Data acquisition method for intelligent training ground with body and electronic equipment

Abstract

The invention provides a data acquisition method and electronic equipment for an intelligent training ground with a body. The method comprises the following steps: in response to a received acquisition task, determining an acquisition data parameter and an output data parameter corresponding to the acquisition task; determining a plurality of source devices and configuration information respectively corresponding to the plurality of source devices according to the acquired data parameters; generating a target driving program according to the acquired data parameters and the configuration information of the plurality of source devices; running the target drive program to obtain collection data corresponding to the collection task; and according to the output data parameters, format conversion processing and data annotation processing are performed on the acquired data to generate the output data corresponding to the acquisition task, so that the data acquisition efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of data acquisition, and more particularly to a data acquisition method and electronic device for embodied intelligence training grounds. Background Technology

[0002] As various industries upgrade towards digitalization and intelligence, multimodal data from multiple sources has become the foundational data for precise robot control, algorithm model training, and scenario-based service optimization.

[0003] Related data acquisition technologies are usually designed for single-scenario data acquisition. For example, for devices from specific manufacturers, a matching dedicated driver is manually selected, acquisition parameters are manually configured, raw data is acquired using independent tools, and then data annotation is done manually.

[0004] While the above data acquisition process can meet the single data acquisition needs in simple scenarios, it cannot meet the data acquisition needs of multi-source heterogeneous devices, resulting in low data acquisition efficiency. Summary of the Invention

[0005] This application provides a data acquisition method and electronic device for a embodied intelligent training field, in order to solve the problem of low data acquisition efficiency in related technologies.

[0006] In a first aspect, embodiments of this application provide a data acquisition method for an embodied intelligent training field, the method comprising:

[0007] In response to the received acquisition task, determine the acquisition data parameters and output data parameters corresponding to the acquisition task;

[0008] Based on the collected data parameters, determine the configuration information of multiple source devices and their respective counterparts.

[0009] Based on the collected data parameters and configuration information of multiple source devices, generate the target driver;

[0010] Run the target driver to obtain the acquisition data corresponding to the acquisition task;

[0011] Based on the output data parameters, the collected data undergoes format conversion and data annotation to generate the output data corresponding to the collection task.

[0012] In one possible implementation, the data acquisition task includes acquiring quality requirement information and acquisition time limit requirement information, and the data acquisition parameters include acquisition priority; in response to the received acquisition task, the data acquisition parameters and output data parameters corresponding to the acquisition task are determined, including:

[0013] Based on the information on the required collection quality and the required collection time, determine the collection priority for each collection task.

[0014] In one possible implementation, the data acquisition quality requirements information includes the scarcity of the data to be acquired, current acquisition environment information, and acquisition cost; based on the data acquisition quality requirements information and acquisition time limit requirements information, the acquisition priority corresponding to the acquisition task is determined, including:

[0015] The expected information to be collected is determined based on the scarcity of the data to be collected, the current collection environment, and the collection cost.

[0016] Based on the expected data collection information and the time limit requirements for data collection, determine the collection priority corresponding to the collection task.

[0017] In one possible implementation, a target driver is generated based on the acquired data parameters and configuration information of multiple source devices, including:

[0018] Based on the collected data parameters and the configuration information of multiple source devices, generate task templates corresponding to each of the multiple source devices;

[0019] Based on the task templates corresponding to multiple source devices, the preset driver is configured to obtain the target driver.

[0020] In one possible implementation, based on the collected data parameters and the configuration information of multiple source devices, task templates corresponding to each of the multiple source devices are generated, including:

[0021] Based on the collected data parameters, at least one acquisition element is determined. The at least one acquisition element includes at least one of the following: data modality type, acquisition frequency, data precision, and data storage format.

[0022] Based on the configuration information of at least one acquisition element and multiple source devices, generate acquisition trigger conditions, data caching modes and exception handling modes corresponding to multiple source devices respectively;

[0023] Based on the acquisition trigger conditions, data caching mode, and exception handling mode corresponding to each of the multiple source devices, generate task templates corresponding to each of the multiple source devices.

