing module is used to parse the syntax and semantics of capability objects, generate capability dependency graphs, and detect capability conflicts. The tool mapping module is used to automatically map capability objects to tool interfaces that can be called by the AI ​​agent, and generate parameter structure definitions (Schema) and callback interfaces; The execution control module is used to generate execution plans based on task dependencies, and to perform topology sorting, risk-level execution, conflict resolution, and redundancy optimization for execution on multiple devices. The status feedback module is used to receive the device execution status and return it to the AI ​​entity, while also performing anomaly monitoring and multi-device status aggregation. The dynamic capability update module is used to implement incremental updates, hot updates, and rollbacks of capability objects, ensuring cross-device capability synchronization. The storage and logging module is used to store capability objects, execution plans, device status, and exception logs.

[0011] Furthermore, the execution control module dynamically adjusts the task execution order through a risk-based execution strategy to ensure that high-risk operations are constrained, while low-risk operations are executed first.

[0012] Furthermore, the capability registration and management module supports unified capability abstraction for environmental appliances across brands and models, forming a standardized interface protocol.

[0013] Furthermore, the tool mapping module supports zero-code capability invocation, which is used to automatically map newly added capability objects to tool interfaces that can be invoked by the artificial intelligence agent.

[0014] Furthermore, the dynamic capability update module works in conjunction with the storage and log module to enable hot updates of capability objects, version rollback, and traceable storage of historical execution data.

[0015] Furthermore, the environmental appliances include air purifiers, humidifiers, dehumidifiers, and other home or office environmental control devices, and each device is connected to the system through a unified capability abstraction interface to achieve multi-device collaborative control.

[0016] This invention also provides a native control method for environmental electrical appliances for artificial intelligence agents, comprising the following steps: Step 1: Receive control tasks from the AI ​​agent; Step 2: Register the functions of environmental electrical appliances as capability objects, and perform dynamic management and version control; Step 3: Perform syntactic and semantic parsing on the capability objects to generate a capability dependency graph and detect conflicts; Step 4: Map the capability object to a tool interface that the AI ​​agent can call, and generate the parameter structure definition (Schema) and callback interface; Step 5: Generate an execution plan based on task dependencies, and control the execution of tasks through risk-based execution, multi-device collaboration, conflict resolution, and redundancy optimization. Step 6: Collect device execution status and abnormal events, return the results to the AI ​​agent, and perform multi-device status aggregation; Step 7: Perform incremental updates, hot updates, or rollbacks on the capability objects to ensure cross-device synchronization, and store the relevant execution data and logs to the storage medium.

[0017] Furthermore, step 5 includes: dynamically adjusting the task execution priority according to the risk level, with low-risk tasks executed first and high-risk tasks executed in a controlled manner.

[0018] The present invention also provides a native control device for environmental appliances for artificial intelligence agents, including a processor, a memory and a program module connected to the processor, the program module being configured to execute the method described.

[0019] The present invention also provides a non-volatile storage medium storing computer-executable instructions for causing a computer to perform the method described herein.

[0020] By employing the above-described solution, the environmental electrical native control system, method, apparatus, and storage medium for artificial intelligence agents achieve the following technical effects: 1) AI-native control enables zero-code capability invocation: This invention uses a capability abstraction and registration module and a tool mapping module to abstract all functions of environmental appliances into capability objects, which can be directly invoked by the artificial intelligence agent (AI Agent) without additional programming. This can significantly reduce system integration costs, improve the efficiency of AI-hardware interaction, and support personalized scene configuration for users. AI can autonomously complete environmental control tasks.

[0021] 2) Multi-device collaboration and dynamic aggregation execution: The execution control module supports collaborative scheduling, redundancy optimization, and topology sorting scheduling across multiple devices, enabling multiple environmental appliances to execute safely and reliably according to task dependencies. This improves the accuracy and response speed of environmental control, while achieving optimal utilization of equipment resources. In complex scenarios, such as simultaneously controlling air purifiers, humidifiers, and dehumidifiers, the system can automatically optimize the operation sequence to ensure safe and efficient execution.

[0022] 3) Risk-based execution and safety assurance: This invention proposes a risk-based execution mechanism with levels L0 to L4. Through the execution control module and the status feedback module, the execution of tasks is dynamically monitored and risk assessed. The execution strategy can be adjusted according to the risk level of the task to achieve safe and controllable operation of electrical appliances in the environment. This avoids the direct execution of high-risk operations and ensures user safety and equipment stability.

[0023] 4) Capability hot update and version controllability: The dynamic capability update module supports incremental updates, hot updates and rollback operations for capability objects, ensuring that the system can be upgraded and iterated without interrupting operation. The device functions can be continuously expanded with AI capabilities and scenario requirements, making it highly adaptable. When new sensors are introduced or new control strategies are added, the system can be upgraded without stopping, ensuring long-term availability and continuous system evolution.

