A method of controlling a device through an AI dialogue and reasoning large model

Abstract

The application discloses a kind of by AI dialogue and reasoning big model control equipment method, including the following steps: utilize enhanced output restriction module, using intelligent output monitoring module, by general protocol conversion framework, with efficient data input module, backup and switching mechanism of model are constructed.The through AI dialogue and reasoning big model control equipment method, system stability is improved, introduce backup and switching mechanism of model, periodically backup AI big model, when main model appears fault, performance drops or is attacked, can be quickly switched to backup model, this ensures the continuity of equipment control process, avoids because big model anomaly leads to equipment control failure, with smart home system as example, even if main model encounters data anomaly or network attack, backup model can replace work in time, maintain the normal operation of household appliance equipment, guarantee the life of user is not affected, improve the reliability and stability of entire system.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method for controlling devices through AI dialogue and reasoning large models. Background Technology

[0002] With the rapid development of artificial intelligence technology, AI dialogue models have achieved remarkable results in many software application fields such as intelligent customer service, content creation, and knowledge Q&A, thanks to their powerful language understanding and generation capabilities. They have become a powerful assistant in people's work and life. However, in the field of hardware control, the application of AI dialogue models faces many challenges.

[0003] Currently, large-scale AI dialogue models are essentially software functions trained on massive amounts of data. Internally, they involve complex computational processes, but externally they appear as black-box systems of input and output, lacking the ability to directly control external hardware devices. Although some methods have been attempted to enable large-scale AI dialogue models to control hardware, they generally suffer from the following problems:

[0004] Reliance on a single model and poor stability: Existing solutions rely too heavily on large AI models. If the large model fails, is attacked, or its performance degrades, the entire hardware control system will be paralyzed. For example, in smart homes, it may incorrectly control home appliances, leading to security risks.

[0005] The instruction format is limited and lacks flexibility: it mostly uses fixed prompts and label formats to identify instructions, which makes it difficult to adapt to complex and diverse equipment control needs, and is prone to misidentification. Even slight changes in the user's expression may lead to control failure.

[0006] Protocol conversion is difficult and lacks versatility: Although the protocol conversion module has some customization capabilities, it is difficult to adapt to a wide variety of complex device communication protocols, resulting in high development costs, poor versatility and scalability, especially in the field of industrial control.

[0007] Inefficient data processing and poor real-time performance: Data goes through many processing stages between devices, prompts, and large models, and parsing and encapsulation are complex. In scenarios with high real-time requirements, such as intelligent transportation and industrial production, it is unable to respond to control needs in a timely manner.

[0008] In summary, overcoming these problems in existing technologies and achieving efficient, stable, flexible, and universal control of hardware devices by AI dialogue models has become a key technical challenge that urgently needs to be addressed in the current intersection of artificial intelligence and the Internet of Things. Based on this background, this invention aims to propose an optimized solution to promote the widespread application of AI in the field of hardware control. Summary of the Invention

[0009] (a) Technical problems to be solved

[0010] To address the shortcomings of existing technologies, this invention provides a method for controlling devices through AI dialogue and reasoning large-scale models. This method offers advantages such as improved system stability, enhanced command recognition flexibility, improved protocol adaptability and universality, optimized data processing efficiency and real-time performance, and expanded application scenario possibilities. It solves the problems of insufficient stability, flexibility, universality, and real-time performance in current AI dialogue and reasoning large-scale model hardware control.

[0011] (II) Technical Solution

[0012] To achieve the above objectives, the present invention provides the following technical solution: a method for controlling a device through an AI dialogue and reasoning large model, comprising the following steps:

[0013] S1. Utilizing the enhanced output restriction module, the AI ​​dialogue and reasoning model is guided to output content that meets the requirements through a diversified prompt word library and semantic understanding technology. The diversified prompt word library provides multiple instruction guidance methods according to different device control scenarios and user habits. The semantic understanding technology incorporates a description of the instruction semantics into the prompt words.

[0014] S2. An intelligent output monitoring module is adopted, which uses machine learning classification technology and natural language processing technology to perform multi-dimensional analysis on the output of the AI ​​dialogue and reasoning model, and to identify and parse control commands.

[0015] S3. Through a general protocol conversion framework, the parsed control instructions are converted into the control protocols of the corresponding devices. The general protocol conversion framework includes a protocol parsing layer, a data adaptation layer, and an interface layer. The protocol parsing layer is used to parse the communication protocols of different devices and convert them into a unified internal data format. The data adaptation layer adapts the data according to the device control requirements and data format specifications. The interface layer provides a unified calling interface.

