IoT device network configuration method and device combining edge cloud and large model

Abstract

The present disclosure provides an IoT device network configuration method and device combining an end cloud and a large model, which calls a large language model to analyze a user natural language network configuration request through a cloud-side intelligent agent, generates a structured network configuration task plan in combination with device network configuration characteristics, and performs network configuration interaction with a device side according to the task plan by an APP-side intelligent agent; meanwhile, end cloud collaborative control is realized in the network credential writing, device activation binding and abnormality processing processes. Thus, the user manual selection and repeated trial and error operations can be reduced, the network configuration process can be dynamically adjusted according to the device type, user context and on-site abnormal state, and the convenience, security, abnormality repair capability and network configuration success rate of the IoT device network configuration are improved.

Description

Technical Field

[0001] This disclosure relates to the field of IoT device network configuration technology, and more specifically, to an IoT device network configuration method and apparatus that combines edge and cloud computing in a large-scale model. Background Technology

[0002] With the development of IoT applications such as smart homes, smart security, and smart lighting, an increasing number of IoT devices need to connect to users' home networks upon first use or when the network environment changes. Existing IoT devices typically connect to the network via Bluetooth, SoftAP, or QR code pairing. In a typical pairing process, users need to select the device type, enter the Wi-Fi network name and password in the app, and follow the app's preset pages to complete operations such as device scanning, credential writing, network connection waiting, and device binding. The cloud server is typically used to generate pairing tokens, verify device legitimacy, and bind the device to the user's account after successful network connection.

[0003] However, existing network configuration solutions often use fixed pages and processes to guide users. The app typically only allows users to follow preset steps such as device selection, permission request, password input, and waiting for connection, making it difficult to understand users' network configuration needs expressed in natural language. For example, when a user says, "Help me connect the new camera I bought in the living room to the internet," existing solutions usually still require the user to manually select the device type, model, or network configuration entry point. The interaction is relatively mechanical, and the user threshold is high.

[0004] Meanwhile, different categories and models of IoT devices differ in terms of network configuration capabilities, supported wireless network frequency bands, near-field communication methods, methods for indicating network configuration status, and activation conditions. Existing network configuration processes are usually pre-defined in the APP or cloud logic for specific devices. When the device type or the user's on-site network environment changes, it is difficult to dynamically generate appropriate network configuration steps based on the device's network configuration characteristics and the user's context, resulting in insufficient universality and adaptability of the network configuration process. Summary of the Invention

[0005] This disclosure provides at least one method and apparatus for IoT device network configuration using a cloud-edge integrated big data model. The method involves a cloud-based intelligent agent that uses a big data model to parse user natural language network configuration requests, generating a structured network configuration task plan based on device network configuration characteristics. The intelligent agent on the app then interacts with the device according to this task plan. Simultaneously, cloud-edge collaborative control is achieved during network credential writing, device activation and binding, and anomaly handling. This reduces manual selection and repeated trial-and-error operations by the user, allowing the network configuration process to dynamically adjust based on device type, user context, and on-site anomaly conditions, thereby improving the convenience, security, anomaly recovery capabilities, and success rate of IoT device network configuration.

[0006] This disclosure provides a method for configuring IoT devices using a combined edge-cloud and large-scale model, including: Obtain the natural language network configuration request input by the user for the IoT device to be configured, and send the natural language network configuration request and user context information to the cloud intelligent agent; The cloud-based intelligent agent invokes a large language model to perform intent recognition and slot filling on the natural language network configuration request, thereby obtaining network configuration intent information and device feature information. Based on the network distribution intent information, the device feature information, and the preset device knowledge base, a structured network distribution task plan matching the IoT device to be distributed to is generated. The structured network distribution task plan includes multiple network distribution step nodes and execution actions and user guidance information corresponding to each network distribution step node. The structured distribution network task plan is sent to the smart agent on the APP, so that the smart agent on the APP can execute the distribution network guidance according to the structured distribution network task plan and interact with the IoT device to be distributed through the near field communication link; When executing the network credential writing step, the network credential input by the user is encrypted to generate an encrypted credential, and the encrypted credential is transmitted to the IoT device to be configured through the smart agent of the APP. The IoT device to be configured decrypts the encryption credential to obtain network connection information, and after accessing the target network based on the network connection information, sends an activation request to the cloud intelligent agent; the cloud intelligent agent verifies the activation request, and after the verification is successful, completes the binding of the IoT device to be configured with the user account.

[0007] In one optional implementation, the natural language network distribution request is subjected to intent recognition and slot filling to obtain network distribution intent information and device characteristic information, specifically including: Identify at least one network configuration intent from the natural language network configuration request: adding a device, reconfiguring the network, or connecting the device to the network; Extract at least one slot information from the natural language network configuration request, including equipment category, equipment location, equipment appearance features, equipment model, and equipment status; The slot information is matched with the user's historical device binding records, user home space information, and device knowledge base to determine the device feature information corresponding to the IoT device to be configured.

[0008] In one optional implementation, based on the distribution network intent information, the device feature information, and a preset device knowledge base, a structured distribution network task plan matching the IoT device to be distributed to is generated, specifically including: Based on the device feature information, the network configuration capability information of the IoT device to be configured is retrieved from the preset device knowledge base. The network configuration capability information includes at least one of the following: supported near-field communication methods, supported wireless network frequency bands, network configuration status indication methods, and activation conditions. The network distribution path corresponding to the IoT device to be distributed is determined based on the network distribution capability information. A structured network distribution task plan is generated based on the network distribution path, including at least some of the steps in the following steps: status confirmation, permission request, device scanning, target confirmation, credential input, credential writing, activation waiting, and configuration completion.

[0009] In one optional implementation, the APP-based intelligent agent executes network distribution guidance according to the structured network distribution task plan, specifically including: When executing each distribution network step node, the current distribution network execution status is collected and the current distribution network execution status is fed back to the cloud intelligent agent; The cloud-based intelligent agent invokes the large language model to dynamically generate the current user guidance script based on the current network configuration execution status and historical interaction context. The current user guidance script is sent to the intelligent agent on the APP and output to the user through interface display or voice broadcast.

[0010] In one optional implementation, data interaction with the IoT device to be networked is performed via a near-field communication link, specifically including: The APP-based intelligent agent invokes the Bluetooth communication link or SoftAP communication link to scan for nearby candidate IoT devices. If multiple candidate IoT devices are detected, the list of candidate IoT devices is sent to the cloud-based intelligent agent; The cloud-based intelligent agent filters and sorts multiple candidate IoT devices based on user historical binding records, device feature information, and candidate IoT device signal information, and generates target device confirmation information. The IoT device to be configured is determined based on the user's confirmation operation regarding the target device confirmation information.

[0011] In one optional implementation, the network credentials input by the user are encrypted to generate encrypted credentials, specifically including: Obtain the device public key or session key corresponding to the IoT device to be configured on the network; Based on the device public key or session key, the wireless network identifier and wireless network password entered by the user are encrypted to obtain the encryption credential; The APP-side intelligent agent does not store or parse the wireless network password locally, and transmits the encrypted credentials in ciphertext form to the designated communication characteristic value or designated data receiving interface of the IoT device to be configured.

[0012] In one optional implementation, the method further includes: When an abnormal event is detected during the distribution network process, abnormal information corresponding to the abnormal event is obtained. The abnormal information includes at least one of the following: distribution network error code, equipment model information, equipment communication capability information, current network environment information, and historical similar case information. The abnormal information is organized into abnormal diagnosis prompts and input into the large language model; The large language model is used to perform fault root cause reasoning on the abnormal events in the distribution network, generating abnormal root cause information and corresponding natural language repair suggestions. The natural language repair suggestions are sent to the smart agent on the APP to guide the user to perform the anomaly repair operation.

[0013] In one optional implementation, the method further includes: The large language model generates task plan adjustment instructions based on the anomaly root cause information. Based on the task plan adjustment instruction, a repair step node is inserted into the original structured distribution network task plan or the subsequent retry strategy is adjusted to obtain the corrected structured distribution network task plan. The revised structured distribution network task plan is subjected to security verification, and after the security verification is passed, it is sent to the APP-side intelligent agent, so that the APP-side intelligent agent continues to execute the distribution network process according to the revised structured distribution network task plan.