[0024] In one possible implementation, a preset driver is configured according to the task templates corresponding to multiple source devices to obtain a target driver, including:

[0025] The task templates corresponding to multiple source devices are parsed and processed to extract driver configuration parameters. The driver configuration parameters include at least one of the communication protocol type, acquisition trigger condition, data transmission rate and anomaly handling threshold corresponding to multiple source devices.

[0026] Based on the driver configuration parameters, parameters are assigned to at least one subroutine in the preset driver to obtain the target driver. The at least one subroutine includes at least one of the following: protocol adaptation subroutine, acquisition control subroutine, data caching subroutine, and exception response subroutine.

[0027] In one possible implementation, the target driver includes a protocol adaptation subroutine, a data acquisition control subroutine, a data caching subroutine, and an exception response subroutine. Running the target driver acquires the acquisition data corresponding to the acquisition task, including:

[0028] Run the protocol adaptation subroutine to establish communication links with multiple source devices;

[0029] Run the acquisition control subroutine to send acquisition commands to multiple source devices respectively. The acquisition commands include acquisition trigger conditions, acquisition frequency and data precision.

[0030] After receiving the first intermediate data sent by each source device, the data caching subroutine is run to store the first intermediate data corresponding to each source device in real time according to the data caching mode corresponding to each source device.

[0031] Run the exception response subroutine to perform exception detection processing on the first intermediate data corresponding to each source device, and obtain the detection result corresponding to each first intermediate data; the detection result is either detection passed or detection failed.

[0032] Multiple first intermediate data points that passed the test were identified as collected data.

[0033] In one possible implementation, the collected data undergoes format conversion and data annotation processing based on the output data parameters to generate output data corresponding to the data collection task, including:

[0034] Based on the output data parameters, determine the target data format and annotation parameters;

[0035] According to the target data format, the collected data is converted to obtain the second intermediate data, which includes structured data and / or unstructured data.

[0036] Based on the annotation parameters, the structured and / or unstructured data are annotated to generate the output data corresponding to the data acquisition task.

[0037] In one possible implementation, structured and unstructured data are labeled according to annotation parameters to generate output data corresponding to the data acquisition task, including:

[0038] Based on the annotation parameters, a first annotation model and a second annotation model are determined. The first annotation model is used to annotate structured data, and the second annotation model is used to annotate unstructured data.

[0039] Based on the first annotation model, the structured data is subjected to the first annotation process to obtain the first output data;

[0040] Based on the second annotation model, the unstructured data is processed with a second annotation to obtain the second output data;

[0041] The output data includes first output data and second output data.

[0042] Secondly, embodiments of this application provide a data acquisition device for an embodied intelligent training field, the device comprising:

[0043] The first determining module is used to determine the acquisition data parameters and output data parameters corresponding to the received acquisition task in response to the acquisition task.

[0044] The second determining module is used to determine multiple source devices and their respective configuration information based on the collected data parameters.

[0045] The generation module is used to generate the target driver based on the collected data parameters and the configuration information of multiple source devices;

[0046] The runtime module is used to run the target driver program to obtain the acquisition data corresponding to the acquisition task.

[0047] The processing module is used to perform format conversion and data annotation on the collected data according to the output data parameters, and generate the output data corresponding to the collection task.

[0048] In one possible implementation, the data acquisition task includes acquiring quality requirement information and acquisition time limit requirement information, and the data acquisition parameters include acquisition priority; the first determining module is specifically used for:

[0049] Based on the information on the required collection quality and the required collection time, determine the collection priority for each collection task.

[0050] In one possible implementation, the information required for collecting quality data includes the scarcity of the data to be collected, current collection environment information, and collection cost; the first determining module is specifically used for:

[0051] The expected information to be collected is determined based on the scarcity of the data to be collected, the current collection environment, and the collection cost.

[0052] Based on the expected data collection information and the time limit requirements for data collection, determine the collection priority corresponding to the collection task.