[0024] 5) Closed-loop feedback and status traceability: The status feedback module is combined with the storage log module to realize closed-loop feedback and historical record storage of equipment status, execution results and abnormal events, realize traceable and auditable environmental electrical control, facilitate abnormal event analysis and optimization of system strategies, and allow users or maintenance personnel to view equipment operation history and AI decision trajectory, thereby improving system transparency and maintainability.

[0025] 6) Unified and standardized cross-device capabilities: Through capability abstraction and registration mechanisms, the capabilities of environmental appliances of different brands and models are standardized to form a unified interface protocol, which can enhance the system's compatibility and scalability, and lay a standard foundation for the integration of future smart home ecosystems and artificial intelligence agents (AI agents). Air purifiers, humidifiers and dehumidifiers from different manufacturers can all be controlled in a unified manner through the same artificial intelligence agent (AI agent), reducing secondary development costs.

[0026] 7) Intelligent scene optimization and adaptive control: By combining the strategy generation and tool interface call of the artificial intelligence agent, this invention can autonomously optimize the control strategy according to user needs, environmental conditions and historical data, realize adaptive scene control, improve user experience and equipment efficiency, and reduce energy waste. For example, it can automatically adjust air purification and humidity when the user leaves the room, and the environment has been optimized when the user returns, thus achieving intelligent automatic management.

[0027] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0028] Figure 1 This is a structural diagram of the native environmental electrical control system for artificial intelligence agents according to the present invention; Figure 2 This is a schematic diagram of the system function implementation process of the present invention; Figure 3 This is a flowchart of the native environmental electrical control method for artificial intelligence agents according to the present invention; Figure 4 This is a schematic diagram of an embodiment of the control method of the present invention; Figure 5Schematic diagram of an embodiment of the control system and storage medium of the present invention. Detailed implementation manners

[0029] The following combines the accompanying drawings and embodiments to further describe in detail the specific implementation manners of the present invention. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0030] Refer Figure 1 As shown, this embodiment provides a native control system for environmental appliances for an artificial intelligence agent, including: An artificial intelligence agent interface module 10, configured to receive control tasks from the artificial intelligence agent and return the execution status to the artificial intelligence agent.

[0031] A capability registration and management module 20, configured to abstract various functions of environmental appliances into capability objects, and support dynamic registration, update, and version management. It includes: structurally describing the functions of electrical equipment, including capability ID, parameter definition, status variables, execution constraints, dependency relationships, and risk levels; outputting standardized capability data objects. Responsible for registering device capabilities with the system to generate a unique capability index; supporting dynamic registration and version management.

[0032] A capability parsing module 30, configured to parse the syntax and semantics of capability objects, generate a capability dependency graph, and detect capability conflicts. It includes: parsing the capability data object into an internal model understandable by AI; constructing a dependency graph and a conflict relationship matrix.

[0033] A tool mapping module 40, configured to automatically map capability objects to tool interfaces callable by the artificial intelligence agent, and generate a parameter structure definition (Schema) and a callback interface. It includes: automatically generating an AI tool interface from the parsed capability model; supporting the generation of a function call structure definition (Schema) and callback interface definition.

[0034] An execution control module 50, configured to generate an execution plan according to the task dependency relationship, perform topological sorting on multi-device execution, risk-graded execution ( ), conflict resolution, and redundancy optimization. It includes: generating an execution plan according to an AI instruction; verifying the risk level and execution constraints; scheduling the device interface to execute an operation; performing risk-graded control on the operation; blocking, recording, or manually confirming an abnormal operation.

[0035] A status feedback module 60, configured to receive the device execution status and return it to the artificial intelligence agent, and at the same time perform abnormal monitoring and multi-device status aggregation. It includes: real-time transmitting the device status and operation results; supporting abnormal monitoring and an event triggering mechanism.

[0036] The dynamic capability update module 70 is used to implement incremental updates, hot updates, and rollbacks of capability objects, ensuring cross-device capability synchronization and guaranteeing consistency in multi-device collaborative scenarios.

[0037] The storage and logging module 80 is used to store capacity objects, execution plans, device status, and exception logs.

[0038] In this embodiment, the execution control module 50 dynamically adjusts the task execution order through a risk-based execution strategy to ensure that high-risk operations are constrained, while low-risk operations are executed first.

[0039] In this embodiment, the capability registration and management module 20 supports unified capability abstraction for environmental electrical appliances across brands and models, forming a standardized interface protocol.

[0040] In this embodiment, the tool mapping module 40 supports zero-code capability calls, which is used to automatically map newly added capability objects to tool interfaces that can be called by the artificial intelligence agent.

[0041] In this embodiment, the dynamic capability update module 70 and the storage and log module 80 work together to realize hot updates of capability objects, version rollback, and traceable storage of historical execution data.