[0014] In one optional implementation, the method further includes: After the IoT device to be configured to the network completes the configuration and is bound to the user account, the device name and space ownership information corresponding to the IoT device to be configured are generated based on the device location slot and device category slot in the natural language configuration request. Update the device shadow status based on the device name and spatial ownership information; The large language model generates intelligent scene recommendation information related to the IoT device to be configured based on the network intent information, and pushes the intelligent scene recommendation information to the intelligent agent on the APP.

[0015] This disclosure also provides an IoT device network configuration device that combines edge and cloud computing in a large-scale model, including: The data acquisition module is used to acquire the natural language network configuration request input by the user for the IoT device to be configured, and send the natural language network configuration request and user context information to the cloud intelligent agent; The information recognition module is used to call the large language model through the cloud intelligent agent to perform intent recognition and slot filling on the natural language network distribution request, and obtain network distribution intent information and device feature information. The plan generation module is used to generate a structured distribution network task plan that matches the IoT device to be distributed to the network based on the distribution network intent information, the device feature information and the preset device knowledge base. The structured distribution network task plan includes multiple distribution network step nodes and execution actions and user guidance information corresponding to each distribution network step node. The guidance control module is used to send the structured distribution network task plan to the APP-side intelligent agent, so that the APP-side intelligent agent can execute the distribution network guidance according to the structured distribution network task plan and interact with the IoT device to be distributed through the near-field communication link; The encryption processing module is used to encrypt the network credentials input by the user when the network credentials writing step is executed, generate encrypted credentials, and transmit the encrypted credentials to the IoT device to be configured through the smart agent of the APP. The activation binding module is used to decrypt the encryption credential through the IoT device to be configured, obtain network connection information, and send an activation request to the cloud intelligent agent after accessing the target network based on the network connection information; the cloud intelligent agent verifies the activation request, and completes the binding between the IoT device to be configured and the user account after the verification is successful.

[0016] This disclosure also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they execute the steps of the above-described edge-cloud combined big data model IoT device network configuration method, or any possible implementation of the above-described edge-cloud combined big data model IoT device network configuration method.

[0017] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described edge-cloud combined big data model IoT device network configuration method, or any possible implementation of the above-described edge-cloud combined big data model IoT device network configuration method.

[0018] This disclosure also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the above-described IoT device network configuration method for the edge-cloud combined big data model, or any possible implementation of the above-described IoT device network configuration method for the edge-cloud combined big data model.

[0019] This disclosure provides an IoT device configuration method and apparatus that combines edge and cloud computing with a large-scale model. The method involves a cloud-based intelligent agent calling a large language model to parse the user's natural language configuration request, generating a structured configuration task plan based on the device's configuration characteristics, and then having the intelligent agent on the APP interact with the device according to this task plan. Simultaneously, edge-cloud collaborative control is achieved during network credential writing, device activation and binding, and anomaly handling. This reduces manual selection and repeated trial-and-error operations by the user, allowing the configuration process to dynamically adjust based on device type, user context, and on-site anomaly conditions, thereby improving the convenience, security, anomaly recovery capabilities, and configuration success rate of IoT device configuration.

[0020] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. These drawings are incorporated in and constitute a part of this specification. They illustrate embodiments conforming to this disclosure and, together with the specification, serve to explain the technical solutions of this disclosure. It should be understood that the following drawings only show some embodiments of this disclosure and should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without creative effort.

[0022] Figure 1 This document illustrates a flowchart of an IoT device network configuration method based on an embodiment of an edge-cloud combined large-scale model provided by this disclosure. Figure 2 A flowchart is shown below illustrating another IoT device network configuration method based on a large-scale edge-cloud combined model provided in this disclosure embodiment; Figure 3 This diagram illustrates a schematic of an IoT device network configuration device based on a large-scale edge-cloud combined model provided in an embodiment of this disclosure; Figure 4 A schematic diagram of an electronic device provided in an embodiment of this disclosure is shown. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0024] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0025] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0026] Research has revealed that existing network configuration solutions often employ fixed pages and processes to guide user operations. Apps typically only execute pre-defined steps such as device selection, permission requests, password input, and connection waiting, failing to understand users' network configuration needs expressed in natural language. For example, when a user requests, "Connect my newly purchased living room camera to the internet," existing solutions usually still require the user to manually select the device type, model, or network configuration entry point, resulting in a mechanical interaction and a high barrier to entry. Furthermore, different types and models of IoT devices vary in their network configuration capabilities, supported wireless network frequency bands, near-field communication methods, pending network configuration status indicators, and activation conditions. Existing network configuration processes are usually pre-defined in the app or cloud logic for specific devices. When device types or the user's on-site network environment change, it is difficult to dynamically generate adapted network configuration steps based on device configuration characteristics and user context, leading to insufficient universality and adaptability of the network configuration process.

[0027] Based on the above research, this disclosure provides an IoT device configuration method and apparatus that combines edge and cloud computing with a large-scale model. The method involves a cloud-based intelligent agent calling a large language model to parse the user's natural language configuration request, generating a structured configuration task plan based on the device's configuration characteristics, and then having the intelligent agent on the APP interact with the device according to this task plan. Simultaneously, edge-cloud collaborative control is achieved during network credential writing, device activation and binding, and anomaly handling. This reduces manual selection and repeated trial-and-error operations by the user, allowing the configuration process to dynamically adjust based on device type, user context, and on-site anomaly conditions, thereby improving the convenience, security, anomaly recovery capabilities, and configuration success rate of IoT device configuration.

[0028] To facilitate understanding of this embodiment, a detailed description of the edge-cloud integrated IoT device network configuration method disclosed in this disclosure embodiment is provided first. The execution entity of the edge-cloud integrated IoT device network configuration method provided in this disclosure embodiment is generally a computer device with certain computing capabilities. This computer device may include, for example, a terminal device, a server, or other processing devices. The terminal device may be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc. In some possible implementations, this edge-cloud integrated IoT device network configuration method can be implemented by a processor calling computer-readable instructions stored in memory.

[0029] See Figure 1 The diagram shows a flowchart of an IoT device network configuration method based on a large-scale edge-cloud combined model provided in this disclosure. The method includes steps S101 to S106, wherein: S101. Obtain the natural language network configuration request input by the user for the IoT device to be configured, and send the natural language network configuration request and user context information to the cloud intelligent agent.

[0030] In practice, users can input natural language network configuration requests for IoT devices through the network configuration interaction interface provided by the APP. These natural language requests can be text input requests or voice input requests that are then processed into text via speech recognition. The natural language network configuration request expresses the user's intention to perform network configuration operations on a specific IoT device; examples include "Connect the new camera I bought in the living room to the internet," "Add this device with a screen to the home network," and "Reconnect the smart light in the bedroom."

[0031] Here, after the APP-side intelligent agent receives the natural language network configuration request input by the user, it can perform basic preprocessing on the natural language network configuration request. The basic preprocessing includes at least one of speech-to-text conversion, text format normalization, sensitive character filtering, language type recognition, and invalid input filtering.

[0032] Specifically, when sending a natural language network configuration request to the cloud-based intelligent agent, the APP-side intelligent agent can also simultaneously obtain and send user context information. This user context information assists the cloud-based intelligent agent in understanding the user's current network configuration needs and may include at least one of the following: user identifier, user account information, user home space information, historical device binding records, current APP page status, user authorized permission status, user's home or room information, current network environment information, and candidate device information scanned by the APP.

[0033] Among them, user home space information can include space division information such as living room, bedroom, and kitchen; historical device binding records can include information such as device name, device category, device location, device model, and device communication identifier of the bound devices; current network environment information can include the currently connected Wi-Fi name, network frequency band, router-related status, or wireless network information that can be detected by the APP.

[0034] For example, when a user inputs "Help me connect the new camera with a screen in the living room to the internet," the smart agent on the app can send this natural language network configuration request, the user's account identifier, existing room information in the user's home, a list of bound devices, and the scan information of candidate devices detected by the app to the cloud-based smart agent. After receiving the above information, the cloud-based smart agent can combine the user's context information to determine the device location, device status, device appearance characteristics, and device category corresponding to the user's expressions such as "living room," "newly purchased," "with screen," and "camera," thereby providing a data foundation for subsequent intent recognition, slot filling, and structured network configuration task plan generation.