[0053] In one possible implementation, the generation module is specifically used for:

[0054] Based on the collected data parameters and the configuration information of multiple source devices, generate task templates corresponding to each of the multiple source devices;

[0055] Based on the task templates corresponding to multiple source devices, the preset driver is configured to obtain the target driver.

[0056] In one possible implementation, the generation module is specifically used for:

[0057] Based on the collected data parameters, at least one acquisition element is determined. The at least one acquisition element includes at least one of the following: data modality type, acquisition frequency, data precision, and data storage format.

[0058] Based on the configuration information of at least one acquisition element and multiple source devices, generate acquisition trigger conditions, data caching modes and exception handling modes corresponding to multiple source devices respectively;

[0059] Based on the acquisition trigger conditions, data caching mode, and exception handling mode corresponding to each of the multiple source devices, generate task templates corresponding to each of the multiple source devices.

[0060] In one possible implementation, the generation module is specifically used for:

[0061] The task templates corresponding to multiple source devices are parsed and processed to extract driver configuration parameters. The driver configuration parameters include at least one of the communication protocol type, acquisition trigger condition, data transmission rate and anomaly handling threshold corresponding to multiple source devices.

[0062] Based on the driver configuration parameters, parameters are assigned to at least one subroutine in the preset driver to obtain the target driver. The at least one subroutine includes at least one of the following: protocol adaptation subroutine, acquisition control subroutine, data caching subroutine, and exception response subroutine.

[0063] In one possible implementation, the target driver includes a protocol adaptation subroutine, an acquisition control subroutine, a data buffering subroutine, and an exception response subroutine. The runtime module is specifically used for:

[0064] Run the protocol adaptation subroutine to establish communication links with multiple source devices;

[0065] Run the acquisition control subroutine to send acquisition commands to multiple source devices respectively. The acquisition commands include acquisition trigger conditions, acquisition frequency and data precision.

[0066] After receiving the first intermediate data sent by each source device, the data caching subroutine is run to store the first intermediate data corresponding to each source device in real time according to the data caching mode corresponding to each source device.

[0067] Run the exception response subroutine to perform exception detection processing on the first intermediate data corresponding to each source device, and obtain the detection result corresponding to each first intermediate data; the detection result is either detection passed or detection failed.

[0068] Multiple first intermediate data points that passed the test were identified as collected data.

[0069] In one possible implementation, the processing module is specifically used for:

[0070] Based on the output data parameters, determine the target data format and annotation parameters;

[0071] According to the target data format, the collected data is converted to obtain the second intermediate data, which includes structured data and / or unstructured data.

[0072] Based on the annotation parameters, the structured and / or unstructured data are annotated to generate the output data corresponding to the data acquisition task.

[0073] In one possible implementation, the processing module is specifically used for:

[0074] Based on the annotation parameters, a first annotation model and a second annotation model are determined. The first annotation model is used to annotate structured data, and the second annotation model is used to annotate unstructured data.

[0075] Based on the first annotation model, the structured data is subjected to the first annotation process to obtain the first output data;

[0076] Based on the second annotation model, the unstructured data is processed with a second annotation to obtain the second output data;

[0077] The output data includes first output data and second output data.

[0078] Thirdly, this application provides a chip on which a computer program is stored. When the computer program is executed by the chip, it implements the data acquisition method of the embodied intelligent training field as described in any of the first aspects.

[0079] Fourthly, this application provides a chip module on which a computer program is stored. When the computer program is executed by the chip module, it implements the data acquisition method of the embodied intelligent training field as described in any of the first aspects.

[0080] Fifthly, embodiments of this application provide an electronic device, including: a memory, a processor, and a transceiver;

[0081] The memory stores instructions that the computer executes;

[0082] The processor executes computer execution instructions stored in memory to implement a data acquisition method for an embodied intelligent training field, as described in any of the first aspects.

[0083] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the data acquisition method for the embodied intelligent training field of any one of the first aspects.

[0084] In a seventh aspect, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the data acquisition method for the embodied intelligent training field of any one of the first aspects.

