A signal source intelligent control method based on a large language model

Abstract

The application discloses a signal source intelligent control method and system based on a large language model and a computer readable storage medium. The method comprises the following steps: obtaining a natural language control instruction input by a user; performing text processing; performing semantic analysis based on a large language model to obtain structured control information comprising a target device, an operation type and target parameter information; determining a target parameter value in an absolute setting operation, obtaining a current state reference value in a relative adjustment operation, identifying an adjustment direction and an adjustment amplitude level, and determining the target parameter value based on a preset step mapping rule; performing legality verification on the target parameter value; generating a control instruction and sending the control instruction to a signal source for execution. The method can also obtain an execution result and update state information, and is suitable for a multi-device collaborative control scene. The application can reduce operation complexity and improve the accuracy and convenience of natural language control.

Description

Technical Field

[0001] This invention relates to the field of electronic test and measurement instrument control technology, and in particular to a signal source intelligent control method based on a large language model. Background Technology

[0002] Signal generators are important instruments and equipment in the field of electronic test and measurement. They are typically used to generate test signals with set frequency, power, modulation method and other parameters. They are widely used in research and development, debugging, production testing and performance verification in fields such as communications, radar, electronic countermeasures, semiconductors, and aerospace.

[0003] In existing technologies, signal source control methods mainly include local panel operation and remote program control. For local panel operation, users typically need to manually set parameters such as output frequency, output power, RF output switch, and modulation mode using physical buttons, knobs, or a touch interface on the instrument's front panel. This method usually requires operators to be familiar with a relatively complex menu hierarchy and understand the meaning and corresponding operation logic of relevant technical parameters. When performing complex parameter configurations or combinations, the operation steps are numerous, resulting in a high learning curve.

[0004] For remote program control, the host computer typically sends standard programmable instrument commands to the signal source via instrument interfaces such as GPIB, LAN, and USB to achieve remote control of the instrument. While this method can improve the level of automation, it usually requires users to have a certain level of programming ability and a precise understanding of the relevant command syntax and parameter formats. In practical use, it still suffers from problems such as cumbersome debugging processes, inconvenience in temporary adjustments, and a lack of intuitive interaction.

[0005] As the complexity of electronic testing tasks continues to increase, users not only need to quickly set basic parameters such as frequency and power during actual testing, but also frequently need to dynamically adjust parameters according to the test status, or coordinate with other testing equipment to complete collaborative operations. However, most existing signal source control methods rely on users directly inputting explicit parameter values ​​or manually writing control commands, lacking the ability to understand natural language control commands, especially struggling to handle conversational, ambiguous, and relative adjustment commands, thus failing to meet the needs for natural and convenient equipment control.

[0006] Therefore, there is an urgent need in the existing technology for a signal source control technology solution that can semantically parse the natural language control commands input by the user and convert the parsing results into executable control parameters, especially for relatively adjustment control commands that do not directly give the absolute value of the target parameter. Summary of the Invention

[0007] The purpose of this invention is to provide a signal source intelligent control method based on a large language model, so as to solve the problems of complex operation, high requirements for users' professional skills, difficulty in effectively semantically parsing natural language control commands, difficulty in handling relative adjustment control needs, and inconvenience in multi-device collaborative control in existing signal source control methods.

[0008] To achieve the above objectives, this invention provides a signal source intelligent control method based on a large language model, comprising the following steps: acquiring a natural language control command input by a user; processing the natural language control command into text; performing semantic parsing on the text-processed natural language control command based on a large language model to obtain structured control information, wherein the structured control information includes at least target device, operation type, and target parameter information; when the operation type is an absolute setting operation, determining the target parameter value based on the target parameter information; when the operation type is a relative adjustment operation, acquiring the current state reference value corresponding to the target parameter, identifying the adjustment direction and adjustment magnitude level, and determining the adjustment amount based on a preset step mapping rule, so as to determine the target parameter value based on the current state reference value, the adjustment direction, and the adjustment amount; performing parameter validity verification on the target parameter value; generating a corresponding control command based on the verified target parameter value; and sending the control command to the signal source to control the signal source to perform the corresponding operation.

[0009] In one embodiment, the natural language control instruction includes voice instructions and text instructions; when the natural language control instruction is a voice instruction, the voice instruction is subjected to speech recognition to obtain the corresponding text information; when the natural language control instruction is a text instruction, the text instruction is directly used as the text to be parsed and input into the subsequent processing flow.

[0010] In one embodiment, the semantic parsing of the text-processed natural language control instructions based on a large language model includes: identifying the target device controlled by the user; identifying the operation type corresponding to the user's intent; extracting the parameter name, parameter value, and unit of the parameter to be controlled; and when the operation type is a relative adjustment operation, identifying the adjustment direction and degree expression in the natural language control instructions, and mapping the degree expression to a preset adjustment range level.