[0035] In some embodiments, user context information is not limited to information collected locally on the app, but may also include historical context information queried by the cloud-based intelligent agent based on the user's account. For example, the cloud-based intelligent agent can query information such as the binding status of existing IoT devices under the user's name, home space configuration, past network configuration failure records, and device usage preferences based on the user's identifier, and use this information as an auxiliary basis for subsequent semantic parsing and task plan generation. Thus, even if the user's natural language expression is relatively brief, the cloud-based intelligent agent can still complete the network configuration requirements by combining the context, reducing the user's manual operation of selecting device type, device location, or network configuration entry point.

[0036] Thus, this embodiment does not require users to select device types and input network distribution parameters one by one on a fixed page. Instead, it allows users to describe their network distribution needs in natural language, and the intelligent agent on the APP sends the natural language network distribution request and user context information to the cloud intelligent agent. This enables the cloud intelligent agent to perform subsequent intelligent parsing and dynamic task planning based on the user's expression and context information, thereby improving the ease of use of the network distribution entry and the adaptive capability of the network distribution process.

[0037] S102. The cloud-based intelligent agent calls the large language model to perform intent recognition and slot filling on the natural language network distribution request, thereby obtaining network distribution intent information and device feature information.

[0038] In practice, after receiving the natural language network configuration request and user context information from the APP-based intelligent agent, the cloud-based intelligent agent invokes a large language model to perform semantic understanding on the natural language network configuration request, in order to identify the user's network configuration intent and extract slot information related to the IoT devices to be configured. The large language model can be deployed on a cloud server or invoked by the cloud-based intelligent agent through a model service interface, and is used to output structured network configuration parsing results based on the user's input text and context information.

[0039] Specifically, the cloud-based intelligent agent can combine natural language network configuration requests, user context information, candidate device information, and preset parsing prompt templates to generate model input information, and then input this model input information into a large language model. The preset parsing prompt templates can be used to constrain the large language model to output network configuration intent information and device characteristic information according to preset fields, avoiding the output of only unstructured text. The preset fields may include at least one of the following: intent field, device category field, device location field, device appearance characteristic field, device model field, device status field, network configuration action field, and confidence level field.

[0040] In some embodiments, intent identification is used to determine the type of network configuration operation the user wishes to perform. The network configuration intent information may include at least one of adding a new device, reconfiguring the network, restoring device connectivity, switching the device's network connection, or continuing a previously failed network configuration process. For example, when a user enters "Help me connect the new camera I bought in the living room to the internet," the large language model can identify the network configuration intent as adding a new device; when a user enters "The smart light in my bedroom can't connect after I changed the router, please reconnect it," the large language model can identify the network configuration intent as reconfiguring the network or restoring device connectivity.

[0041] In some embodiments, slot filling is used to extract or combine user context information from the natural language network configuration request to complete the device characteristic information related to the IoT device to be configured. The device characteristic information may include at least one of the following: device category, device location, device appearance features, device model, device status, device usage scenario, and network configuration target described by the user.

[0042] The device categories can include cameras, smart lights, smart speakers, smart sockets, sensors, etc.; the device locations can include living room, bedroom, kitchen, doorway, etc.; the device appearance characteristics can include having a screen, black, round, small, newly purchased, etc.; the device status can include first-time network configuration, powered on, indicator light flashing, scanned, previously failed to bind, etc.

[0043] For example, for a natural language network configuration request like "Help me connect the new camera with a screen in my living room to the internet," the cloud-based intelligent agent, after calling the large language model, can obtain the following structured parsing results: the configuration intent is to add a device, the device category is a camera, the device location is the living room, the device appearance feature is that it has a screen, and the device status is newly purchased or first-time network configuration. These structured parsing results can serve as the basis for subsequent retrieval of the device knowledge base, determination of the network configuration method, and generation of a structured network configuration task plan.

[0044] In some embodiments, when the user's input lacks partial slot information, the large language model can complete the input or generate confirmation information based on the user's context. For example, if the user only inputs "Help me connect this device to the internet," the cloud-based intelligent agent can combine candidate device information scanned by the app, current room information, historical device binding records, etc., to infer the device category and status of the candidate device. If the inference results in multiple possible objects, a prompt message requiring user confirmation can be generated, such as "Two network-configurable devices have been found nearby. Do you want to add this new camera?"

[0045] In some embodiments, the cloud-based intelligent agent can also verify the intent recognition results and slot filling results output by the large language model. Specifically, it can perform consistency verification on the device category, device location, and device model output by the large language model based on the device category table, device model table, home space information under the user account, and candidate device broadcast information in the device knowledge base. When the verification result shows that the confidence of a certain slot information is insufficient, the slot can be marked as pending confirmation, and the corresponding user confirmation step can be inserted in the subsequent task plan generation process.

[0046] In this way, the cloud-based intelligent agent can convert users' freely expressed natural language network distribution requests into structured network distribution intent information and equipment characteristic information. This allows the subsequent network distribution process to no longer rely on fixed entry points and manual selection, but can automatically determine network distribution goals and requirements based on the user's actual expression, equipment characteristics, and contextual information, thereby providing an accurate data foundation for dynamically generating structured network distribution task plans.

[0047] S103. Based on the network distribution intent information, the device feature information, and the preset device knowledge base, generate a structured network distribution task plan that matches the IoT device to be distributed. The structured network distribution task plan includes multiple network distribution step nodes and execution actions and user guidance information corresponding to each network distribution step node.

[0048] In practical implementation, after obtaining network distribution intent information and device characteristic information, the cloud-based intelligent agent can generate a structured network distribution task plan that matches the IoT device to be distributed, based on the network distribution intent information, the device characteristic information, and a preset device knowledge base. The structured network distribution task plan describes the sequence of steps required to complete the network distribution for the IoT device, the corresponding actions for each step, and the guidance information output to the user, enabling the intelligent agent on the APP to execute subsequent network distribution guidance and device interaction according to the structured network distribution task plan.

[0049] In some embodiments, the preset device knowledge base may include network configuration capability information and network configuration rule information for different categories and models of IoT devices. The network configuration capability information may include at least one of the following: the near-field communication methods supported by the device, the supported wireless network frequency bands, the method for indicating the pending network configuration status, the method for writing network credentials, the device permission requirements, and the device activation conditions.

[0050] The near-field communication methods can include Bluetooth communication, SoftAP communication, acoustic communication, QR code-assisted communication, etc.; the supported wireless network frequency bands can include those that only support 2.4GHz, those that support both 2.4GHz and 5GHz, etc.; the methods for indicating the network configuration status can include flashing indicator lights, displaying the network configuration code on the screen, and voice prompts to enter the network configuration mode, etc.

[0051] Specifically, the cloud-based intelligent agent can search a pre-defined device knowledge base based on device characteristic information to determine the network configuration characteristics of the IoT device to be configured. For example, when the device characteristic information indicates that the IoT device to be configured is a "camera with a screen," the cloud-based intelligent agent can retrieve the typical network configuration characteristics corresponding to this type of camera, such as supporting Bluetooth network configuration, supporting only 2.4GHz wireless networks, a fast-flashing indicator light after power-on indicating that it is in a network configuration state, requiring the APP to obtain Bluetooth and location permissions, and needing to send an activation request to the cloud with a network configuration token after the device successfully connects to the network.

[0052] Furthermore, after determining the network configuration characteristics of the devices, the cloud-based intelligent agent can determine the task type of this network configuration task based on the network configuration intent information, and generate a structured network configuration task plan in conjunction with the device network configuration characteristics. For example, when the network configuration intent information is to add a new device, the structured network configuration task plan may include steps such as status confirmation, permission request, device scanning, target confirmation, network credential input, encrypted credential writing, activation waiting, and configuration completion; when the network configuration intent information is to reconfigure the network, the structured network configuration task plan may also include steps such as deactivating the old network status, resetting the device network configuration status, or skipping the completed binding verification.