[0085] This application provides a data acquisition method and electronic device for an embodied intelligence training field. In response to a received acquisition task, the method determines the acquisition data parameters and output data parameters corresponding to the task. Based on the acquisition data parameters, it determines multiple source devices and their respective configuration information. Based on the acquisition data parameters and the configuration information of the multiple source devices, it generates a target driver program. The target driver program is then run to acquire the acquisition data corresponding to the task. Based on the output data parameters, the acquired data undergoes format conversion and data annotation to generate the output data corresponding to the task. This achieves accurate adaptation and efficient data acquisition from multiple source devices in the embodied intelligence training field. Through format conversion and data annotation, standardized, high-quality output data is generated, effectively supporting the training needs of embodied intelligence models and improving data acquisition efficiency. Attached Figure Description

[0086] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0087] Figure 1 This application provides an embodiment of the architecture diagram of a system for a data acquisition method in an embodied intelligent training field.

[0088] Figure 2 A flowchart illustrating a data acquisition method for an embodied intelligence training field provided in this application. Figure 1 ;

[0089] Figure 3 A flowchart illustrating a data acquisition method for an embodied intelligence training field provided in this application. Figure 2 ;

[0090] Figure 4 A flowchart illustrating a data acquisition method for an embodied intelligence training field provided in this application. Figure 3 ;

[0091] Figure 5 A schematic diagram of the structure of a data acquisition device for an embodied intelligent training field provided in this application;

[0092] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0093] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0094] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of test systems consistent with some aspects of this application as detailed in the appended claims.

[0095] In this application, the term "comprising" and its variations can refer to non-limiting inclusion; the term "or" and its variations can refer to "and / or". The terms "first", "second", etc., in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0096] As various industries upgrade towards digitalization and intelligence, multimodal data from multiple sources has become the foundational data for precise robot control, algorithm model training, and scenario-based service optimization.

[0097] Related data acquisition technologies are usually designed for single-scenario data acquisition. For example, for devices from specific manufacturers, a matching dedicated driver is manually selected, acquisition parameters are manually configured, raw data is acquired using independent tools, and then data annotation is done manually.

[0098] While the above data acquisition process can meet the single data acquisition needs in simple scenarios, it cannot meet the data acquisition needs of multi-source heterogeneous devices, resulting in low data acquisition efficiency.

[0099] To address the aforementioned technical issues, this application provides a data acquisition method for an embodied intelligence training ground. Targeting the data acquisition needs of an embodied intelligence training ground, the method uses the acquisition data parameters and output data parameters corresponding to the acquisition task as drivers. First, it matches multiple source devices and their corresponding configuration information using the acquisition data parameters and generates a target driver program to acquire the acquired data. Then, based on the output data parameters, it performs format conversion and annotation processing on the acquired data, ultimately achieving standardized data output that meets the task requirements. This achieves accurate adaptation and efficient data acquisition from multiple source devices in the embodied intelligence training ground. Through format conversion and data annotation, standardized, high-quality output data is formed, effectively supporting the training needs of embodied intelligence models and improving data acquisition efficiency.

[0100] Below, in conjunction with Figure 1 The system architecture of the data acquisition method used in the embodied intelligence training field is illustrated with an example.

[0101] Figure 1 This is a schematic diagram of the system architecture for a data acquisition method used in an embodied intelligent training field, as provided in an embodiment of this application. Please refer to... Figure 1 , Figure 1 It can include a data acquisition system, which can include a task parsing layer, a device management layer, a driver generation layer, a data acquisition layer, and a data processing layer. Each layer can work together efficiently through data interfaces to complete the entire data acquisition process.

[0102] The task parsing layer can serve as the system's instruction entry point. It can receive externally initiated data acquisition tasks and extract and determine the data acquisition parameters and output data parameters based on the task description information.

[0103] For example, when the task parsing layer receives the task of "collecting visual and force data of the grasping action of the embodied intelligent robot", the layer can automatically parse the collected data parameters as "visual image data: resolution 1920×1080, frame rate 30fps, force sensing data: sampling rate 1000Hz, output data parameters: JSON format, including action type and force range labeling".

[0104] The device management layer can pre-store attribute information of various source devices in the embodied intelligent training field. For example, the attribute information may include device type, data acquisition range, communication protocol, etc.