[0016] S4. With the help of the high-efficiency data input module, asynchronous processing and parallel computing technology are used to collect and process device data, and encapsulate it into prompt words to feed back to the AI ​​dialogue and reasoning model. The asynchronous processing is used to realize real-time data collection and caching, and the parallel computing technology is used to accelerate data parsing and encapsulation.

[0017] S5. Construct a model backup and switching mechanism to regularly back up the large AI dialogue and inference model. When the main model fails, its performance degrades, or it is attacked, it will automatically switch to the backup model through the model switching module to ensure the continuity of equipment control.

[0018] Preferably, in a smart home scenario, the enhanced output limiting module automatically selects appropriate prompts based on the user's past usage habits and device type. If the user frequently uses colloquial commands to control the device, then a specific format of colloquial prompts is selected.

[0019] Preferably, the intelligent output monitoring module uses a trained convolutional neural network model to classify the model output text in real time. After recognizing the control command, it uses a natural language processing tool to parse the device, operation, and parameter information in the command.

[0020] Preferably, under the general protocol conversion framework, users can develop corresponding protocol conversion plugins according to the device's communication protocol. For smart devices using the MQTT protocol, an MQTT protocol parser plugin can be developed to realize the conversion of instruction and data formats.

[0021] Preferably, the high-efficiency data input module uses an asynchronous data acquisition interface to connect to the device at the data acquisition end, collects data in real time and caches it in memory, and uses multi-threading technology to process the collected data in parallel at the data processing end, quickly parses and encapsulates it into prompt word format and then transmits it to the AI ​​dialogue and reasoning big model.

[0022] Preferably, the model backup and switching mechanism backs up the AI ​​dialogue and reasoning big model every day. The backup data is stored on a remote server. When the main model is abnormal, the model switching module completes the model switching within 1 second and starts the repair and optimization process of the main model. After the repair is completed, it is used as the main model again.

[0023] (III) Beneficial Effects

[0024] Compared with existing technologies, this invention provides a method for controlling devices through AI dialogue and reasoning large models, which has the following beneficial effects:

[0025] 1. This method of controlling devices through AI dialogue and reasoning models improves system stability. It introduces a model backup and switching mechanism, regularly backing up the AI ​​model. When the main model malfunctions, experiences performance degradation, or is attacked, it can quickly switch to the backup model. This ensures the continuity of device control and avoids device control failure due to abnormalities in the main model. Taking a smart home system as an example, even if the main model encounters data anomalies or network attacks, the backup model can take over in time to maintain the normal operation of home appliances, ensuring that users' lives are not affected and improving the reliability and stability of the entire system.

[0026] 2. This method of controlling devices through AI dialogue and reasoning models enhances the flexibility of command recognition. The enhanced output restriction module utilizes a diverse prompt vocabulary and semantic understanding technology to make the AI ​​model output more aligned with complex control needs. The intelligent output monitoring module employs machine learning classification and natural language processing technologies to analyze model output from multiple dimensions. This enables the system to more accurately identify and parse various forms of control commands, whether colloquial or complex, and to process them precisely. For example, in a smart office scenario, users can give commands in more natural language, such as "dim the lights in the meeting room and turn on the projector," which the system can accurately understand and execute, greatly improving the flexibility and accuracy of command recognition.

[0027] 3. This method of controlling devices through AI dialogue and reasoning models improves protocol adaptability and versatility. The universal protocol conversion framework includes a protocol parsing layer, a data adaptation layer, and an interface layer. Users can quickly develop conversion plugins for different protocols based on this framework, which greatly reduces the difficulty of adapting to new protocols. This enables the system to easily connect to devices with various communication protocols on the market, whether it is the common MQTT and HTTP protocols or the niche serial port proprietary protocols. In the field of industrial IoT, this framework can help enterprises quickly integrate devices of different brands and protocols, achieve unified management and control, reduce device access costs, and improve the versatility and scalability of the system.

[0028] 4. This method of controlling equipment through AI dialogue and reasoning models optimizes data processing efficiency and real-time performance. The high-efficiency data input module adopts asynchronous processing and parallel computing technology. The asynchronous data acquisition interface enables real-time data acquisition and caching, while parallel computing accelerates data parsing and encapsulation. This optimizes the flow and processing of data between various modules, reduces data processing time, and improves data processing efficiency. In intelligent transportation systems with extremely high real-time requirements, the system can quickly collect traffic equipment data, analyze it, and issue control commands to adjust traffic lights in a timely manner, alleviate traffic congestion, and meet real-time requirements.