[0053] In some embodiments, the structured distribution network task plan includes multiple distribution network step nodes. Each distribution network step node may include at least one of the following: step type, execution action, user guidance information, execution conditions, status feedback requirements, and exception branch rules. The step type is used to identify the functional category of the current step in the distribution network process; the execution action is used to indicate the local capabilities that the APP-side intelligent agent needs to call or the data interaction that needs to be performed with the device; the user guidance information is used to explain to the user the operation that needs to be performed in the current step; the execution conditions are used to determine whether the current step is completed and whether it can proceed to the next step; the status feedback requirements are used to indicate the execution status that the APP-side intelligent agent needs to report to the cloud-side intelligent agent; and the exception branch rules are used to trigger corresponding exception handling logic when an exception occurs in the current step.

[0054] For example, a status confirmation node in a structured network configuration task plan might include: a step type of status confirmation, an action of displaying the device's pending network configuration status confirmation interface, user guidance information prompting the user to confirm whether the device's indicator light is flashing, and an execution condition of user confirmation that the device has entered the pending network configuration state. A permission request node might include: an action of calling the APP's Bluetooth and location permission request interface, user guidance information prompting the user to authorize Bluetooth and location permissions for searching nearby devices, and an execution condition of successful permission authorization. A device scanning node might include: an action of calling the Bluetooth scanning function to search for nearby candidate IoT devices, user guidance information prompting the user to keep the device powered on and close to the phone, and an execution condition of detecting a candidate device or reaching a preset scanning duration.

[0055] For example, the credential writing node can include: the action of obtaining network credentials and writing encrypted credentials to the IoT device to be configured via a near-field communication link; the user guidance information prompting the user to enter the home Wi-Fi password and keep the APP running in the foreground; and the execution condition being that the device returns a successful credential reception result. The activation waiting node can include: the action of waiting for the IoT device to be configured to connect to the target network and then initiating an activation request to the cloud-based smart agent; the user guidance information prompting the user to wait for the device to connect to the network; and the execution condition being that the cloud-based smart agent receives and verifies the device activation request.

[0056] In some embodiments, the cloud-based intelligent agent can also invoke a large language model to generate user guidance information corresponding to each network configuration step node based on the current device's network configuration characteristics and user context information. Unlike preset fixed text, the user guidance information can be dynamically generated by combining user expression, device type, device location, and the current step. For example, for a camera device, guidance information such as "Please confirm that the indicator light on the camera screen is flashing" can be generated; for a smart light device, guidance information such as "Please turn the light on and off three times consecutively to make the light flash" can be generated. Thus, different devices can correspond to different step instructions and operation prompts.

[0057] In some embodiments, the structured distribution network task plan can be represented using a structured data format, such as JSON, XML, or other task description formats. Each distribution network step node in the structured distribution network task plan can be arranged in a preset order, or it can include conditional jump relationships, so that the intelligent agent on the APP can switch subsequent steps according to the actual execution status. For example, when a single candidate device is scanned, it can directly enter the target confirmation node; when multiple candidate devices are scanned, it can enter the candidate device sorting and user confirmation node; when no device is scanned, it can enter the rescan or device status check node.

[0058] In this way, the cloud-based intelligent agent does not send fixed page processes to the app-based intelligent agent. Instead, it dynamically generates a structured network configuration task plan that matches the IoT devices to be configured, based on network configuration intent information, device characteristic information, and a preset device knowledge base. This structured network configuration task plan includes both network configuration actions that can be executed by the app and user-oriented guidance information, enabling the app-based intelligent agent to complete network configuration interactions such as device search, user confirmation, credential writing, and activation waiting according to the plan. This improves the adaptability of the network configuration process to different IoT devices and different user scenarios.

[0059] S104. The structured distribution network task plan is sent to the APP-side intelligent agent, so that the APP-side intelligent agent executes the distribution network guidance according to the structured distribution network task plan and interacts with the IoT device to be distributed through the near-field communication link.

[0060] In practice, after the cloud-based intelligent agent generates a structured distribution network task plan, it can send the structured distribution network task plan to the intelligent agent on the APP. After receiving the structured distribution network task plan, the intelligent agent on the APP can parse the multiple distribution network step nodes contained therein, and execute the distribution network guidance operation sequentially according to the execution actions, user guidance information, and status feedback requirements corresponding to each distribution network step node.

[0061] Specifically, the cloud-based intelligent agent can send the structured network distribution task plan to the APP-side intelligent agent via a persistent internet connection, message push connection, or request-response connection. The persistent internet connection can use communication methods such as MQTT, SSE, or WebSocket. After receiving the structured network distribution task plan, the APP-side intelligent agent can determine the current network distribution step node to be executed based on the step sequence or conditional jump relationship in the structured network distribution task plan, and then invoke the corresponding local capabilities of the APP to execute the corresponding action of the network distribution step node.

[0062] In some embodiments, the smart agent on the APP can display corresponding operation prompts in the APP interface and / or output corresponding operation prompts to the user through voice broadcast, based on the user guidance information of the current network configuration step node. For example, at the status confirmation node, the smart agent on the APP can prompt the user to confirm whether the IoT device to be configured is powered on, whether the indicator light is flashing, and whether the screen displays the network configuration icon; at the permission request node, the smart agent on the APP can prompt the user to enable Bluetooth permission, location permission, or LAN access permission; at the device scanning node, the smart agent on the APP can prompt the user to keep the mobile phone close to the IoT device to be configured and wait for the scan results.

[0063] Here, during the execution of each distribution network step, the APP-based intelligent agent can also collect the current distribution network execution status and feed it back to the cloud-based intelligent agent. The current distribution network execution status may include at least one of the following: current step identifier, step execution result, authorization result, device scan result, candidate device list, user confirmation result, credential writing result, device return status, activation waiting time, and error code. Based on the current distribution network execution status fed back by the APP-based intelligent agent, the cloud-based intelligent agent can determine whether to proceed to the next distribution network step, or whether it is necessary to call the large language model to regenerate the user guidance script or adjust subsequent distribution network steps.

[0064] In some embodiments, the APP-based smart agent can interact with the IoT device to be configured via a near-field communication link. This near-field communication link can be a Bluetooth link, a SoftAP link, an NFC link, an acoustic link, or other data communication links suitable for near-field network configuration. The APP-based smart agent can activate the corresponding near-field communication module according to the communication method specified in the structured network configuration task plan, and establish a communication connection with the IoT device to be configured, performing at least one of the following operations: device scanning, device identification, device status reading, encrypted credential writing, and network configuration result reception.

[0065] In some embodiments, when the structured network distribution task plan indicates the use of Bluetooth communication, the smart agent on the APP can enable Bluetooth scanning to scan for nearby candidate IoT devices in the network distribution state and obtain at least one of the following: broadcast information, device name, device type, device identifier, signal strength, and service characteristics. The smart agent on the APP can then feed back the list of candidate devices obtained from the scan to the smart agent on the cloud. The smart agent on the cloud will then filter and sort the candidate IoT devices based on the network distribution requirement information, device characteristic information, user historical binding records, and candidate device communication information, and generate a target device recommendation result.

[0066] For example, when multiple candidate IoT devices are detected, the cloud-based intelligent agent can combine user-inputted device characteristics such as "living room," "camera," "newly purchased," and "with screen," as well as existing device binding records under the user's account, to filter out already bound devices or devices that do not match the current network configuration requirements. The agent then sends the recommended devices as target devices to the app-based intelligent agent. The app-based intelligent agent can display the target device recommendations to the user and, based on the user's confirmation of the recommended device, determine the IoT device to be configured in this network configuration interaction.

[0067] In some embodiments, when the structured network distribution task plan indicates the use of SoftAP communication, the smart agent on the APP can guide the user to connect their mobile phone to the hotspot network provided by the IoT device to be networked, or automatically redirect to the system network settings page for the user to connect to the device's hotspot. After the smart agent on the APP establishes a SoftAP communication connection with the IoT device to be networked, it can read the device status, send network distribution requests, or write encrypted credentials through the local area network communication interface. For other near-field communication methods, the smart agent on the APP can also invoke the corresponding communication capabilities according to the execution actions defined in the structured network distribution task plan.