[0042] In this embodiment, the environmental appliances include air purifiers, humidifiers, dehumidifiers, and other home or office environmental control devices. Each device connects to the system through a unified capability abstraction interface (device interface module) to achieve multi-device collaborative control. The device interface module is used to drive the electrical devices to perform specific functions and supports multiple communication protocols (such as Wi-Fi, ZigBee, Bluetooth, Modbus, etc.).

[0043] The control system will now be described in further detail.

[0044] The core of this system is: 1. Abstract the functions and states of environmental electrical appliances into standardized capability objects; 2. Enable the AI ​​entity to automatically analyze and invoke the capabilities of environmental electrical appliances; 3. Introduce risk level and execution constraint mechanisms to achieve safe and controllable automated control; 4. Supports multi-device collaboration, dynamic updates, and cross-platform calls; 5. Establish a closed-loop control system, including perception, decision-making, execution, and feedback.

[0045] This native control system for environmental appliances is suitable for smart environmental appliances such as air purifiers, humidifiers, dehumidifiers, and air quality monitors. Its overall structure includes: 1. AI Agent Interface Module Function: Enables two-way communication with the AI ​​agent, including task reception, policy invocation, and result feedback; Features: Supports zero-code capability invocation and is compatible with multiple types of AI agents.

[0046] 2. Capability Registration and Management Module Function: Responsible for the dynamic registration, index management, version control, and capability aggregation of capability objects; Features: Supports hot updates, rollbacks, and cross-device synchronization.

[0047] 3. Capability Parser Module Functionality: Performs syntax parsing, semantic mapping, dependency and conflict detection on registration capabilities; Features: Generates internal standardized capability models, providing foundational data for tool mapping and execution.

[0048] 4. Tool Mapping Module Functionality: Maps the capability model to AI tool interfaces, including function signatures, parameter structure definitions (Schema), and callback interfaces; Features: Enables zero-code API calls and supports event-driven capabilities.

[0049] 5. Execution Controller Module Functions: Execution plan generation, risk classification and control, multi-device collaboration, conflict resolution, and redundancy optimization; Features: Supports multi-device aggregation, topology sorting, and dependency scheduling, ensuring safe and reliable execution.

[0050] 6. Device Interface Module Function: Responsible for issuing device commands and protocol adaptation (Wi-Fi, ZigBee, Bluetooth, Modbus, etc.); Features: Compatible with devices from multiple manufacturers, supports asynchronous / synchronous execution modes and sensor status acquisition.

[0051] 7. Status & Feedback Module Function: Collect device status, execution results and abnormal events in real time, and send them back to the AI ​​Agent. Features: Closed-loop control, multi-device status aggregation, and support for log recording and auditing.

[0052] 8. Capability Update Module Functionality: Enables incremental updates, version management, and rollback of capability objects; Features: Supports hot updates without interrupting system operation, ensuring consistency across multiple devices.

[0053] 9. Storage & Audit Module Functions: Stores capability objects, execution plans, status data, logs, and exception records; Features: Supports long-term historical tracing and audit analysis.

[0054] System data flow and control flow: 1. Control Flow: The AI ​​Agent initiates tasks through the interface module → Capability registration / resolution → Tool mapping → Execution control → Issuance of device commands → Execution control module monitors execution → Status feedback module sends back AI in a closed loop.

[0055] 2. Data Flow: The device interface module collects sensor data → the status feedback module aggregates the data → the storage module records the data → the AI ​​Agent interface module can read the data, forming a closed-loop data flow.

[0056] 3. Dynamic Capability Flow: New capability object registration → Capability resolution → Tool mapping → Execution control → Hot update / rollback → Synchronous update across all devices.

[0057] System hardware and software implementation: 1. Hardware Components The main control chip / microcontroller (MCU) or SoC is responsible for module computation and task scheduling; Network communication interfaces: Wi-Fi, ZigBee, Bluetooth, etc.; Memory: EEPROM, NAND / Flash, used for storing capability objects, configurations, and logs; Input / output interfaces: used to connect environmental electrical actuators and sensors.

[0058] 2. Software Components Operating system: Real-time operating system (RTOS) or embedded Linux; Modular software architecture: Each of the above functional modules operates independently and supports hot loading and updating; Communication protocol stack: Implements communication between AI agents and device protocol adaptation; Security and logging module: Ensures the security and auditability of control instructions and status data.

[0059] Parameter Figure 2 As shown, the system function implementation process includes: 1. Ability generation and registration: When the electrical device is powered on or initialized, an ability object is generated.

[0060] Registered to the system through the ability registration module and a unique index is generated.

[0061] 2. Ability parsing: The system analyzes the semantics and parameter structure of the ability object.

[0062] Construct an ability dependency graph and a conflict detection matrix.

[0063] 3. Automatic tool mapping: Map the ability object to an AI tool interface, and automatically generate call functions and parameter structure definitions (Schema).

[0064] 4. AI agent call: The AI agent executes operations according to the policy or task scheduling tool interface.

[0065] Perform risk level check and dependency constraint verification before the call.