[0011] In one embodiment, determining the adjustment amount based on a preset step mapping rule and determining the target parameter value according to the current state reference value, the adjustment direction, and the adjustment amount includes: querying a preset mapping relationship between parameter type, adjustment magnitude level, and step value to obtain the standard step value corresponding to the target parameter; adding the standard step value to the current state reference value when the adjustment direction is increasing; and subtracting the standard step value from the current state reference value when the adjustment direction is decreasing.

[0012] In one embodiment, the parameter validity verification of the target parameter value includes: performing boundary verification and validity judgment on the target parameter value according to the parameter range, unit requirements, format rules or security restrictions of the corresponding parameter of the target device; when the target parameter value exceeds the allowable range, the unit does not match or there is a format abnormality, generating an abnormal prompt message and refusing to execute the corresponding control command, or correcting the target parameter value to the boundary value within the allowable range before executing the subsequent control process.

[0013] In one embodiment, generating the corresponding control instruction based on the verified target parameter value includes generating a Programmable Instrument Standard Command (SCPI) instruction; sending the control instruction to the signal source includes sending the SCPI instruction to the signal source via the VISA Universal Instrument Interface.

[0014] In one embodiment, the method further includes: after sending the control command, obtaining the execution result returned by the signal source; updating the status information corresponding to the target device according to the execution result; and outputting the execution result and / or abnormal information.

[0015] In one embodiment, when the natural language control instruction includes multiple device control requirements, the natural language control instruction is parsed to obtain structured control information corresponding to each of the multiple devices; corresponding control instructions are generated based on the structured control information corresponding to each device; and the corresponding control instructions are sent to the signal source and other test devices for execution.

[0016] This invention also provides an intelligent control system for a signal source based on a large language model, comprising an input processing unit, a semantic parsing unit, a parameter determination unit, a parameter verification unit, an instruction generation unit, and an instrument driving unit. The input processing unit is used to acquire natural language control instructions input by the user and perform textualization processing on the natural language control instructions. The semantic parsing unit is used to perform semantic parsing on the textualized natural language control instructions based on the large language model. The parameter determination unit is used to determine target parameter values. The parameter verification unit is used to perform parameter validity verification on the target parameter values. The instruction generation unit is used to generate corresponding control instructions. The instrument driving unit is used to send the control instructions to the signal source to control the signal source to perform corresponding operations.

[0017] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is run on a processor, the processor executes the above-described intelligent control method for signal sources based on a large language model.

[0018] Compared with existing technologies, the present invention has at least the following beneficial effects: The present invention performs semantic parsing of natural language control instructions based on a large language model and converts the parsing results into control instructions executable by the signal source, thereby reducing the dependence of signal source control on professional instruction syntax and manual operation procedures; In the relative adjustment operation scenario, the present invention determines the target parameter value by combining the current state reference value, adjustment direction, adjustment amplitude level and preset step mapping rules, thereby realizing the parameterization of relative adjustment control requirements; In addition, the present invention can also perform legality verification of the target parameter value, support execution result feedback and state update, and can be extended to multi-device collaborative control scenarios. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall architecture of a signal source intelligent control system based on a large language model according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the processing flow of a signal source intelligent control method based on a large language model according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of a user interface according to an embodiment of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention. For those skilled in the art, various modifications, substitutions and variations can be made to the technical solution of the present invention without departing from the spirit and substance of the present invention, and all such modifications, substitutions and variations should fall within the protection scope of the present invention.

[0023] In the description of this invention, it should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, system, product, 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, system, product, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, system, product, or apparatus that includes said element.

[0024] In the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance, nor should they be construed as implying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0025] In this invention, Standard Commands for Programmable Instruments (SCPI) is a standard command set for controlling programmable test and measurement instruments; Virtual Instrument Software Architecture (VISA) is a software interface architecture for communication between computers and instruments; and Graphics User Interface (GUI) is an interface for user interaction with the system. For frequency units used in this invention, Hertz, kilohertz, megahertz, and gigahertz can be denoted as Hz, kHz, MHz, and GHz, respectively, and decibels and milliwatts can be denoted as dBm.

[0026] Furthermore, in this invention, General Purpose Interface Bus (GPIB) is a bus interface used for instrument communication; Local Area Network (LAN) is a local area network used for device network communication; Universal Serial Bus (USB) is a serial bus interface used for device connection and data transmission; Radio Frequency (RF) refers to radio frequency signals or radio frequency output; Automatic Speech Recognition (ASR) is a speech recognition technology used to implement speech-to-text conversion; and JavaScript Object Notation (JSON) is a data format used to represent structured control information.