[0029] 5. This method of controlling devices through AI dialogue and reasoning models expands the application scenarios. Combining the above advantages, this technical solution breaks through many limitations between AI and hardware, opening up a wider space for the application of AI dialogue models in the field of hardware control. In addition to common smart home and smart office fields, it can also be applied to many emerging fields such as smart medical equipment control, smart agricultural environmental regulation, and smart factory production automation, promoting the intelligent upgrading of various industries and creating more application value and business opportunities. Attached Figure Description

[0030] Figure 1This is a schematic diagram illustrating the principle of an AI dialogue model for a method of controlling devices through a large AI dialogue and reasoning model proposed in this invention.

[0031] Figure 2 This is a diagram illustrating the collaborative relationships between software modules in a method for controlling devices through AI dialogue and reasoning large-scale models proposed in this invention.

[0032] Figure 3 This is a user flow diagram illustrating a method for controlling devices through AI dialogue and reasoning large-scale models proposed in this invention.

[0033] Figure 4 This invention presents a method for controlling devices through AI dialogue and reasoning using a large model, and the data input module flowchart is shown below.

[0034] Figure 5 This invention proposes a method for controlling devices through AI dialogue and reasoning large models. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Please see Figure 1-5 A method for controlling devices through AI dialogue and reasoning large models, characterized by the following steps:

[0037] S1. Utilizing an enhanced output constraint module, the AI ​​dialogue and reasoning model is guided to output content that meets requirements through a diversified prompt word library and semantic understanding technology. The diversified prompt word library provides multiple instruction guidance methods based on different device control scenarios and user habits, while semantic understanding technology incorporates descriptions of the instruction's semantics into the prompt words. In smart home scenarios, the enhanced output constraint module automatically selects appropriate prompt words based on the user's past usage habits and device type. If the user frequently uses conversational commands to control devices, then a specific format of conversational prompt words is selected.

[0038] The system automatically selects appropriate prompt words from a diverse prompt word library based on the user's usage habits and device types in smart home scenarios. Assuming the user is accustomed to using conversational commands, the system selects prompt words that, when hearing phrases like "turn on XX" or "turn off XX," output the command format as "[Device Name] On / Off, with an identifier." For example, when the user says, "I'm home, turn on the living room light," this prompt word guides the AI ​​dialogue and reasoning model to output something like "Welcome home, the living room light has been turned on [#] Living room light on [#]." This process leverages semantic understanding technology, enabling the model to more accurately generate commands that meet the requirements, providing a foundation for subsequent device control.

[0039] S2. An intelligent output monitoring module is adopted, which uses machine learning classification technology and natural language processing technology to perform multi-dimensional analysis on the output of the AI ​​dialogue and reasoning model, and identifies and parses control commands. The intelligent output monitoring module uses a trained convolutional neural network model to classify the model output text in real time. After identifying the control commands, it uses natural language processing tools to parse the device, operation and parameter information in the commands.

[0040] After receiving the output from the AI ​​big data model, the intelligent output monitoring module first uses a trained convolutional neural network model to classify the output text in real time. When it recognizes a control command such as "turn on the living room light", it uses a natural language processing tool to parse the device (living room light) and operation (turn on) information in the command. Then, it sends the parsed command to the subsequent control module. This multi-dimensional analysis method, compared with the traditional method that only relies on fixed format matching, can more accurately identify and parse commands, effectively improve the accuracy of command recognition, and reduce misjudgments.

[0041] S3. A general protocol conversion framework is used to convert the parsed control commands into the control protocols of the corresponding devices. The general protocol conversion framework includes a protocol parsing layer, a data adaptation layer, and an interface layer. The protocol parsing layer is used to parse the communication protocols of different devices and convert them into a unified internal data format. The data adaptation layer adapts the data according to the device control requirements and data format specifications. The interface layer provides a unified calling interface. Under the general protocol conversion framework, users can develop corresponding protocol conversion plugins according to the communication protocols of the devices. For smart devices using the MQTT protocol, an MQTT protocol parser plugin is developed to realize the conversion of commands and data formats.