[0105] Once the data collection parameters are obtained, this level can use a parameter matching algorithm to filter out multiple source devices that meet the collection requirements, and retrieve the corresponding configuration information of each source device, including device communication address, working mode parameters, data transmission protocol, etc.

[0106] For example, for visual and force data acquisition tasks, two types of source devices can be matched: high-definition industrial cameras and six-dimensional force sensors, and specific information such as camera exposure parameters and sensor range configuration can be determined.

[0107] After receiving the collected data parameters from the task parsing layer and the device configuration information from the device management layer, the driver generation layer can generate target drivers that can be adapted to work collaboratively with various source devices through the built-in driver template library and parameter mapping rules.

[0108] The target driver can include startup instructions for each source device, data acquisition frequency control logic, and can also integrate a multi-device data synchronization transmission protocol to ensure that the data collected by different types of devices remains consistent in the time dimension, avoiding the problem of training data failure caused by data asynchrony between devices.

[0109] The data acquisition layer can be used to run the target driver, establish a stable communication link with each source device, acquire the raw data corresponding to the acquisition task in real time, and temporarily store the raw data in the system's temporary data buffer.

[0110] During the data acquisition process, this level can also be used for data integrity detection. When a data transmission interruption or data packet loss is detected in a source device, the driver's reconnection mechanism will be automatically triggered to ensure the continuity of the acquired data.

[0111] The data processing layer can serve as the system's output end. Based on the output data parameters determined by the task parsing layer, it performs format conversion and data annotation on the raw collected data in the temporary buffer. The processed data is then determined to be the output data that meets the requirements of the collection task and will be transmitted to the embodied intelligent training database to provide high-quality data support for model training.

[0112] For example, format conversion processing can uniformly convert RAW format images output by industrial cameras and binary force value data output by force sensors into JSON format required by the output parameters.

[0113] Data annotation processing can add semantic annotations to data based on preset annotation rules, such as labeling images of grasping actions with action labels such as "clamping" and "releasing", and labeling force data with status labels such as "safe range" and "overload".

[0114] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0115] Figure 2 A flowchart illustrating a data acquisition method for an embodied intelligence training field provided in this application. Figure 1 ,like Figure 2 As shown, the method includes:

[0116] S201. In response to the received acquisition task, determine the acquisition data parameters and output data parameters corresponding to the acquisition task.

[0117] A data collection task can refer to a data collection request instruction initiated in an embodied intelligence training field to support model training.

[0118] The methods for receiving data collection tasks include, but are not limited to, user input through a human-computer interaction interface, external training systems calling through API interfaces, and automatic triggering by the system according to the training plan.

[0119] A data collection task can include descriptive information such as the data collection object, the collection scenario, and the purpose of the data.

[0120] For example, the data collection task could be "collecting visual and tactile data of the robot's actions in grasping a water cup".

[0121] Data acquisition parameters can be data parameters that characterize the requirements of the acquisition task for the raw data.

[0122] Data acquisition parameters can include information such as data type, acquisition precision, acquisition frequency, and data dimensions.

[0123] Data types can include visual images, force sensing data, posture data, etc.

[0124] Acquisition accuracy can include resolution, measurement error range, etc.

[0125] Data dimensions can include frame rate, sampling rate, etc.

[0126] Output data parameters can be used to characterize the format and semantic requirements of the final delivered data for the acquisition task.

[0127] Output data parameters can include information such as target data format, annotation rules, and data storage path.

[0128] The target data format can include JSON, CSV, TFRecord, etc.

[0129] Labeling rules can include label type, labeling granularity, etc.

[0130] Once the system receives a data acquisition task, it can parse and process the task. Based on the task description, it can extract data acquisition features such as data type, acquisition accuracy requirements, acquisition frequency, and data acquisition duration. It can also extract output features such as target data format, annotation rules, and data storage path. Based on the acquisition and output features, it can determine the acquisition data parameters and output data parameters corresponding to the acquisition task.