[0027] Furthermore, in this invention, PyVISA is a Python software library used to implement communication control between programs and instruments; TCP / IP is the Transmission Control Protocol / Internet Protocol; and Prompt is a prompt constraint message used to constrain and guide the output of structured control information by a large language model. For other English abbreviations or terms used herein, their common meanings in the art shall apply.

[0028] The system configuration and control method of the present invention will be described in detail below with reference to the accompanying drawings.

[0029] like Figure 1 As shown, one embodiment of the present invention provides a signal source intelligent control system 100 based on a large language model. The signal source intelligent control system 100 may include a user input layer 210, a voice interaction layer 110, a semantic parsing layer 120, a state management layer 130, an instrument driving layer 140, and a feedback display layer 150. The instrument driving layer 140 is communicatively connected to the signal source 200 to send control commands to the signal source 200 and receive execution feedback 220.

[0030] In one embodiment, the user input layer 210 is used to receive natural language control instructions input by the user. The natural language control instructions can be voice instructions or text instructions. When the natural language control instruction is a voice instruction, the user input layer 210 can send the voice instruction to the voice interaction layer 110; when the natural language control instruction is a text instruction, the text instruction can be directly sent to the semantic parsing layer 120 for subsequent processing.

[0031] In one embodiment, the voice interaction layer 110 is used to acquire, preprocess, and perform speech recognition on user voice signals to convert natural language control commands in voice form into text information. The voice interaction layer 110 can employ a speech recognition engine to achieve speech-to-text conversion and supports real-time recording input and offline voice input.

[0032] In one embodiment, the semantic parsing layer 120 is used to perform semantic parsing on the text information based on a large language model to output structured control information. Specifically, the semantic parsing layer 120 can receive text instructions and preset prompts, and use the large language model to perform device identification, operation type identification, parameter extraction, adjustment direction identification, degree identification, and intent disambiguation processing on the text instructions, thereby generating structured control information that can be used for subsequent control, such as structured control information in JSON format.

[0033] In one embodiment, the device identified by the semantic parsing layer 120 may include a signal source, a spectrum analyzer, or multiple devices in a multi-device collaborative control scenario; the identified operation type may include absolute setting operations and relative adjustment operations; the extracted parameters may include output frequency, output power, RF output switch status, and scan-related parameters; the degree recognition is used to map degree expressions such as "a little," "some," "obvious," or "significant" to preset adjustment levels. When the operation type is an absolute setting operation, the target parameter value can be directly determined based on the parameter values ​​in the structured control information.

[0034] In one embodiment, the state management layer 130 is used to maintain the current state information of the controlled device. The current state information may include the output frequency, output power, RF switch status, and scan-related parameters of the signal source, and may also include the state information of other test equipment associated with the signal source. The state management layer 130 can provide a current state reference value for relative adjustment control.

[0035] In one embodiment, the state management layer 130 can also be used to respond to relative adjustment commands. When the semantic parsing layer 120 recognizes that the natural language control command input by the user is a relative adjustment command, the state management layer 130 can determine the adjustment amount of the target parameter based on the current state reference value and a preset step mapping rule, and calculate the target parameter value accordingly. The preset step mapping rule can be used to determine the standard step value of the corresponding parameter based on the identified adjustment direction and adjustment magnitude level.

[0036] In one embodiment, the instrument driver layer 140 generates instrument-recognizable control commands based on the structured control information and the determined target parameter values, and sends the control commands to the signal source 200 for execution. Before generating the control commands, the system can also perform parameter validity verification on the target parameter values. Preferably, the control commands are Programmable Instrument Standard Command (SCPI) commands, and the instrument driver layer 140 can establish a communication connection with the signal source 200 through the VISA Universal Instrument Interface to complete command sending and execution feedback reception.

[0037] In one embodiment, the feedback display layer 150 is used to output execution feedback information to the user. The feedback display layer 150 can construct a graphical interactive interface for displaying the current status of the device, natural language parsing results, structured control information, parameter adjustment information, instruction execution results, and exception information.

[0038] In one embodiment, the intelligent control system 100 for the signal source can support collaborative control of multiple devices. The semantic parsing layer 120 can parse a single natural language control command into structured control information corresponding to multiple devices. Based on this, the instrument driving layer 140 generates control commands for the corresponding devices and sends them to the signal source and other test devices for execution.

[0039] In one embodiment, each layer of the signal source intelligent control system 100 can be implemented by software, hardware, or a combination of both. For example, the voice interaction layer 110, semantic parsing layer 120, state management layer 130, instrument driving layer 140, and feedback display layer 150 can be implemented by one or more processors executing program instructions in memory; or, some functions in each layer can also be implemented by independent circuits, interface units, or dedicated processing units.

[0040] like Figure 2 As shown, one embodiment of the present invention provides a signal source intelligent control method based on a large language model, the method including the following steps:

[0041] S110, acquire natural language control commands input by the user.