[0068] In some embodiments, the app-side intelligent agent does not simply mechanically execute fixed page processes, but rather provides adjustable network configuration guidance based on the structured network configuration task plan issued by the cloud-side intelligent agent. When a user fails to complete an operation, refuses authorization, detects multiple devices, or fails to detect any devices at a certain step, the app-side intelligent agent can send the corresponding status back to the cloud-side intelligent agent. The cloud-side intelligent agent can then combine the historical interaction context and the current execution status to generate new natural language guidance scripts using a large language model and send them to the app-side intelligent agent. For example, when device scanning takes a long time, the user can be prompted with "Please confirm that the device has entered the flashing state and try to bring your phone closer to the device"; when multiple devices are detected, the user can be prompted with "Multiple devices have been detected nearby, please confirm whether to add this new camera."

[0069] In this way, the structured distribution network task plan can form executable task instructions between the cloud-based intelligent agent and the APP-based intelligent agent. The APP-based intelligent agent completes user guidance and near-field device interaction based on the task instructions, so that the distribution network process has both the understanding and planning capabilities of the cloud-based large language model, and can actually complete device scanning, target confirmation, status feedback and data interaction through the local capabilities of the APP, thereby improving the executability of the distribution network process and the continuity of user operation.

[0070] S105. When executing the network credential writing step, the network credential entered by the user is encrypted to generate an encrypted credential, and the encrypted credential is transmitted to the IoT device to be configured through the smart agent of the APP.

[0071] In practical implementation, when the structured network distribution task plan reaches the network credential writing step, the smart agent on the APP can guide the user to input network credentials for accessing the target network. These network credentials may include a wireless network identifier and a wireless network password, such as a Wi-Fi SSID and password; in other embodiments, the network credentials may also include authentication information for accessing a local area network, a private network, or a device management network.

[0072] Here, after the user inputs the network credentials, the network credentials can be encrypted to generate an encrypted credential. Specifically, the network credentials can be encrypted based on the device public key, preset key, session key, or temporary negotiation key corresponding to the IoT device to be configured, resulting in an encrypted credential that can only be decrypted by the IoT device to be configured.

[0073] The device public key can be pre-stored in the device knowledge base, device certificate store, or cloud device management service, or it can be provided by the IoT device to be configured through a secure communication process after the smart agent on the APP establishes a near-field communication connection with the IoT device to be configured. The session key can be generated by at least two of the cloud smart agent, the smart agent on the APP, and the IoT device to be configured through a key negotiation process.

[0074] In some embodiments, encryption processing can be performed by the cloud-based intelligent agent calling a security encryption service. In this case, after the user inputs the network credentials, the APP-based intelligent agent can submit the network credentials to the cloud-based intelligent agent's security processing module. The cloud-based intelligent agent obtains the corresponding device public key or session key based on the device identity information of the IoT device to be configured, and uses this device public key or session key to encrypt the network credentials, generating an encrypted credential. Subsequently, the cloud-based intelligent agent sends the encrypted credential to the APP-based intelligent agent, which then continues to perform the pass-through operation.

[0075] In other embodiments, encryption can also be performed within a secure execution environment on the APP-side intelligent agent's local machine. Specifically, the APP-side intelligent agent can obtain the device public key or temporary session key corresponding to the IoT device to be configured from the cloud-based intelligent agent, and encrypt the network credentials entered by the user in a protected security module to generate encrypted credentials. This method reduces the risk of plaintext network credentials being exposed in the ordinary business logic layer.

[0076] Furthermore, after generating the encrypted credential, the smart agent on the APP can transmit the encrypted credential to the IoT device to be configured via a near-field communication link established with it. This near-field communication link can be a Bluetooth communication link, a SoftAP communication link, an NFC communication link, or other short-range data transmission links. When using a Bluetooth communication link, the smart agent on the APP can write the encrypted credential into the preset Bluetooth signature value of the IoT device to be configured; when using a SoftAP communication link, the smart agent on the APP can send the encrypted credential to the preset data receiving interface of the IoT device to be configured via a local area network interface.

[0077] In some embodiments, during the transmission of encrypted credentials, the smart agent on the APP side is only responsible for receiving the encrypted credentials, determining the near-field communication channel, and writing the encrypted credentials to the IoT device to be configured. It does not decrypt the encrypted credentials, nor does it persistently store the plaintext data of the network credentials locally. Furthermore, the smart agent on the APP side can clear the input cache, temporary variables, and session data after the credentials are written to reduce the risk of network credential leakage.

[0078] In some embodiments, the network credential writing step node in the structured distribution network task plan may include a credential input action, an encryption processing action, a communication channel selection action, a credential writing action, and a write result feedback action. The APP-side intelligent agent can sequentially execute the above actions according to the configuration of this step node, and receive the write result returned by the IoT device to be distributed after completing the credential writing. The write result may include status information such as successful reception, verification failure, write timeout, and device disconnection. The APP-side intelligent agent can send the write result back to the cloud-side intelligent agent so that the cloud-side intelligent agent can determine whether to enter the activation waiting step or trigger an exception handling process.

[0079] For example, when the IoT device to be configured is a camera that only supports Bluetooth configuration, the cloud-based smart agent can determine from the device knowledge base that the camera's credential writing method is Bluetooth signature value writing, and obtain the corresponding device public key. After the user enters the home Wi-Fi name and password, the cloud-based smart agent or the app-based smart agent encrypts the Wi-Fi name and password based on the device public key, generating an encrypted credential; the app-based smart agent then writes this encrypted credential to the camera's specified Bluetooth signature value via the Bluetooth communication link. After receiving the encrypted credential, the camera can decrypt it on the device side to obtain network connection information for accessing the target network.

[0080] S106. The IoT device to be configured decrypts the encryption credential to obtain network connection information, and after accessing the target network based on the network connection information, sends an activation request to the cloud intelligent agent; the cloud intelligent agent verifies the activation request, and after the verification is successful, completes the binding of the IoT device to be configured with the user account.

[0081] In practice, after receiving the encrypted credentials transmitted by the smart agent on the APP, the IoT device to be configured can decrypt the credentials through its on-device security decryption module to obtain network connection information. This network connection information may include at least one of the following: wireless network identifier, wireless network password, network authentication parameters, configuration token, and cloud connection address. The wireless network identifier and wireless network password are used to enable the IoT device to access the target network, while the configuration token is used to prove the legitimacy of the configuration task to the cloud smart agent after the device accesses the network.

[0082] Specifically, the IoT device to be configured can decrypt the encrypted credentials using its own pre-set device private key, pre-set key, or session key. When the encrypted credentials are generated based on the device public key, the IoT device to be configured can decrypt them using its own stored device private key; when the encrypted credentials are generated based on the session key, the IoT device to be configured can decrypt them using the session key negotiated during the configuration process. After decryption, the IoT device to be configured can perform integrity and format verification on the decrypted network connection information to determine whether the network connection information is usable.

[0083] In some embodiments, if the IoT device to be configured fails to decrypt the encrypted credential, or the decrypted network connection information does not meet the format requirements, the IoT device to be configured can return status information such as decryption failure, invalid credential, or incorrect format to the smart agent on the APP via the near-field communication link. The smart agent on the APP can send this status information back to the smart agent on the cloud, so that the smart agent on the cloud can determine whether to regenerate the encrypted credential, prompt the user to re-enter the network credential, or trigger an exception handling process.

[0084] Here, after the network connection information verification is successful, the IoT device to be configured can access the target network based on the network connection information. Taking a Wi-Fi network as an example, the IoT device to be configured can search for the target wireless network based on the wireless network identifier and use the wireless network password to authenticate and connect. After successful connection, the IoT device to be configured can obtain the local area network address and establish a communication connection with the Internet or cloud server. During the process of accessing the target network, the IoT device to be configured can also detect the network connection status and return the status of successful connection, connection failure, authentication failure, and timeout failure to connect to the APP-side smart agent through the near-field communication link.

[0085] In some embodiments, after the IoT device to be configured connects to the target network, it can send an activation request to the cloud-based intelligent agent. The activation request may include at least one of the following: a configuration token, device identity information, device model information, device public key identifier, firmware version information, network connection status, and identifier of this configuration task.