[0131] Optionally, the data acquisition task may include data acquisition quality requirements and data acquisition time limit requirements, and the data acquisition parameters may include acquisition priority. The data acquisition parameters and output data parameters corresponding to the data acquisition task may be determined in response to the received data acquisition task in the following manner: the acquisition priority corresponding to the data acquisition task is determined based on the data acquisition quality requirements and data acquisition time limit requirements.

[0132] Information on quality requirements for data collection can be key constraints to ensure that the collected data meets the needs of model training.

[0133] Information on data collection time limits can be constrained information that balances data utility with collection efficiency.

[0134] Information on collection time limits may include task completion deadline, data real-time performance level, and scheduling response window.

[0135] Collection priority can reflect the value weight and urgency of a task.

[0136] The quality requirements and time limits for data collection can be input into a preset priority prediction model to obtain the collection priority corresponding to the data collection task.

[0137] Optionally, the collection quality requirements information includes the scarcity of the data to be collected, the current collection environment information, and the collection cost. The collection priority corresponding to the collection task can be determined based on the collection quality requirements information and the collection time limit requirements information in the following ways: determine the expected data collection information based on the scarcity of the data to be collected, the current collection environment information, and the collection cost; determine the collection priority corresponding to the collection task based on the expected data collection information and the collection time limit requirements information.

[0138] Among them, the scarcity of data to be collected can refer to the completeness of the target data corresponding to the task to be collected in the existing dataset or model knowledge system.

[0139] The higher the scarcity, the greater the value of the data to business needs.

[0140] The current data acquisition environment information can refer to the real-time environmental conditions and related constraints of the target acquisition area when the acquisition task is performed. The scarcity of the data to be acquired and the current data acquisition environment information can be user-defined or pre-set.

[0141] Real-time environmental conditions can include sensor operating status, communication link stability, and so on.

[0142] Data collection cost can refer to the quantitative indicators of various resources required to complete the data collection task.

[0143] Data collection costs can include robot travel distance, task execution time, energy consumption, equipment wear and tear, etc.

[0144] The expected information in data collection can refer to the potential value information about the effectiveness of the data collection task.

[0145] An expected information prediction function can be constructed to determine the expected information for data collection based on the scarcity of the data to be collected, the current collection environment information, and the collection cost.

[0146] This step transforms vague data collection requirements into clear and actionable parameters, providing clear guidance for subsequent equipment matching, data collection, and data processing.

[0147] S202. Based on the collected data parameters, determine the configuration information of multiple source devices and their respective counterparts.

[0148] Source devices can refer to hardware devices with data acquisition capabilities in an embodied intelligent training field.

[0149] Source devices may include, but are not limited to, industrial cameras, six-dimensional force sensors, lidar, inertial measurement units, microphones, etc.

[0150] Configuration information can refer to a set of key parameters that enable the source device to function properly and collect data.

[0151] Configuration information may include device communication address, operating mode parameters, data transmission protocol, etc.

[0152] The device communication address may include IP address, serial port number, etc., and is not limited here.

[0153] Operating mode parameters may include camera exposure time, sensor range, etc., which are not limited here.

[0154] Data transmission protocols may include TCP / IP, CAN bus, USB 3.0, etc., and are not limited here.

[0155] Based on the collected data parameters, multiple source devices that meet the collection requirements can be selected from the device resource pool of the embodied intelligent training field, and the configuration information corresponding to each source device can be retrieved.

[0156] For example, if the data acquisition parameters are "visual image resolution 1920×1080, frame rate 30fps", then select high-definition industrial cameras that support this resolution and frame rate; if the data acquisition parameters are "joint torque sampling rate 500Hz, measurement range 0-50N・m", then select six-dimensional force sensors that meet this sampling rate and range.

[0157] Optionally, it can support the simultaneous filtering of multiple types of source devices to meet multi-dimensional data acquisition needs. For example, when acquiring visual, force, and posture data at the same time, it can match three types of source devices: industrial cameras, force sensors, and IMUs.

[0158] Optionally, the configuration information corresponding to the source device can be adapted to the collected data parameters and the device's own characteristics.