[0042] In one embodiment, the natural language control command can be a voice command or a text command input by the user. The natural language control command can be used to express parameter settings, on / off control, parameter adjustments, or multi-device collaborative control of the signal source. For example, the user can input a natural language control command such as "Set the signal source output frequency to 5.8GHz, power 0dBm, turn on RF output".

[0043] S120, the natural language control instructions are processed into text.

[0044] In one embodiment, when the natural language control command is a voice command, it can be converted into corresponding text information by a speech recognition engine; when the natural language control command is a text command, it can be directly input as the text to be parsed into subsequent processing. Furthermore, the text processing may also include text normalization processing, which can be used to remove redundant words, unify parameter expression formats, identify unit information, and standardize equipment name representations.

[0045] S130, perform semantic parsing on the text information to obtain structured control information.

[0046] In one embodiment, the text information and preset prompts can be input into a large language model, which then outputs corresponding structured control information. The structured control information may include device identification results, operation identification results, and parameter extraction results, and can be output in a structured data format such as JSON. Furthermore, the semantic parsing may also include keyword recognition, intent recognition, parameter extraction, and structured output format verification.

[0047] In one embodiment, the device identification result is used to determine the target device controlled by the user. The target device may be a signal source or a test device such as a spectrum analyzer that is associated with the signal source. The operation identification result is used to determine the operation type corresponding to the user's intention. The operation type may include absolute setting operation and relative adjustment operation. The parameter extraction result is used to determine the parameter name, parameter value and unit of the parameter to be controlled.

[0048] In one embodiment, the semantic parsing can also identify degree expressions in natural language control instructions and map the degree expressions to preset adjustment range levels for subsequent adjustment calculations. The degree expressions may include at least one of "a little," "some," "obvious," or "significant," and the preset adjustment range levels may include at least one of small, medium, and significant.

[0049] S140, determine whether the structured control information corresponds to relative adjustment control.

[0050] In one embodiment, if the structured control information corresponds to absolute setting control, then the target parameter value is determined based on the structured control information, and the process proceeds to step S190; if the structured control information corresponds to relative adjustment control, then the process proceeds to step S150. Relative adjustment control refers to a control method where the user does not directly provide the absolute value of the target parameter, but instead uses relative expressions such as "increase a little," "decrease a bit," or "slightly increase" for control.

[0051] S150, obtain the current state baseline value corresponding to the target parameter.

[0052] In one embodiment, the current state information corresponding to the target parameter can be read from the state management layer as a reference value for subsequent relative adjustment calculations. For example, when a user requests a relative adjustment to the output frequency, the current output frequency can be read as the current state reference value; when a user requests a relative adjustment to the output power, the current output power can be read as the current state reference value.

[0053] S160 identifies the adjustment direction and adjustment level.

[0054] In one embodiment, based on the structured control information output by the large language model, the user's desired adjustment direction (increase or decrease) can be identified, along with the corresponding adjustment level. The adjustment level may include preset levels such as small, medium, or large. For example, for expressions like "increase a little" or "decrease a little," the corresponding adjustment direction and small adjustment level can be identified respectively.

[0055] S170, determine the adjustment amount according to the preset step mapping rule.

[0056] In one embodiment, corresponding step mapping rules can be pre-set for different parameters, and different adjustment levels can be mapped to corresponding standard step values. The standard step values ​​can be preset according to the target parameter type. For example, for frequency parameters, small amplitude levels can be mapped to step values ​​of 0.1 GHz; for power parameters, they can be mapped to corresponding dBm adjustment amounts according to preset rules; and for scan-related parameters, they can also be mapped to corresponding scan step values ​​according to preset rules.

[0057] In one embodiment, the preset step mapping rule can be used to establish a correspondence between degree description, adjustment direction, adjustment magnitude level and parameter step value, and query the corresponding standard step value based on the target parameter type and the adjustment magnitude level, thereby converting the relative adjustment intention in natural language into an executable adjustment amount.

[0058] S180, determine the target parameter value based on the current state reference value and the adjustment amount.

[0059] In one embodiment, the current state reference value can be increased or decreased according to the adjustment direction to obtain the target parameter value. For example, when the current output frequency is 3.8 GHz and the step value corresponding to the small amplitude level is 0.1 GHz, when the "increase the frequency a little" command is detected, the target parameter value can be determined to be 3.9 GHz.

[0060] In one embodiment, when the adjustment direction is to increase, the standard step value can be added to the current state reference value; when the adjustment direction is to decrease, the standard step value can be subtracted from the current state reference value to obtain the target parameter value.

[0061] S190, Perform parameter validity verification on the target parameter value.