[0066] 5. Execution scheduling and device control: The execution control module distributes instructions according to topological sorting and policies.

[0067] The device interface module receives the instructions and executes them.

[0068] 6. Status feedback and closed loop: The execution results and device status are sent back to the AI agent through the status feedback module.

[0069] 7. Abnormal events trigger the risk control mechanism and log record.

[0070] The system functions are as follows: Multi-device collaboration: Illustrated by air purifiers, humidifiers, and dehumidifiers, demonstrating redundancy optimization and aggregation capabilities.

[0071] Risk grading execution: The execution control module reflects The risk level control strategy.

[0072] Dynamic Capability Registration: The capability registration module can dynamically register new devices and new capabilities, and supports version management, hot updates, and rollback.

[0073] Closed-loop control: The status feedback module sends the execution results back to the AI ​​Agent, forming a closed loop.

[0074] Separation of data flow and control flow: Figure 2 The middle arrow indicates that control commands are sent downwards, while status data is transmitted upwards.

[0075] Tool Mapping: The AI ​​call interface is generated by the tool mapping module, supporting zero-code capability calls.

[0076] The system employs a risk-based hierarchical execution mechanism, a multi-device collaborative control mechanism, and a dynamic capability update mechanism, as detailed below: Risk-based Implementation Mechanism: This invention introduces risk level management: Risk levels are tied to capability parameters and operational constraints.

[0077] The execution control module determines the execution strategy based on the risk level.

[0078] The risk level can be dynamically adjusted to adapt to changes in the scenario.

[0079] Multi-device collaborative control mechanism: Capability aggregation: Similar capabilities are aggregated across multiple devices to generate a unified API.

[0080] Redundancy optimization: Multiple devices with the same capabilities can be optimized for execution based on location, power consumption, and status.

[0081] Load balancing: Supports concurrent task allocation to avoid overloading of a single device.

[0082] State synchronization: Ensures that state information is consistent across different devices to avoid conflicting execution.

[0083] Dynamic capability update mechanism: Incremental update: When adding or modifying capabilities of a device, only the changed parts are synchronized.

[0084] Full update: The system supports full synchronization of capabilities to ensure data consistency.

[0085] Hot update: Update capabilities without affecting operation.

[0086] Version rollback: In abnormal situations, it can roll back to the previous capability version.

[0087] System Interfaces and Compatibility: The system-generated AI tool interface supports multiple call protocols, including: OpenAI function call format, JSON-RPC, local Agent framework, and cloud control API.

[0088] Supports cross-platform artificial intelligence agents (AI Agents) calls to ensure device ecosystem compatibility.

[0089] Data flow and control flow design: Data flow: Used to transfer device status, sensor data, and event information.

[0090] Control flow: Used to transfer AI execution instructions and operation plans.

[0091] The dual-channel design ensures control security and auditability.

[0092] Through this technical solution, the system can achieve: 1) Standardized abstraction of electrical equipment capabilities; [[ID=2l]]2) Zero-code automatic call ability of artificial intelligence agents (AI Agents); 3) Risk-graded controllable execution; 4) Multi-device collaborative operation and redundancy optimization; 5) Dynamic ability expansion and version evolution; 6) Cross-platform and cross-vendor compatibility; x7) Data and control closed-loop feedback to ensure safety and reliability.

[0093] See [[ID=3S]] Figure 3 As shown, an environment electrical appliance native control method for artificial intelligence agents includes the following steps: Step S1, receiving a control task sent by an artificial intelligence agent.

[0094] Step S2, registering the functions of the environment electrical appliances as capability objects and performing dynamic management and version control.

[0095] Step S3, parsing the syntax and semantics of the capability objects, generating a capability dependency graph, and detecting conflicts.

[0096] Step S4, mapping the capability objects to tool interfaces callable by artificial intelligence agents, and generating a parameter structure definition (Schema) and callback interfaces Step S5, generating an execution plan according to the task dependency relationship, and executing the control task through risk-graded execution, multi-device collaboration, conflict resolution, and redundancy optimization.

[0097] Step S6, collecting the device execution status and abnormal events, returning the results to the artificial intelligence agent, and aggregating the multi-device status.

[0098] Step S7: Perform incremental updates, hot updates, or rollbacks on the capability objects to ensure cross-device synchronization, and store the relevant execution data and logs to the storage medium.

[0099] Furthermore, step S5 includes: classifying risks according to risk level. The priority of task execution is dynamically adjusted, with low-risk tasks executed first and high-risk tasks executed in a controlled manner.

[0100] The control method will be described in further detail below.

[0101] The core of this method lies in: 1. Abstract the functions of environmental electrical appliances into standardized capability objects; 2. Implement dynamic registration, resolution, and tool mapping of capability objects; 3. Achieve safe and controllable execution through risk classification and dependency constraints; 4. Supports multi-device collaboration, redundancy optimization, and dynamic capability updates; 5. Achieve a closed loop between data flow and control flow to ensure that the status is traceable and auditable.