[0159] Configuration information can include communication configuration, operating parameters, transmission protocols, and so on.

[0160] For example, the communication configuration of an industrial camera is the IP address 192.168.1.100; the communication configuration of a force sensor is the serial port number COM3.

[0161] For example, the operating parameters of an industrial camera are an exposure time of 10ms and an automatic white balance mode; the operating parameters of a force sensor are a range of 0-50N·m and a filter frequency of 100Hz.

[0162] For example, industrial cameras can use the TCP / IP protocol to transmit image data, while force sensors can use the CAN bus protocol to transmit torque data.

[0163] This step enables a precise match between data collection needs and device resources, laying the hardware foundation for subsequent driver generation and data collection.

[0164] S203. Generate the target driver based on the collected data parameters and the configuration information of multiple source devices.

[0165] A target driver can refer to program code generated based on the acquired data parameters and source device configuration information, which can drive multiple source devices to work together.

[0166] The target driver may include a protocol adaptation subroutine, an acquisition control subroutine, a data caching subroutine, and an exception response subroutine.

[0167] The target driver can be used to achieve collaborative data acquisition from multiple sources and ensure that the acquired data meets the parameter requirements.

[0168] The target driver generation process can be based on the collected data parameters and source device configuration information. Through the built-in driver template library and preset parameter mapping rules, the collected data parameters are converted into control commands that the device can recognize.

[0169] The driver template library can include basic driver frameworks and general data processing logic for different types of source devices.

[0170] For example, the basic driver framework may include camera startup templates, sensor data reading templates, etc.

[0171] Based on the source device, the corresponding basic driver framework can be called from the driver template library. Based on the configuration information and collected data parameters, device control instructions can be generated through parameter mapping rules to form the target driver program.

[0172] S204. Run the target driver program to obtain the acquisition data corresponding to the acquisition task.

[0173] Acquired data can refer to the raw data collected by the source device according to the control logic of the target driver, without undergoing format conversion or semantic annotation processing.

[0174] The acquired data may include RAW format images output by the camera, binary force data output by the sensor, etc.

[0175] The target driver can be run, enabling it to establish a stable communication link with each source device based on the communication address and transmission protocol in the configuration information. This completes the device handshake and initialization, sends synchronous acquisition commands to each source device, and triggers the devices to start data acquisition according to the data acquisition parameter requirements. The raw data acquired by each source device is transmitted to the system in real time through a preset transmission protocol. The system temporarily stores the raw data in a temporary data buffer and monitors the data transmission status in real time during the acquisition process. If abnormalities such as data packet loss or transmission interruption are detected, the system automatically triggers a reconnection mechanism and re-acquires missing data to ensure the continuity and integrity of the acquired data and obtain the acquisition data corresponding to the acquisition task.

[0176] Optionally, the target driver includes a protocol adaptation subroutine, an acquisition control subroutine, a data caching subroutine, and an exception response subroutine. The target driver can be run in the following manner to obtain the acquisition data corresponding to the acquisition task: The protocol adaptation subroutine is run to establish communication links with multiple source devices; the acquisition control subroutine is run to send acquisition commands to the multiple source devices respectively, the acquisition commands including acquisition trigger conditions, acquisition frequency, and data precision; after receiving the first intermediate data sent by each source device, the data caching subroutine is run to store the first intermediate data corresponding to each source device in real time according to the data caching mode corresponding to each source device; the exception response subroutine is run to perform exception detection processing on the first intermediate data corresponding to each source device, obtaining the detection result corresponding to each first intermediate data; the detection result is either detection passed or detection failed; the multiple first intermediate data with detection results of passing are determined as the acquisition data.

[0177] The protocol adaptation subroutine can be used to adapt to the communication protocols of different source devices and establish and maintain a stable communication link with the source devices.

[0178] The acquisition control subroutine can be used to generate standardized acquisition commands and accurately distribute them to multiple source devices, and manage data acquisition triggering, frequency, accuracy and other behaviors.

[0179] The data caching subroutine can be used to receive and store the first intermediate data returned by the source device in real time, and supports caching according to device-specific rules.