[0062] In one embodiment, the target parameter value can be boundary-checked and its legality judged based on the parameter range, unit requirements, format rules, or security restrictions of the target device. When the target parameter value exceeds the allowed range, the unit does not match, or there is an abnormal format, an error message can be generated, and the corresponding control command can be refused to be executed. In other embodiments, the target parameter value can be corrected to a boundary value within the allowed range, or a prompt message can be output to guide the user to confirm before execution. When the target parameter value is legal, the subsequent command generation step is initiated.

[0063] S200, generate SCPI instructions.

[0064] In one embodiment, based on the verified target parameter value and the corresponding target device, operation type, and parameter name, an SCPI instruction for controlling the signal source to perform the corresponding operation is generated. For natural language control instructions containing multiple control actions, one or more SCPI instruction sequences can be generated to achieve the combined execution of multiple actions such as frequency setting, power setting, RF output switch control, and scan parameter setting.

[0065] S210, send the SCPI command to the signal source for execution.

[0066] In one embodiment, the SCPI command can be sent to a signal source via the VISA universal instrument interface, whereby the signal source performs the corresponding parameter setting or control operation. For multi-device collaborative control scenarios, corresponding control commands can also be sent to the signal source and other test devices separately.

[0067] S220: Obtain the execution result and update the status information.

[0068] In one embodiment, after sending the SCPI command, the execution result returned by the signal source can be received, and the device state cache in the state management layer can be updated according to the execution result to support the continuous processing of subsequent relative adjustment control commands. Simultaneously, execution success prompts, execution failure prompts, exception information, parameter adjustment information, or parsing logs can be displayed in the user interface.

[0069] In one embodiment, the method can be used for both absolute parameter setting control and relative adjustment control based on fuzzy natural language. In the relative adjustment control, the user does not need to directly input the absolute value of the target parameter. Furthermore, the method can also be used in multi-device collaborative control scenarios, that is, parsing a single natural language control command into structured control information corresponding to multiple devices, and generating control commands for each corresponding device.

[0070] like Figure 3 As shown, one embodiment of the present invention provides a user interface 300 for implementing intelligent control of signal sources based on a large language model. The user interface 300 can be displayed on the display device of a host computer terminal and is used to receive natural language control commands input by the user, display device status information, and provide feedback on control execution results.

[0071] In one embodiment, the user interface 300 may include a device connection control area 310, a shortcut command area 320, a real-time device status area 330, a parsing log display area 340, a command result feedback area 350, and an input area 360.

[0072] In one embodiment, the device connection control area 310 is used to display the current connection status of the controlled device and provide operation entry points such as device connection, disconnection, or device switching. The device connection control area 310 can also display device identification information, communication interface information, and the type of the currently connected target device.

[0073] In one embodiment, the shortcut command area 320 is used to display preset commonly used control command templates. These templates can correspond to operations such as frequency setting, power setting, RF output switch control, scan-related parameter setting, or multi-device collaborative control. Users can quickly generate corresponding input content or trigger corresponding control flows by selecting the commonly used control command templates, thereby reducing repetitive input and improving the execution efficiency of commonly used control operations.

[0074] In one embodiment, the real-time device status area 330 is used to display the current operating status of the controlled device. The current operating status may include output frequency, output power, RF output switch status, scan-related parameters, and other test status information. In a multi-device collaborative control scenario, the real-time device status area 330 can also simultaneously display the status information corresponding to multiple devices. The current status information displayed in the real-time device status area 330 can also provide a reference for understanding the execution results of relative adjustment-type control and for subsequent continuous control.

[0075] In one embodiment, the parsing log display area 340 is used to display the parsing results or intermediate processing results of natural language control commands. The parsing log display area 340 can display device recognition results, operation recognition results, parameter extraction results, adjustment direction, adjustment range level, and structured control information output by the large language model.

[0076] In one embodiment, the instruction result feedback area 350 is used to display the execution result of the control instruction. The execution result may include information such as instruction sending success notification, execution completion notification, execution failure notification, parameter out-of-bounds notification, format error notification, or communication exception notification. Furthermore, the instruction result feedback area 350 may also display the target parameter value, parameter adjustment amount, parameter changes before and after adjustment, and corresponding execution time information.

[0077] In one embodiment, the input area 360 is used to receive natural language control commands input by the user. The input area 360 may include a text input sub-area and a voice input sub-area. The user can input natural language control commands via text input or voice input, which are then automatically converted into text information by the system before proceeding to the subsequent semantic parsing process. The input area 360 can also support the continuous input of multiple natural language control commands.

[0078] In one embodiment, when a user inputs a natural language control command through the input area 360, the user interface 300 can send the natural language control command to the background control processing module for processing. After the control command is executed, the corresponding execution result, status update information, and exception prompt information are displayed in the real-time device status area 330, the parsing log display area 340, and the command result feedback area 350. The background control processing module may include at least some of the functional modules of the aforementioned semantic parsing layer, status management layer, and instrument driver layer.