[0086] Among them, the device identity information can be the device serial number, device unique identifier, chip identifier, certificate identifier, or other information that can characterize the device identity; the network distribution token can be a temporary credential generated by the cloud intelligent agent in this network distribution process and associated with the user account, the network distribution task to be distributed, and the device characteristic information.

[0087] Furthermore, after receiving an activation request from the IoT device to be configured in the network, the cloud-based intelligent agent can verify the activation request. Specifically, the cloud-based intelligent agent can verify whether the configuration token is valid, expired, associated with the current user account, and corresponds to the current structured network configuration task plan, as well as whether the device identity information matches the legitimate device records in the device knowledge base, device certificate library, or device management platform. The cloud-based intelligent agent can also verify whether the model information, communication capability information, or firmware version information currently reported by the device is consistent with the aforementioned device characteristic information to avoid incorrect binding of non-target devices.

[0088] In some embodiments, after the activation request verification is successful, the cloud-based intelligent agent can complete the binding between the IoT device to be configured and the user account. Specifically, the cloud-based intelligent agent can establish a binding relationship between the user account and the IoT device to be configured in the device management database, and record at least one of the following: device name, device category, device model, device location, device communication identifier, binding time, and configuration task identifier. Furthermore, the cloud-based intelligent agent can also update the device shadow status corresponding to the IoT device to be configured, updating the device status to activated, connected to the network, or available, and push the successful configuration result to the intelligent agent on the APP.

[0089] In some embodiments, after receiving a successful network configuration result pushed by the cloud-based smart agent, the app-side smart agent can display a configuration completion notification on the network configuration interface and proceed to the configuration completion node according to the structured network configuration task plan. For example, the app-side smart agent can notify the user that "the device has successfully connected to the network and bound to the current account," and display the device name, the room it belongs to, or the entry point for subsequent scenario configuration. If the cloud-based smart agent finds that the network configuration token is invalid, the device identity does not match, or the device has been bound to another account during the activation verification process, it can generate a corresponding activation failure reason and notify the user or trigger an exception handling process through the app-side smart agent.

[0090] In this way, after receiving the encrypted credentials, the IoT device to be configured only decrypts and uses the network connection information on the device side, and actively initiates an activation request to the cloud intelligent agent after successfully accessing the target network. The cloud intelligent agent verifies the configuration token and the device identity information before performing account binding, which can ensure that there is a corresponding relationship between the device's network connection behavior, activation behavior and user account binding behavior, thereby improving the security and reliability of the IoT device configuration binding process.

[0091] See Figure 2 The diagram shows a flowchart of an IoT device network configuration method based on a large-scale edge-cloud combined model provided in this disclosure. The method includes steps S201 to S204, wherein: S201. When an abnormal event is detected during the distribution network process, abnormal information corresponding to the abnormal event is obtained. The abnormal information includes at least one of the following: distribution network error code, equipment model information, equipment communication capability information, current network environment information, and historical similar case information.

[0092] S202. Organize the abnormal information into abnormal diagnosis prompts and input them into the large language model.

[0093] S203. Using the large language model, perform fault root cause reasoning on the distribution network anomaly event to generate anomaly root cause information and corresponding natural language repair suggestions.

[0094] S204. The natural language repair suggestion is sent to the smart agent on the APP to guide the user to perform the anomaly repair operation.

[0095] In practical implementation, during the process of the intelligent agent on the APP executing network distribution guidance and device interaction according to the structured network distribution task plan, the cloud-based intelligent agent can detect whether network distribution anomalies have occurred based on the current network distribution execution status returned by the intelligent agent on the APP, the device-side return status, and the cloud activation status. These network distribution anomalies may include at least one of the following: device scan timeout, no target device detected, multiple suspected devices detected, authorization failure, network credential writing failure, device decryption failure, target network connection failure, device network timeout, device activation failure, and device binding failure.

[0096] Here, when a distribution network anomaly event is detected, the cloud-based intelligent agent can obtain the anomaly information corresponding to the event. The anomaly information may include at least one of the following: distribution network error code, device model information, device communication capability information, current network environment information, and historical similar case information.

[0097] Among them, the network configuration error code can be reported by the smart agent on the APP or the IoT device to be configured, and is used to characterize the type of abnormality; the device model information can be used to determine the network configuration rules and capability limitations corresponding to the IoT device to be configured; the device communication capability information can include whether the device supports Bluetooth network configuration, whether it supports SoftAP network configuration, whether it only supports 2.4GHz wireless network, whether it supports 5GHz wireless network, whether it supports specific Bluetooth service feature values, etc.; the current network environment information can include the current Wi-Fi name, wireless network frequency band, network signal strength, whether dual-band merging exists, the current mobile network connection status, etc.; the historical similar case information can include the causes of failure and repair methods of similar devices in similar error codes and similar network environments.

[0098] In some embodiments, abnormal information can come from multiple data sources. For example, the APP-based intelligent agent can report the current network configuration step node, user operation results, permission status, Bluetooth scan results, SoftAP connection results, candidate device list, credential writing status, and activation waiting time; IoT devices to be configured can report device status codes, network connection failure reasons, credential verification results, or device restart status via near-field communication links; the cloud-based intelligent agent can query device models, device certificate status, device activation status, network configuration token status, and historical network configuration records based on the device management platform. The cloud-based intelligent agent can integrate the above information to form an abnormal information set corresponding to the current abnormal network configuration event.

[0099] In some embodiments, the cloud-based intelligent agent can organize the abnormal information into abnormal diagnostic prompts and input them into a large language model. The abnormal diagnostic prompts may include a description of the current abnormal event, the current network configuration step node, state changes before and after the abnormality, device network configuration characteristics, network environment information, historical similar cases, and the result format required for the large language model to output. By organizing the abnormal information into abnormal diagnostic prompts, the large language model can perform fault analysis under clear context and output constraints, avoiding the generation of only generalized error messages.

[0100] For example, when an anomaly occurs during the device scanning phase where the target device is not detected, the anomaly diagnosis message may include: the current step is the device scanning step; the APP has not detected a candidate device matching the device's feature information; the IoT device to be paired with is a certain model of camera; this device supports Bluetooth pairing; the indicator light should flash rapidly when the device enters the pairing state; and the user has not yet confirmed the indicator light status. The large language model can infer possible causes based on the above information, such as the device not entering the pairing state, Bluetooth permission not being enabled, the phone and device being too far apart, or the device being bound to another account, and generate corresponding natural language repair suggestions.

[0101] For example, when a timeout occurs during the activation waiting phase, the anomaly diagnosis message may include: the current step is the activation waiting step; the IoT device to be configured has received the encryption certificate but has not sent an activation request to the cloud smart agent within the preset time; the device model information indicates that the device only supports 2.4GHz wireless networks; the current network environment information indicates that the user's router may have dual-band merging enabled; and historical similar cases indicate that such devices may not be able to recognize connectable 2.4GHz networks in a dual-band merging environment. The large language model can infer from the above information that the root cause of the anomaly may be a mismatch in the target network frequency band or dual-band merging preventing the device from accessing the network, and generate repair suggestions prompting the user to disable dual-band merging, switch to the 2.4GHz network, or re-enter the 2.4GHz network certificate.

[0102] In some embodiments, after performing root cause inference on abnormal network events, the large language model can output abnormal root cause information and corresponding natural language repair suggestions. The abnormal root cause information represents the possible causes of the fault inferred by the large language model based on the abnormal information, and may include at least one of cause category, cause description, confidence level, and supporting data. The natural language repair suggestions are used to explain executable repair operations to the user, and may include suggestions such as device status check, permission granting, proximity to the device, rescanning, switching network configuration methods, modifying router settings, re-entering network credentials, and waiting for the device to restart.

[0103] In some embodiments, the cloud-based intelligent agent can perform security and executability checks on the natural language repair suggestions output by the large language model. Specifically, it can determine whether the natural language repair suggestions involve unauthorized operations, whether they require users to provide unnecessary sensitive information, whether they conflict with the device's network configuration characteristics, and whether they can be guided and executed by the current app-side intelligent agent. For repair suggestions that do not meet security or executability requirements, the cloud-based intelligent agent can filter, rewrite, or re-request generation from the large language model.