[0102] This control method is applicable to air purifiers, humidifiers, dehumidifiers, air quality monitors, and other intelligent environmental electrical appliances.

[0103] This control method includes the following steps: (1) Capability abstraction and registration 1. Electrical equipment generates capability objects, which include: Capability ID, name, and function description; Parameter definition and value range; State variables and their readable / writable properties; Dependency relationships and conflict relationships; Risk level designation (L0~L4).

[0104] 2. Capability objects are registered with the system through the capability registration module, generating a unique index.

[0105] 3. After registration, the system will store the capability object in the capability database and establish version management information.

[0106] (2) Ability Analysis 1. The system performs syntactic and semantic parsing on the registered capability objects to generate an internal capability model.

[0107] 2. Construct a dependency graph and conflict detection matrix to mark the calling order and possible conflicts between capabilities.

[0108] 3. Associate the capability model with the tool mapping module to prepare for generating AI callable interfaces.

[0109] (3) Tool interface generation and mapping 1. An interface for automatically generating capability models, including: Function signature Parameter structure definition (Schema) Callback event interface 2. The generated interface is called by the AI ​​Agent, enabling zero-code invocation.

[0110] (4) Execution plan generation and risk classification control 1. The AI ​​Agent initiates an operation request or task scheduling.

[0111] 2. The execution control module generates an execution plan based on capability dependencies and the conflict matrix, and sorts the capability call order.

[0112] 3. Perform risk classification assessment for each capability operation: : Executes automatically and records operation logs; L2: Conditional execution, triggered by the environment or state; L3: Requires manual confirmation before execution; L4: Disable automatic execution, only log exceptions.

[0113] 4. Combining risk control strategies with AI strategies ensures safe and controllable execution.

[0114] (5) Multi-device collaboration and redundancy optimization 1. When multiple similar devices exist, the system aggregates capabilities to form a unified calling interface.

[0115] 2. Redundancy optimization strategy: Select the optimal device to perform the operation based on the device status, location, power consumption and load.

[0116] 3. Supports load balancing to avoid overloading of a single device.

[0117] 4. Ensure consistent execution results from multiple devices to achieve collaborative control.

[0118] (6) Command issuance and device interface control 1. The execution control module converts the sorted execution plan into device instructions.

[0119] 2. The device interface module sends instructions to the corresponding device according to the communication protocol (such as Wi-Fi, ZigBee, Bluetooth, Modbus, etc.).

[0120] 3. Support asynchronous or synchronous instruction execution modes, dynamically selected according to the task type.

[0121] (7)Status collection and feedback closed-loop 1. The device interface module collects the device status and execution results in real time.

[0122] 2. The status feedback module transmits the execution results back to the artificial intelligence agent (AI Agent) to achieve closed-loop control.

[0123] 3. Abnormal events trigger risk management strategies, record logs and can trigger manual intervention.

[0124] 4. Support the aggregation of multi-device status to ensure cross-device consistency.

[0125] (8)Dynamic capability update and version management 1. The system supports the dynamic registration and incremental update of capability objects, and automatically identifies newly added or modified capabilities.

[0126] 2. Support hot update to complete capability upgrade without stopping the system operation.

[0127] 3. Support version rollback. In case of anomalies or conflicts, the historical version can be restored. <00,00384>

[0128] (9)Separation of data flow and control flow 1. Data flow: Responsible for transmitting device status, sensor data and event information back.

[0129] 2. Control flow: Responsible for issuing AI instructions and execution scheduling.

[0130] 3. The dual-channel design ensures control security and operation auditability, while enhancing system stability.

[0131] Through the above control methods, the present invention can achieve: 1) Standardize and abstract the capabilities of electrical appliances to improve the generality of AI calls; 2) The zero-code automatic call capability of the artificial intelligence agent (AI Agent) reduces development costs; 3) Risk-graded execution, controllable automated operations, and enhanced security; 4) Multi-device collaboration and redundant optimization to improve system stability and efficiency; 5) Dynamic capability update and version management to ensure the continuous evolution of the system; 6) Closed-loop data flow and control flow, supporting status traceability and event auditing; <00,00399>7) Cross-platform and cross-vendor device compatibility to build an open smart home ecosystem.

[0132] Ref. Figure 4As shown in the figure, the innovation points of this control method are as follows: 1. AI-native control: The artificial intelligence agent (AI Agent) directly schedules the device capabilities and calls them with zero code.

[0133] 2. Standardized abstraction of electrical appliance capabilities: The device functions are uniformly abstracted into capability objects, which is convenient for cross-device and cross-vendor calls.

[0134] 3. Risk grading execution: L0-L4 level control is introduced throughout the process to achieve safe and controllable automation.

[0135] 4. Multi-device collaboration: The execution control module supports aggregation, redundant optimization, and load balancing.