[0079] In one embodiment, the user interface 300 can also support multi-round interactions. When the natural language control command input by the user is ambiguous, parameters are missing, or device indication is unclear, the user interface 300 can output prompt information to guide the user to supplement the corresponding parameters or reconfirm the control intention, and continue to execute the subsequent control process after confirmation.

[0080] In one embodiment, the position, size, and content of each display area in the user interface 300 can be adjusted according to actual application needs. The above interface layout is only an illustrative example and does not constitute a limitation of the present invention.

[0081] In one embodiment, the user needs to perform an absolute setting operation on the signal source. Specifically, the user can input natural language control commands via voice input, such as "set the signal source output frequency to 5.8GHz, power to 0dBm, and turn on RF output." The system first converts the voice command into corresponding text information through the voice interaction layer, and then the semantic parsing layer performs semantic parsing on the text information based on a large language model to obtain the corresponding structured control information. The structured control information may include the target device being the signal source, the output frequency parameter being 5.8GHz, the output power parameter being 0dBm, and the RF output status being on. When the natural language control command corresponds to an absolute setting operation, the system can determine the target parameter value based on the parameter values ​​in the structured control information. Subsequently, the instrument driver layer generates one or more corresponding SCPI commands, or generates a corresponding SCPI command sequence, based on the structured control information and the target parameter value, and sends them to the signal source for execution via the VISA universal instrument interface. After the signal source completes the execution, the feedback display layer can display a successful setting message on the user interface and simultaneously update the device status information in the status management layer.

[0082] In one embodiment, the user needs to perform a relative adjustment operation on the signal source parameters. Specifically, if the current output frequency of the signal source is recorded as 3.8 GHz and the output power as -10 dBm in the state management layer, the user can input the natural language control command "increase the frequency slightly". After the semantic parsing layer performs semantic parsing on the natural language control command using a large language model, it can identify the command as a relative adjustment command and identify the target parameter as the output frequency, the adjustment direction as increase, the degree as "slightly", and the adjustment magnitude as small. Subsequently, the state management layer determines the standard step value corresponding to the output frequency at the small magnitude level as 0.1 GHz according to the preset step mapping rules, and calculates the target parameter value of 3.9 GHz based on the current state reference value of 3.8 GHz. After that, the system performs parameter validity verification on the target parameter value. After the verification is passed, the instrument driver layer generates the corresponding SCPI command and sends it to the signal source for execution. After execution, the feedback display layer can display information such as the original parameter value, adjustment amount, target value, and execution result. In this embodiment, the user does not need to directly input the absolute value of the target parameter; the system can complete the calculation of the target parameter value and control execution based on the relative adjustment intention in natural language.

[0083] In one embodiment, a user can control the signal source using fuzzy natural language. Specifically, the user can input natural language control commands such as "reduce the signal power slightly to avoid overloading the device." The semantic parsing layer performs semantic understanding of the natural language control commands based on a large language model, identifying that "reduce slightly" corresponds to a small reduction in the power parameter, and extracts the target parameter, adjustment direction, and adjustment magnitude related to the control. For additional semantic information related to the control objective but not directly constituting the parameter value, such as "avoid overloading the device," the system can treat it as background semantics and extract the core control intent. Subsequently, the state management layer reads the current output power as the state reference value and determines the corresponding small power adjustment amount according to a preset step mapping rule, thereby calculating the target power parameter value. After completing parameter validity verification, the instrument driver layer generates the corresponding control command and sends it to the signal source for execution. After execution, the feedback display layer can output the corresponding execution result and state update information. In this embodiment, even if the user does not explicitly provide a specific parameter value, the system can still complete the corresponding control operation based on the natural language control command.

[0084] In one embodiment, the method can also be applied to multi-device collaborative control scenarios. Specifically, during testing, the user can input a natural language control command: "Set both the signal source and spectrum analyzer to 3.8GHz, and set the signal source power to -5dBm." After parsing the natural language control command, the semantic parsing layer can identify that the command contains multi-device control requirements for the signal source and spectrum analyzer, and decompose the single natural language control command into multiple control sub-tasks corresponding to each device, thereby extracting the corresponding target parameter information for each. Subsequently, the system can determine the corresponding target parameter values ​​for different devices and generate corresponding structured control information and control commands. Since the control objects and command formats of different devices may be different, the instrument driver layer can generate corresponding control commands suitable for the signal source and spectrum analyzer respectively, and send the control commands to the signal source and spectrum analyzer for execution. After execution, the feedback display layer outputs the execution results of each device, and the state management layer updates the corresponding state information of each device, thereby forming a closed-loop processing process for multi-device collaborative control.