[0104] After generating natural language repair suggestions, the cloud-based intelligent agent can send these suggestions to the app-based intelligent agent. Upon receiving the suggestions, the app-based intelligent agent can output them to the user through at least one of the following methods: network configuration interface display, pop-up prompts, step cards, voice broadcast, or graphic instructions, to guide the user in performing the anomaly repair operation. For example, the app-based intelligent agent might prompt the user, "Please check if the camera indicator light is flashing rapidly. If not, please press and hold the device button for 5 seconds to enter network configuration mode"; or it might prompt the user, "The current device may only support 2.4GHz networks. It is recommended to temporarily disable the router's dual-band merging function and then reconfigure the network."

[0105] In some embodiments, the app-based smart agent can also collect repair confirmation results after the user performs a repair operation and feed the repair confirmation results back to the cloud-based smart agent. The cloud-based smart agent can determine, based on the repair confirmation results, whether to continue executing the original structured distribution network task plan, re-execute the current distribution network step node, or generate a revised structured distribution network task plan. For example, after the user confirms that the dual-band merging function has been disabled, the cloud-based smart agent can instruct the app-based smart agent to return to the network credential input step or re-execute the device networking step.

[0106] In some embodiments, after the cloud-based intelligent agent generates root cause information based on the anomaly information, it can further generate task plan adjustment instructions based on the root cause information. These task plan adjustment instructions are used to adjust the original structured network configuration task plan, such as inserting an anomaly repair step node into the original process, or adjusting subsequent retry strategies. The anomaly repair step node can be used to prompt the user to check the device's network configuration status, re-enable the device's Bluetooth, switch to a 2.4GHz wireless network, disable the router's dual-band merging function, or re-enter network credentials, etc.

[0107] Here, upon receiving the task plan adjustment instruction, the cloud-based intelligent agent can partially modify the original structured distribution network task plan based on the instruction, resulting in a revised structured distribution network task plan. For example, when the root cause information indicates that the device may fail to connect to the network due to lack of 5GHz network support, a repair step of "switching to 2.4GHz network" or "retrying after disabling dual-band merging" can be inserted into the original task plan, and the subsequent steps can be adjusted to rewrite the network credentials and wait for the device to activate again.

[0108] In some embodiments, before sending the revised structured distribution network task plan to the smart agent on the APP, the cloud-based smart agent can also perform a security check on the revised structured distribution network task plan to determine whether it contains unauthorized operations, sensitive information acquisition operations, or operations that are incompatible with the device's distribution network capabilities. If the security check passes, the revised structured distribution network task plan is sent to the smart agent on the APP, allowing the smart agent on the APP to continue executing the distribution network guidance according to the new task plan.

[0109] In some embodiments, after the IoT device to be configured has completed network configuration and been bound to the user account, the cloud-based intelligent agent can generate the device name and spatial affiliation information corresponding to the IoT device to be configured based on the device location slot and device category slot in the natural language network configuration request. For example, when the user expresses "living room camera" in the network configuration request, the device that has completed network configuration can be automatically named "living room camera" and assigned to the "living room" group in the user's home space.

[0110] Furthermore, the cloud-based intelligent agent can update the device shadow status or device management information based on the generated device name and space ownership information, enabling the APP to display the device's latest binding status, space ownership, and online status.

[0111] In some embodiments, the cloud-based intelligent agent can also invoke a large language model to generate intelligent scene recommendation information related to the IoT device based on network configuration intent information, device category, device location, and user's historical usage habits, and push the intelligent scene recommendation information to the intelligent agent on the APP. For example, after the camera is configured, intelligent scenes such as arming when leaving home, turning off monitoring when returning home, or abnormal movement alerts can be recommended to the user. This allows for further integration of device usage configuration after network configuration is completed, improving the efficiency of IoT device configuration and the intelligent experience.

[0112] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0113] Based on the same inventive concept, this disclosure also provides an IoT device configuration device for an IoT device configuration model that corresponds to the IoT device configuration method of the edge-cloud combined big model. Since the principle of the device in this disclosure for solving the problem is similar to the IoT device configuration method of the edge-cloud combined big model described above in this disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0114] Please see Figure 3 , Figure 3 This is a schematic diagram of an IoT device network configuration device based on a large-scale edge-cloud combined model, provided as an embodiment of this disclosure. Figure 3 As shown in the embodiments of this disclosure, the IoT device configuration device 300 with a combined edge-cloud and large-scale model includes: The data acquisition module 310 is used to acquire the natural language network configuration request input by the user for the IoT device to be configured, and send the natural language network configuration request and user context information to the cloud intelligent agent.

[0115] The information recognition module 320 is used to call the large language model through the cloud intelligent agent to perform intent recognition and slot filling on the natural language network distribution request, and obtain network distribution intent information and device feature information.

[0116] The plan generation module 330 is used to generate a structured distribution network task plan that matches the IoT device to be distributed to the network based on the distribution network intent information, the device feature information and the preset device knowledge base. The structured distribution network task plan includes multiple distribution network step nodes and execution actions and user guidance information corresponding to each distribution network step node.

[0117] The guidance control module 340 is used to send the structured distribution network task plan to the APP-side intelligent agent, so that the APP-side intelligent agent executes the distribution network guidance according to the structured distribution network task plan, and interacts with the IoT device to be distributed through the near-field communication link.

[0118] The encryption processing module 350 is used to encrypt the network credential entered by the user when the network credential writing step is executed, generate an encrypted credential, and transmit the encrypted credential to the IoT device to be configured through the APP terminal smart agent.

[0119] The activation binding module 360 ​​is used to decrypt the encryption credential through the IoT device to be configured, obtain network connection information, and send an activation request to the cloud intelligent agent after accessing the target network based on the network connection information; the cloud intelligent agent verifies the activation request, and completes the binding of the IoT device to be configured and the user account after the verification is successful.

[0120] The processing flow of each module in the device and the interaction flow between each module can be referred to the relevant descriptions in the above method embodiments, and will not be detailed here.

[0121] This disclosure provides an IoT device configuration method and apparatus that combines edge and cloud computing with a large-scale model. The method involves a cloud-based intelligent agent calling a large language model to parse the user's natural language configuration request, generating a structured configuration task plan based on the device's configuration characteristics, and then having the intelligent agent on the APP interact with the device according to this task plan. Simultaneously, edge-cloud collaborative control is achieved during network credential writing, device activation and binding, and anomaly handling. This reduces manual selection and repeated trial-and-error operations by the user, allowing the configuration process to dynamically adjust based on device type, user context, and on-site anomaly conditions, thereby improving the convenience, security, anomaly recovery capabilities, and configuration success rate of IoT device configuration.

[0122] Corresponding to Figure 1 and Figure 2 In this disclosure, an electronic device 400 is also provided, which is a method for configuring IoT devices using a combination of edge-cloud and large-scale models. Figure 4 The diagram shown is a structural schematic of an electronic device 400 provided in an embodiment of this disclosure, including: Processor 41, memory 42, and bus 43; memory 42 is used to store execution instructions, including main memory 421 and external memory 422; the main memory 421, also called internal memory, is used to temporarily store the computational data in processor 41, as well as the data exchanged with external memory 422 such as hard disk. Processor 41 exchanges data with external memory 422 through main memory 421. When the electronic device 400 is running, processor 41 and memory 42 communicate through bus 43, enabling processor 41 to execute... Figure 1 and Figure 2 The steps of the IoT device network configuration method that combines edge cloud and large-scale model.