[0042] If the living room light uses the MQTT protocol for communication, the general protocol conversion framework comes into play. The MQTT protocol parser plugin in the protocol parsing layer converts the "living room light on" command into a message format that conforms to the MQTT protocol specification. The data adaptation layer adapts the parsed data according to the control requirements and data format specifications of the lighting equipment to ensure that the data can be correctly received and executed by the lighting equipment. The interface layer provides a unified calling interface for other modules, making it convenient for the output monitoring module to pass in the command and send back the data fed back by the lighting equipment. In this way, the system can easily adapt to devices with different protocols, greatly improving the universality and flexibility of device access.

[0043] S4. With the help of the high-efficiency data input module, asynchronous processing and parallel computing technologies are used to collect and process device data, and encapsulate it into prompt words to feed back to the AI ​​dialogue and reasoning model. The asynchronous processing is used to realize real-time data collection and caching, and the parallel computing technology is used to accelerate data parsing and encapsulation. The high-efficiency data input module connects to the device using an asynchronous data collection interface at the data collection end, collects data in real time and caches it in memory, and uses multi-threading technology to process the collected data in parallel at the data processing end, quickly parses and encapsulates it into prompt word format and then transmits it to the AI ​​dialogue and reasoning model.

[0044] During operation, the lights provide real-time feedback on their status data, such as brightness and temperature. The efficient data input module uses an asynchronous data acquisition interface to collect this data in real time and cache it in memory to avoid data loss. On the data processing end, multi-threading technology is used to process the collected data in parallel, quickly parsing and encapsulating it into specific format prompts, such as "[#] The living room light brightness is 100%, and the temperature is 30 degrees Celsius [#]". Then, these prompts are sent to the AI ​​dialogue and reasoning model to provide the model with real-time device status information so that it can make more accurate inferences and decisions.

[0045] S5. Construct a model backup and switching mechanism to regularly back up the large AI dialogue and inference model. When the main model fails, its performance degrades, or it is attacked, the model switching module automatically switches to the backup model to ensure the continuity of device control. The model backup and switching mechanism backs up the large AI dialogue and inference model daily. The backup data is stored on a remote server. When the main model is abnormal, the model switching module completes the model switching within 1 second and simultaneously initiates the repair and optimization process for the main model. After the repair is completed, it is reused as the main model.

[0046] During system operation, the model backup and switching mechanism is always under monitoring. The AI ​​dialogue and inference model is backed up daily, and the backup data is stored on a remote server to ensure data security and integrity. If the main model fails, such as due to excessive server load causing performance degradation and inability to process user commands normally, the model switching module will quickly switch the system to the backup model within 1 second. During the switching process, hot backup technology is used to ensure that the device control process is not interrupted, and users will hardly notice the model switching. At the same time, the system initiates the repair and optimization process for the main model. After the repair is completed and tested, it is reused as the main model, thereby ensuring the long-term stable operation of the system.

[0047] In summary, the overall workflow of this method for controlling devices through AI dialogue and reasoning models is as follows:

[0048] 1. User-initiated dialogue: Users make requests to the system in the form of natural language, such as saying "I'm home, the living room is a bit dark" in a smart home scenario, or asking "Can the projector in the meeting room be turned on" in a smart office scenario. These questions are received by the system and enter the subsequent processing flow.

[0049] 2. Enhanced output restriction module processing: The system selects appropriate prompt words from a diverse prompt word library based on the dialogue scenario, user habits, and device type. Combined with semantic understanding technology, it guides the output of the AI ​​big model. For example, in a smart home scenario, if the user frequently uses colloquial commands, the prompt words will guide the big model to output content containing device operation commands, and the commands will be marked with a specific format, such as "The living room light has been turned on for you. The living room light is on."

[0050] 3. Intelligent Output Monitoring Module Parses Instructions: The intelligent output monitoring module receives the output content of the AI ​​large model and uses machine learning classification technology and natural language processing technology. First, it uses a trained convolutional neural network model to determine whether the output is a control instruction. If so, it uses natural language processing tools to parse out information such as the device, operation and parameters. For example, from "turn on the living room light", it can be parsed to determine that the device to be controlled is the living room light and the operation is to turn it on.

[0051] 4. Universal Protocol Conversion Framework Adaptation: The parsed instructions enter the universal protocol conversion framework. If the target device uses a specific communication protocol, such as MQTT, the corresponding plugin in the protocol parsing layer will convert the instructions into the protocol format. The data adaptation layer processes the data according to the device requirements. The interface layer provides a unified interface to facilitate interaction with other modules and achieve accurate transmission of instructions to the device.

[0052] 5. Device Operation and Data Feedback: After receiving the converted instructions, the device executes the corresponding operation, such as turning on the living room lights. Simultaneously, the device feeds back its own status data, such as brightness and temperature.