[0136] 5. Dynamic capability management: Capability registration, hot update, and version rollback are implemented for dynamic evolution.

[0137] 6. Closed-loop feedback: The status feedback module and the log audit module ensure traceability and reviewability.

[0138] 7. Event trigger mechanism: The tool interface callback supports AI event-driven operations, improving the intelligent response ability.

[0139] The present invention also provides an AI-native control device for environmental appliances for an artificial intelligence agent, including a processor, a memory, and a program module connected to the processor. The program module is configured to execute the method, including: Registration and parsing of capability objects; Automatic mapping of tool interfaces; Execution control and risk grading execution; Multi-device collaboration and redundant optimization; Status feedback closed-loop and dynamic capability update.

[0140] The storage medium can be Flash, SSD, EEPROM, or other computer-readable storage media.

[0141] The executable program code can be called through the artificial intelligence agent (AI Agent) interface module to achieve native control of environmental appliances and form an end-to-end closed-loop system.

[0142] See Figure 5 As shown in the figure, through the above control system and storage medium, the present invention can achieve: 1. AI-native control of environmental appliances: Call the capability interface with zero code to achieve automated intelligent control; 2. Risk grading and safe execution: L0-L4 level policies ensure safe and reliable execution; 3. Multi-device collaboration and capability aggregation: The execution control module embodies redundancy optimization, aggregation call and topology scheduling strategies to improve efficiency and stability; device modules such as air purifiers, humidifiers, and dehumidifiers can be executed in parallel and collaboratively, and are uniformly scheduled by the execution control module.

[0143] 4. Dynamic capability management and version control: Capability registration, dynamic updates, hot updates, and version rollback modules are integrated throughout the entire system, supporting hot updates and cross-device synchronization; 5. Closed-loop data and control flow: The process starts from the AI ​​Agent initiating a task → Capability registration / resolution → Tool mapping → Execution control → Device execution → Status feedback → Dynamic capability update → Back to the AI ​​Agent, forming an end-to-end closed loop. This ensures that operations are traceable and auditable, facilitating the expansion of the smart home ecosystem.

[0144] 6. Function of storage media: to store capability objects, execution plans, status data and logs, so as to achieve traceability and auditability.

[0145] Example 1: Air Quality Improvement Scenario Scene description: Users can issue natural language commands to their home air purifiers through an AI agent: "The air is a bit stuffy, please freshen it for me."

[0146] Implementation steps: 1. Intent recognition: The AI ​​analyzes the instructions and identifies them as an intention to improve air quality.

[0147] 2. Environment Inquiry: The system obtains the air purifier's capability description from the device capability abstraction module: PM2.5 reading ability Wind speed regulation capability Mode switching capability Simultaneously read the current environmental data: PM2.5 = 128 μg / m³ 3 .

[0148] 3. Semantic arrangement: The semantic orchestration module determines based on the rule engine: PM2.5 > 100 → Activate Enhanced Mode Wind speed adjustment → Automatically matches to high wind speed 4. Security Verification: The security control module calculates the risk level: low risk, allowing direct execution.

[0149] 5. Execute the command: The environmental electrical appliance execution module received the enhanced mode command and started high-speed air purification.

[0150] 6. Incident Feedback: The air purifier started running, and the event publishing module reported to the AI ​​agent that "enhanced mode has been executed".

[0151] 7. Log Recording: The execution log records include instruction content, execution time, risk level, and environmental data.

[0152] Example 2: Automatic Sleep Mode Adjustment Scenario Scene description: A nighttime user said, "I'm going to sleep now, let's make the air conditioning more comfortable." Implementation steps: 1. Intent recognition: This is interpreted as an intentional sleep comfort mode.

[0153] 2. Environment and Equipment Inquiry: Current room temperature: 27℃ Current humidity: 60% Air purifier current fan speed: 3 levels Air quality: Good 3. Semantic arrangement: The rules engine determines: Set the fan speed to low (silent mode). Humidifier turned on, target humidity 50%. Switch the air purifier to night mode (low noise mode). 4. Risk Calculation: Wind speed adjustment → Low risk Humidifier startup → Medium risk (overflow prevention) Simultaneous operation of fan speed and humidifier → Level 2 risk 5. Downgrade strategy: Risk level 2 → Triggering downgrade execution: The humidifier flow rate is limited to 50%. Low wind speed mode Executes automatically, no human confirmation required. 6. Implementation and Feedback: Each device executes the policy action, and the event publishing module reports the status to the AI ​​Agent: "Night mode has been activated."

[0154] 7. Logs and Monitoring: Record changes in equipment status, execution time, risk level, and environmental data for historical analysis.

[0155] Example 3: Pet Air Treatment Scenario Scene description: The user has a cat and the command is: "Pets are active, keep the air fresh."

[0156] Implementation steps: 1. Intent recognition: The interpretation is that this is intended to treat the air in the pet's air.