[0085] In other embodiments, the method and system may also have the following optional implementations:

[0086] In one embodiment, the preset step mapping rule can be configured using the correspondence between parameter type, adjustment amplitude level, and step value. For example, for the output frequency parameter, small, medium, and large adjustment levels can correspond to 0.1GHz, 0.5GHz, and 1.0GHz, respectively; for the output power parameter, small, medium, and large adjustment levels can correspond to 1dBm, 3dBm, and 10dBm, respectively; for the scan width parameter among scan-related parameters, small, medium, and large adjustment levels can correspond to 10MHz, 50MHz, and 100MHz, respectively; and for the reference level parameter, small, medium, and large adjustment levels can correspond to 5dBm, 10dBm, and 20dBm, respectively. It should be understood that the above step values ​​are merely examples, and those skilled in the art can adjust and configure the step values ​​according to different device types, test scenarios, or control requirements without departing from the technical concept of the present invention.

[0087] In one embodiment, when the target parameter value calculated based on the current state reference value and the adjustment amount exceeds the parameter range allowed by the target device, the system can output an error message and refuse to execute the corresponding control command. In other embodiments, the system can also correct the target parameter value to a boundary value within the allowed range of the target device before executing subsequent control procedures. For example, when the target frequency calculated according to the relative adjustment command is higher than the maximum allowed output frequency of the signal source, the target frequency can be corrected to the maximum output frequency; when the calculated target power is lower than the minimum allowed output power of the device, the target power can be corrected to the minimum output power.

[0088] In one embodiment, the SCPI command can adopt different command formats depending on the target parameter type. For example, output frequency setting can correspond to a frequency setting command, output power setting can correspond to a power setting command, RF output switch control can correspond to an output switch command, and scan-related parameter setting can correspond to a scan parameter setting command. For natural language control commands containing multiple control actions, the system can generate one or more SCPI commands, or generate a corresponding SCPI command sequence, to achieve the combined execution of multiple control actions.

[0089] In one embodiment, the voice interaction process may employ a 16kHz sampling rate and mono mode to capture user voice commands, and then invoke a speech recognition engine to convert speech to text. The speech recognition engine may include Baidu ASR, iFlytek ASR, or other speech recognition engines, and may support real-time recording input and offline voice input. It should be understood that the type of speech recognition engine and the voice acquisition parameters can be selected and configured according to actual application needs, and do not constitute a limitation of the present invention.

[0090] In one embodiment, the semantic parsing layer can perform domain-specific semantic parsing of natural language control commands based on preset prompts and contextual constraints, without requiring specific fine-tuning of the large language model for signal source control tasks, and output corresponding structured control information. The preset prompts can be used to define the target device type, parameter name, parameter unit, output format, and control semantic rules, thereby improving the accuracy and applicability of parsing specialized equipment control commands.

[0091] In one embodiment, the preset prompt information may include prompt constraint information (Prompt) constructed for signal source control scenarios. For example, the prompt constraint information may include device type constraints, parameter name constraints, parameter unit constraints, structured output templates, and semantic disambiguation rules, used to guide the large language model to identify test devices such as signal sources, spectrum analyzers, and power meters, and to identify target parameters such as output frequency, output power, RF output switch status, scan-related parameters, and reference levels, and output corresponding parsing results according to a preset structured data format. Furthermore, the prompt constraint information can also be used to disambiguate ambiguous semantics and standardize colloquial expressions.

[0092] In one embodiment, a standardized mapping relationship can be used between the degree description and the preset adjustment level. For example, "a little" or "slightly" can be mapped to a small adjustment level corresponding to "small," "some" or "obviously" can be mapped to a medium adjustment level corresponding to "medium," and "significantly" can be mapped to a large adjustment level corresponding to "large." It should be understood that the mapping relationship between the degree description and the adjustment level can be adjusted and configured according to different application scenarios, and does not constitute a limitation of the present invention.

[0093] In one embodiment, the instrument driver layer can establish a communication connection with the target device through the PyVISA instrument control library and send the generated SCPI commands to the signal source and other test equipment for execution via the TCP / IP communication protocol. It should be understood that, in addition to PyVISA and TCP / IP, other instrument control libraries, communication protocols, or interface methods can also be used to control the target device without departing from the technical concept of this invention.

[0094] In one embodiment, in a multi-device collaborative control scenario, the other test equipment, in addition to a spectrum analyzer, may also include a power meter or other test and measurement equipment used in conjunction with the signal source. The system can determine the target parameter values ​​corresponding to each device based on a single natural language control command, and generate control commands adapted to the command format of each device, thereby achieving synchronous setting or coordinated control of multi-device parameters.

[0095] In one embodiment, after the control command is successfully executed, the system can synchronously update the device state cache in the state management layer and display information such as the original value of the parameter, the adjustment amount, the new value, and the execution status in the user interface, thereby forming a closed-loop interactive process of input, parsing, execution, and feedback, and providing an accurate current state reference value for subsequent relative adjustment control.