[0053] 6. High-efficiency data input module processes feedback data: The high-efficiency data input module uses an asynchronous data acquisition interface to collect and cache device feedback data in real time, and then uses parallel computing technology to quickly parse and encapsulate it into prompt word format, such as "living room light brightness is 80%, temperature is 28 degrees Celsius", and then sends it to the AI ​​big model.

[0054] 7. Continuous Optimization and Interaction of the AI ​​Big Data Model: The AI ​​big data model learns and infers based on the received device feedback data, and can provide more accurate and realistic answers when users ask related questions again. For example, if a user asks "Is the living room light on?", the big data model, based on the feedback data, answers "The living room light is on, and the brightness is 80%".

[0055] 8. Model Backup and Switching Ensures Stable Operation: Throughout the process, the model backup and switching mechanism monitors the status of the main AI model in real time. The system periodically backs up the main model to a remote server. If the main model fails, experiences performance degradation, or is attacked, the model switching module will automatically switch to the backup model within 1 second to ensure uninterrupted equipment control. The system will switch back to the backup model after the main model has been repaired and optimized.

[0056] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0057] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for controlling a device through an AI dialogue and reasoning large model, characterized in that, Includes the following steps: S1. Utilizing the enhanced output restriction module, the AI ​​dialogue and reasoning model is guided to output content that meets the requirements through a diversified prompt word library and semantic understanding technology. The diversified prompt word library provides multiple instruction guidance methods according to different device control scenarios and user habits. The semantic understanding technology incorporates a description of the instruction semantics into the prompt words. S2. An intelligent output monitoring module is adopted, which uses machine learning classification technology and natural language processing technology to perform multi-dimensional analysis on the output of the AI ​​dialogue and reasoning model, and to identify and parse control commands. S3. Through a general protocol conversion framework, the parsed control instructions are converted into the control protocols of the corresponding devices. The general protocol conversion framework includes a protocol parsing layer, a data adaptation layer, and an interface layer. The protocol parsing layer is used to parse the communication protocols of different devices and convert them into a unified internal data format. The data adaptation layer adapts the data according to the device control requirements and data format specifications. The interface layer provides a unified calling interface. S4. With the help of the high-efficiency data input module, asynchronous processing and parallel computing technology are used to collect and process device data, and encapsulate it into prompt words to feed back to the AI ​​dialogue and reasoning model. The asynchronous processing is used to realize real-time data collection and caching, and the parallel computing technology is used to accelerate data parsing and encapsulation. S5. Construct a model backup and switching mechanism to regularly back up the large AI dialogue and inference model. When the main model fails, its performance degrades, or it is attacked, it will automatically switch to the backup model through the model switching module to ensure the continuity of equipment control.

2. The method for controlling a device through an AI dialogue and reasoning large model according to claim 1, characterized in that: In smart home scenarios, the enhanced output limiting module automatically selects appropriate prompts based on the user's past usage habits and device type. If the user frequently uses conversational commands to control devices, then a specific format of conversational prompts is selected.

3. The method for controlling a device through an AI dialogue and reasoning large model according to claim 1, characterized in that: The intelligent output monitoring module uses a trained convolutional neural network model to classify the model output text in real time. After recognizing control commands, it uses natural language processing tools to parse the device, operation, and parameter information in the commands.

4. The method for controlling a device through an AI dialogue and reasoning large model according to claim 1, characterized in that: Under the general protocol conversion framework, users can develop corresponding protocol conversion plugins according to the device's communication protocol. For smart devices using the MQTT protocol, an MQTT protocol parser plugin can be developed to convert instruction and data formats.

5. The method for controlling a device through an AI dialogue and reasoning large model according to claim 1, characterized in that: The high-efficiency data input module connects to the device using an asynchronous data acquisition interface at the data acquisition end, collects data in real time and caches it in memory, and uses multi-threading technology to process the collected data in parallel at the data processing end, quickly parses and encapsulates it into prompt word format and then transmits it to the AI ​​dialogue and reasoning model.

6. The method for controlling a device through an AI dialogue and reasoning large model according to claim 1, characterized in that: The model backup and switching mechanism backs up the AI ​​dialogue and reasoning model daily. The backup data is stored on a remote server. When the main model malfunctions, the model switching module completes the model switch within 1 second and simultaneously initiates the repair and optimization process for the main model. After the repair is completed, it is reused as the main model.

Images