[0157] 2. Environment Inquiry: Indoor temperature = 24℃ The pet activity detection module detected frequent pet movement. 3. Semantic arrangement: Switch the air purifier to medium-high fan speed Humidifier off (anti-slip / waterproof safety) Air quality sensor continuously samples 4. Risk Calculation: Wind speed increases → Medium risk Humidifier off → Low risk Pet activities → Increased risk factors 5. Decision Implementation: Risk rating: Medium → Automatically activate enhanced mode, but keep the humidifier off.

[0158] 6. Execution Feedback: Air purifier fan speed medium to high The AI ​​agent receives real-time feedback on air quality events. 7. Log Recording: This includes execution instructions, risk levels, pet activity records, and changes in air quality.

[0159] Example 4: Multi-device collaborative control scenario Scene description: User instruction: "Before getting out of bed in the morning, adjust the air and humidity in the bedroom to a comfortable level." Implementation steps: 1. Intent recognition: The interpretation is that it represents the intention to prepare the environment for waking up.

[0160] 2. Equipment Inquiry: Air purifier: mode switching capability, fan speed capability Humidifier: Humidification capacity Electric heater: Temperature control capability 3. Obtaining environmental status: Temperature: 18℃ Humidity: 35% 80 μg / m 3 4. Semantic arrangement: The rule engine generates action sequences: Preheat the electric heater to 22℃ The humidifier is turned on to 50% humidity. Air purifier enhanced mode activated 5. Risk classification and coordination: Wind speed increases → Medium risk Electric heater heating → High risk Humidifiers running simultaneously → Medium risk Multiple devices posing a risk → manual confirmation or triggering a downgrade strategy. 6. Implementation strategy: Downgrade the electric heater's power to 80%. Humidifier flow rate 50% Keep the air purifier fan speed at medium. Executes automatically, no human confirmation required. 7. Feedback and Logs: The status of each device is fed back to the artificial intelligence agent. Generate multi-device collaborative execution logs, including risk level, execution order, and environmental data.

[0161] Example 5: Abnormal Environment Handling Scenario Scene description: Air quality suddenly dropped, and the system handled the situation automatically without user intervention.

[0162] Implementation steps: 1. Event Triggering: Air quality sensor detects PM2.5 > 150 μg / m³ 3 The event publishing module notifies the AI ​​Agent. 2. Intent deduction: The system automatically generates the intent: Emergency air purification intent. 3. Semantic arrangement: Air purifier switch to enhanced mode Set the wind speed to maximum. Users are reminded of potential high-risk exposures. 4. Risk Calculation: Maximum wind speed → High risk Continuous Enhancement Mode → Medium Risk Current environment → High risk Overlapping → Level 3 Risk → Manual Confirmation or Automatic Downgrade 5. Implementation strategy: Automatic execution of enhanced mode Reduce continuous running time and set an automatic sleep threshold. Generate exception event logs 6. Feedback and Logs: The AI ​​agent receives the execution status; The system records execution time, risk level, environmental data, and degradation actions.

[0163] The present invention has the following technical effects: 1) AI-native control enables zero-code capability invocation: This invention uses a capability abstraction and registration module and a tool mapping module to abstract all functions of environmental appliances into capability objects, which can be directly invoked by the artificial intelligence agent (AI Agent) without additional programming. This can significantly reduce system integration costs, improve the efficiency of AI-hardware interaction, and support personalized scene configuration for users. AI can autonomously complete environmental control tasks.

[0164] 2) Multi-device collaboration and dynamic aggregation execution: The execution control module supports collaborative scheduling, redundancy optimization, and topology sorting scheduling across multiple devices, enabling multiple environmental appliances to execute safely and reliably according to task dependencies. This improves the accuracy and response speed of environmental control, while achieving optimal utilization of equipment resources. In complex scenarios, such as simultaneously controlling air purifiers, humidifiers, and dehumidifiers, the system can automatically optimize the operation sequence to ensure safe and efficient execution.

[0165] 3) Risk-based execution and safety assurance: This invention proposes a risk-based execution mechanism with levels L0 to L4. Through the execution control module and the status feedback module, the execution of tasks is dynamically monitored and risk assessed. The execution strategy can be adjusted according to the risk level of the task to achieve safe and controllable operation of electrical appliances in the environment. This avoids the direct execution of high-risk operations and ensures user safety and equipment stability.

[0166] 4) Capability hot update and version controllability: The dynamic capability update module supports incremental updates, hot updates and rollback operations for capability objects, ensuring that the system can be upgraded and iterated without interrupting operation. The device functions can be continuously expanded with AI capabilities and scenario requirements, making it highly adaptable. When new sensors are introduced or new control strategies are added, the system can be upgraded without stopping, ensuring long-term availability and continuous system evolution.