[0096] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Various modifications, equivalent substitutions, or improvements made by those skilled in the art to the above embodiments without departing from the spirit and substance of the present invention should fall within the scope of protection of the present invention.

Claims

1. A method for intelligent control of signal sources based on a large language model, characterized in that, Includes the following steps: Obtain natural language control commands input by the user; The natural language control instructions are digitized; Based on a large language model, the natural language control instructions after text processing are semantically parsed to obtain structured control information, which includes at least target device, operation type and target parameter information. When the operation type is an absolute setting operation, the target parameter value is determined based on the target parameter information; When the operation type is a relative adjustment operation, the current state reference value corresponding to the target parameter is obtained, the adjustment direction and adjustment magnitude level are identified, and the adjustment amount is determined based on a preset step mapping rule, so as to determine the target parameter value according to the current state reference value, the adjustment direction and the adjustment amount; Perform parameter validity verification on the target parameter value; Generate corresponding control commands based on the verified target parameter values; The control command is sent to the signal source to control the signal source to perform the corresponding operation.

2. The intelligent control method for signal sources based on a large language model according to claim 1, characterized in that, The natural language control commands include voice commands and text commands; When the natural language control command is a voice command, the voice command is subjected to speech recognition to obtain the corresponding text information; When the natural language control command is a text command, the text command is directly used as the text to be parsed and input into the subsequent processing flow.

3. The intelligent control method for signal sources based on a large language model according to claim 1, characterized in that, The semantic parsing of the natural language control instructions after text processing based on a large language model includes: Identify the target device controlled by the user; Identify the operation type corresponding to the user's intent; Extract the parameter name, parameter value, and unit of the parameter to be controlled; When the operation type is a relative adjustment operation, the adjustment direction and degree description in the natural language control command are identified, and the degree description is mapped to a preset adjustment range level.

4. The intelligent control method for signal sources based on a large language model according to claim 3, characterized in that, The step of determining the adjustment amount based on a preset step mapping rule, and determining the target parameter value based on the current state reference value, the adjustment direction, and the adjustment amount, includes: Query the preset mapping relationship between parameter type, adjustment level and step value to obtain the standard step value corresponding to the target parameter; When the adjustment direction is to increase, the standard step value is added to the current state reference value; When the adjustment direction is to decrease, the standard step value is subtracted from the current state reference value.

5. The intelligent control method for signal sources based on a large language model according to claim 1, characterized in that, The parameter validity verification of the target parameter value includes: Based on the parameter range, unit requirements, format rules, or security restrictions of the target device, the target parameter values ​​are subjected to boundary verification and legality judgment. When the target parameter value exceeds the allowable range, the units do not match, or there is an abnormal format, an error message is generated and the corresponding control instruction is refused to be executed, or the target parameter value is corrected to the boundary value within the allowable range before the subsequent control process is executed.

6. The intelligent control method for signal sources based on a large language model according to claim 1, characterized in that, The step of generating corresponding control instructions based on the verified target parameter values ​​includes generating programmable instrument standard command (SCPI) instructions. Sending the control command to the signal source includes sending the SCPI command to the signal source via the VISA universal instrument interface.

7. The intelligent control method for signal sources based on a large language model according to claim 1, characterized in that, The method further includes: After sending the control command, obtain the execution result returned by the signal source; Update the status information corresponding to the target device based on the execution result; Output the execution results and / or exception information.

8. The intelligent control method for signal sources based on a large language model according to claim 1, characterized in that, When the natural language control command includes multiple device control requirements, the natural language control command is parsed to obtain structured control information corresponding to each of the multiple devices; Generate corresponding control commands based on the structured control information corresponding to each device; The corresponding control commands are sent to the signal source and other test equipment for execution.

9. A signal source intelligent control system based on a large language model, characterized in that, include: An input processing unit is used to acquire natural language control instructions input by the user and to perform textual processing on the natural language control instructions; The semantic parsing unit is used to perform semantic parsing on the text-processed natural language control instructions based on a large language model to obtain structured control information, which includes at least target device, operation type and target parameter information. The parameter determination unit is used to determine the target parameter value according to the target parameter information when the operation type is an absolute setting operation, and to obtain the current state reference value corresponding to the target parameter, identify the adjustment direction and adjustment magnitude level, and determine the adjustment amount based on a preset step mapping rule when the operation type is a relative adjustment operation, so as to determine the target parameter value according to the current state reference value, the adjustment direction and the adjustment amount. A parameter verification unit is used to verify the validity of the target parameter value; The instruction generation unit is used to generate corresponding control instructions based on the verified target parameter values. The instrument driving unit is used to send the control command to the signal source to control the signal source to perform the corresponding operation.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run on the processor, it causes the processor to execute the intelligent control method for signal sources based on a large language model as described in any one of claims 1 to 8.

Images