[0123] This disclosure also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program performs the steps of the IoT device network configuration method for a combined edge-cloud and large-scale model described in the above method embodiments. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0124] This disclosure also provides a computer program product, which includes computer instructions. When the computer instructions are executed by a processor, they can perform the steps of the IoT device network configuration method with end-to-cloud combined big model described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0125] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A method for configuring IoT devices using a large-scale model combining edge and cloud computing, characterized in that, include: Obtain the natural language network configuration request input by the user for the IoT device to be configured, and send the natural language network configuration request and user context information to the cloud intelligent agent; The cloud-based intelligent agent invokes a large language model to perform intent recognition and slot filling on the natural language network configuration request, thereby obtaining network configuration intent information and device feature information. Based on the network distribution intent information, the device characteristic information, and a preset device knowledge base, a structured network distribution task plan matching the IoT device to be distributed is generated. The structured network distribution task plan includes multiple network distribution step nodes and corresponding step types, execution actions, user guidance information, execution conditions, status feedback requirements, and exception branch rules for each node. The step type identifies the functional category of the current step in the network distribution process. The execution action indicates the local capabilities that the APP-side smart agent needs to invoke or the data interaction that needs to be performed with the device. The user guidance information explains the operations that need to be performed in the current step to the user. The execution conditions determine whether the current step is completed and whether the next step can proceed. The status feedback requirements indicate the execution status that the APP-side smart agent needs to report to the cloud-side smart agent. The exception branch rules trigger corresponding exception handling logic when an exception occurs in the current step. In the structured distribution network task plan, each distribution network step node is arranged in a preset order or contains conditional jump relationships so that the APP-side intelligent agent can switch subsequent steps according to the actual execution status. The structured distribution network task plan is sent to the smart agent on the APP, so that the smart agent on the APP can execute the distribution network guidance according to the structured distribution network task plan and interact with the IoT device to be distributed through the near field communication link; When executing the network credential writing step, the network credential input by the user is encrypted to generate an encrypted credential, and the encrypted credential is transmitted to the IoT device to be configured through the smart agent of the APP. The IoT device to be configured decrypts the encryption credential to obtain network connection information, and after accessing the target network based on the network connection information, sends an activation request to the cloud intelligent agent; the cloud intelligent agent verifies the activation request, and after the verification is successful, completes the binding of the IoT device to be configured with the user account.

2. The method according to claim 1, characterized in that, The natural language network distribution request is subjected to intent recognition and slot filling to obtain network distribution intent information and device characteristic information, specifically including: Identify at least one network configuration intent from the natural language network configuration request: adding a device, reconfiguring the network, or connecting the device to the network; Extract at least one slot information from the natural language network configuration request, including equipment category, equipment location, equipment appearance features, equipment model, and equipment status; The slot information is matched with the user's historical device binding records, user home space information, and device knowledge base to determine the device feature information corresponding to the IoT device to be configured.

3. The method according to claim 1, characterized in that, Based on the network distribution intent information, the device feature information, and a preset device knowledge base, a structured network distribution task plan matching the IoT device to be distributed to is generated, specifically including: Based on the device feature information, the network configuration capability information of the IoT device to be configured is retrieved from the preset device knowledge base. The network configuration capability information includes at least one of the following: supported near-field communication methods, supported wireless network frequency bands, network configuration status indication methods, and activation conditions. The network distribution path corresponding to the IoT device to be distributed is determined based on the network distribution capability information. A structured network distribution task plan is generated based on the network distribution path, including at least some of the steps in the following steps: status confirmation, permission request, device scanning, target confirmation, credential input, credential writing, activation waiting, and configuration completion.

4. The method according to claim 1, characterized in that, The APP-based intelligent agent executes network distribution guidance according to the structured network distribution task plan, specifically including: When executing each distribution network step node, the current distribution network execution status is collected and the current distribution network execution status is fed back to the cloud intelligent agent; The cloud-based intelligent agent invokes the large language model to dynamically generate the current user guidance script based on the current network configuration execution status and historical interaction context. The current user guidance script is sent to the intelligent agent on the APP and output to the user through interface display or voice broadcast.

5. The method according to claim 1, characterized in that, Data interaction with the IoT device to be configured via a near-field communication link specifically includes: The APP-based intelligent agent invokes the Bluetooth communication link or SoftAP communication link to scan for nearby candidate IoT devices. If multiple candidate IoT devices are detected, the list of candidate IoT devices is sent to the cloud-based intelligent agent; The cloud-based intelligent agent filters and sorts multiple candidate IoT devices based on user historical binding records, device feature information, and candidate IoT device signal information, and generates target device confirmation information. The IoT device to be configured is determined based on the user's confirmation operation regarding the target device confirmation information.

6. The method according to claim 1, characterized in that, The user-input network credentials are encrypted to generate encrypted credentials, specifically including: Obtain the device public key or session key corresponding to the IoT device to be configured on the network; Based on the device public key or session key, the wireless network identifier and wireless network password entered by the user are encrypted to obtain the encryption credential; The APP-side intelligent agent does not store or parse the wireless network password locally, and transmits the encrypted credentials in ciphertext form to the designated communication characteristic value or designated data receiving interface of the IoT device to be configured.

7. The method according to claim 1, characterized in that, The method further includes: When an abnormal event is detected during the distribution network process, abnormal information corresponding to the abnormal event is obtained. The abnormal information includes at least one of the following: distribution network error code, equipment model information, equipment communication capability information, current network environment information, and historical similar case information. The abnormal information is organized into abnormal diagnosis prompts and input into the large language model; The large language model is used to perform fault root cause reasoning on the abnormal events in the distribution network, generating abnormal root cause information and corresponding natural language repair suggestions. The natural language repair suggestions are sent to the smart agent on the APP to guide the user to perform the anomaly repair operation.

8. The method according to claim 7, characterized in that, The method further includes: The large language model generates task plan adjustment instructions based on the anomaly root cause information. Based on the task plan adjustment instruction, a repair step node is inserted into the original structured distribution network task plan or the subsequent retry strategy is adjusted to obtain the corrected structured distribution network task plan. The revised structured distribution network task plan is subjected to security verification, and after the security verification is passed, it is sent to the APP-side intelligent agent, so that the APP-side intelligent agent continues to execute the distribution network process according to the revised structured distribution network task plan.

9. The method according to claim 1, characterized in that, The method further includes: After the IoT device to be configured to the network completes the configuration and is bound to the user account, the device name and space ownership information corresponding to the IoT device to be configured are generated based on the device location slot and device category slot in the natural language configuration request. Update the device shadow status based on the device name and spatial ownership information; The large language model generates intelligent scene recommendation information related to the IoT device to be configured based on the network intent information, and pushes the intelligent scene recommendation information to the intelligent agent on the APP.

10. A network configuration device for IoT devices using a combined edge-cloud and large-scale model, characterized in that, include: The data acquisition module is used to acquire the natural language network configuration request input by the user for the IoT device to be configured, and send the natural language network configuration request and user context information to the cloud intelligent agent; The information recognition module is used to call the large language model through the cloud intelligent agent to perform intent recognition and slot filling on the natural language network distribution request, and obtain network distribution intent information and device feature information. The plan generation module is used to generate a structured distribution network task plan matching the IoT device to be distributed, based on the distribution network intent information, the device feature information, and a preset device knowledge base. The structured distribution network task plan includes multiple distribution network step nodes and corresponding step types, execution actions, user guidance information, execution conditions, status feedback requirements, and exception branch rules for each step node. The step type identifies the functional category of the current step in the distribution network process. The execution action indicates the local capabilities that the APP-side smart agent needs to invoke or the data interaction that needs to be performed with the device. The user guidance information explains the operations that need to be performed in the current step to the user. The execution conditions determine whether the current step is completed and whether the next step can proceed. The status feedback requirements indicate the execution status that the APP-side smart agent needs to report to the cloud-side smart agent. The exception branch rules trigger corresponding exception handling logic when an exception occurs in the current step. In the structured distribution network task plan, each distribution network step node is arranged in a preset order or contains conditional jump relationships so that the APP-side intelligent agent can switch subsequent steps according to the actual execution status. The guidance control module is used to send the structured distribution network task plan to the APP-side intelligent agent, so that the APP-side intelligent agent can execute the distribution network guidance according to the structured distribution network task plan and interact with the IoT device to be distributed through the near-field communication link; The encryption processing module is used to encrypt the network credentials input by the user when the network credentials writing step is executed, generate encrypted credentials, and transmit the encrypted credentials to the IoT device to be configured through the smart agent of the APP. The activation binding module is used to decrypt the encryption credential through the IoT device to be configured, obtain network connection information, and send an activation request to the cloud intelligent agent after accessing the target network based on the network connection information; the cloud intelligent agent verifies the activation request, and completes the binding between the IoT device to be configured and the user account after the verification is successful.

Images