[0167] 5) Closed-loop feedback and status traceability: The status feedback module is combined with the storage log module to realize closed-loop feedback and historical record storage of equipment status, execution results and abnormal events, realize traceable and auditable environmental electrical control, facilitate abnormal event analysis and optimization of system strategies, and allow users or maintenance personnel to view equipment operation history and AI decision trajectory, thereby improving system transparency and maintainability.

[0168] 6) Unified and standardized cross-device capabilities: Through capability abstraction and registration mechanisms, the capabilities of environmental appliances of different brands and models are standardized to form a unified interface protocol, which can enhance the system's compatibility and scalability, and lay a standard foundation for the integration of future smart home ecosystems and artificial intelligence agents (AI agents). Air purifiers, humidifiers and dehumidifiers from different manufacturers can all be controlled in a unified manner through the same artificial intelligence agent (AI agent), reducing secondary development costs.

[0169] 7) Intelligent scene optimization and adaptive control: By combining the strategy generation and tool interface call of the artificial intelligence agent, this invention can autonomously optimize the control strategy according to user needs, environmental conditions and historical data, realize adaptive scene control, improve user experience and equipment efficiency, and reduce energy waste. For example, it can automatically adjust air purification and humidity when the user leaves the room, and the environment has been optimized when the user returns, thus achieving intelligent automatic management.

[0170] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A native environmental electrical control system for artificial intelligence agents, characterized in that, include: The AI ​​agent interface module is used to receive control tasks from the AI ​​agent and return the execution status to the AI ​​agent; The capability registration and management module is used to abstract the various functions of environmental electrical appliances into capability objects, and supports dynamic registration, updating and version management; The capability parsing module is used to parse the syntax and semantics of capability objects, generate capability dependency graphs, and detect capability conflicts. The tool mapping module is used to automatically map capability objects to tool interfaces that can be called by the AI ​​agent, and to generate parameter structure definitions and callback interfaces; The execution control module is used to generate execution plans based on task dependencies, and to perform topology sorting, risk-level execution, conflict resolution, and redundancy optimization for execution on multiple devices. The status feedback module is used to receive the device execution status and return it to the AI ​​entity, while also performing anomaly monitoring and multi-device status aggregation. The dynamic capability update module is used to implement incremental updates, hot updates, and rollbacks of capability objects, ensuring cross-device capability synchronization. The storage and logging module is used to store capability objects, execution plans, device status, and exception logs.

2. The native environmental electrical control system for artificial intelligence agents according to claim 1, characterized in that, The execution control module dynamically adjusts the task execution order through a risk-based execution strategy to ensure that high-risk operations are constrained, while low-risk operations are executed first.

3. The native environmental electrical control system for artificial intelligence agents according to claim 1, characterized in that, The capability registration and management module supports unified capability abstraction for environmental appliances across brands and models, forming a standardized interface protocol.

4. The native environmental electrical control system for artificial intelligence agents according to claim 1, characterized in that, The tool mapping module supports zero-code capability calls and is used to automatically map newly added capability objects to tool interfaces that can be called by the artificial intelligence agent.

5. The native environmental electrical control system for artificial intelligence agents according to claim 1, characterized in that, The dynamic capability update module works in conjunction with the storage and log module to enable hot updates of capability objects, version rollback, and traceable storage of historical execution data.

6. The native environmental electrical control system for artificial intelligence agents according to any one of claims 1 to 5, characterized in that, The environmental appliances include air purifiers, humidifiers, dehumidifiers, and other home or office environmental control devices. Each device is connected to the system through a unified capability abstraction interface to achieve collaborative control of multiple devices.

7. A native control method for environmental electrical appliances for artificial intelligence agents, characterized in that, Includes the following steps: Step 1: Receive control tasks from the AI ​​agent; Step 2: Register the functions of environmental electrical appliances as capability objects, and perform dynamic management and version control; Step 3: Perform syntactic and semantic parsing on the capability objects to generate a capability dependency graph and detect conflicts; Step 4: Map the capability object to a tool interface that the AI ​​agent can call, and generate parameter structure definitions and callback interfaces; Step 5: Generate an execution plan based on task dependencies, and control the execution of tasks through risk-based execution, multi-device collaboration, conflict resolution, and redundancy optimization. Step 6: Collect device execution status and abnormal events, return the results to the AI ​​agent, and perform multi-device status aggregation; Step 7: Perform incremental updates, hot updates, or rollbacks on the capability objects to ensure cross-device synchronization, and store the relevant execution data and logs to the storage medium.

8. The native control method for environmental appliances for artificial intelligence agents according to claim 7, characterized in that, Step 5 includes: dynamically adjusting the task execution priority according to the risk level, with low-risk tasks executed first and high-risk tasks executed in a controlled manner.

9. A native control device for environmental appliances for artificial intelligence agents, characterized in that, It includes a processor, a memory, and a program module connected to the processor, the program module being configured to perform the method of claim 7 or 8.

10. A non-volatile storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are used to cause the computer to perform the method of claim 7 or 8.

Images