[0180] The anomaly response subroutine can be used to perform multi-dimensional anomaly detection and classification processing on the first intermediate data returned by the source device.

[0181] This step enables collaborative and efficient data acquisition from multiple sources, obtaining raw data that meets parameter requirements.

[0182] S205. Based on the output data parameters, perform format conversion and data annotation on the collected data to generate the output data corresponding to the collection task.

[0183] Format conversion processing refers to the process of converting heterogeneous raw data output from different source devices into a unified format according to the requirements of the output data parameters, so as to ensure the universality of the data and the compatibility of subsequent processing.

[0184] For example, converting RAW format images from industrial cameras to JPG format and converting binary torque data from force sensors to JSON format numerical data ensures the universality and compatibility of data formats, facilitating subsequent model reading and processing.

[0185] Data labeling processing refers to the process of adding semantic labels to the converted collected data based on the preset labeling rules of the output data parameters, so as to improve the semantic information of the data and meet the training needs of embodied intelligent models.

[0186] Output data can refer to standardized, high-quality data that meets the requirements of the data collection task after format conversion and data labeling, and can be directly used for training, verification or testing of embodied intelligence models.

[0187] Based on the output data parameters, the original collected data in the temporary data buffer can be format-converted and labeled to generate the output data corresponding to the collection task.

[0188] Among them, the format conversion process can uniformly convert heterogeneous raw data output from different source devices according to the target format required by the output data parameters.

[0189] Data annotation processing can add semantic tags to the converted collected data based on the preset annotation rules of the output data parameters.

[0190] The annotation rules can be pre-stored in the system's annotation rule library, including action label library, status label library, and numerical range label library.

[0191] Optionally, the annotation process can combine automatic annotation with manual verification: corresponding labels can be matched according to data features to generate preliminary annotation results; if the data collection task requires high annotation accuracy, the preliminary annotation results can be manually checked and corrected to ensure annotation accuracy.

[0192] Optionally, the output data can be transmitted to the embodied intelligence training database according to the storage path specified by the output data parameters. This data can be directly used for the training, validation or testing of the embodied intelligence model, providing high-quality data support for model performance optimization.

[0193] This application provides a data acquisition method for an embodied intelligence training ground. In response to a received acquisition task, the method determines the acquisition data parameters and output data parameters corresponding to the task. Based on the acquisition data parameters, it determines multiple source devices and their respective configuration information. Based on the acquisition data parameters and the configuration information of the multiple source devices, it generates a target driver program. The target driver program is then run to acquire the acquisition data corresponding to the task. Based on the output data parameters, the acquired data undergoes format conversion and data annotation to generate the output data corresponding to the task. This method achieves accurate adaptation and efficient data acquisition from multiple source devices in the embodied intelligence training ground. Through format conversion and data annotation, standardized, high-quality output data is generated, effectively supporting the training needs of embodied intelligence models and improving data acquisition efficiency.

[0194] Below, in conjunction with Figure 3 The process of generating the target driver based on the collected data parameters and the configuration information of multiple source devices is explained.

[0195] Figure 3 A flowchart illustrating a data acquisition method for an embodied intelligence training field provided in this application. Figure 2 ,like Figure 3 As shown, in this embodiment... Figure 2 Based on the embodiments, a data acquisition method for the embodied intelligence training field is described in detail, the method including:

[0196] S301. Based on the collected data parameters and the configuration information of multiple source devices, generate task templates corresponding to multiple source devices respectively.

[0197] Task templates can be standardized configuration files customized for a single source device.

[0198] The collected data parameters and configuration information of multiple source devices can be input into the task template generation model to obtain the first template corresponding to each of the multiple source devices. The integrity of the first template corresponding to each source device is verified. For any first template, after the first template verification is passed, the first template is determined as the task template.

[0199] The task template generation model can be a pre-trained neural network model, and there are no restrictions on this.

[0200] Optionally, task templates corresponding to multiple source devices can be generated based on the collected data parameters and the configuration information of multiple source devices in the following manner: determine at least one collection element based on the collected data parameters; generate collection trigger conditions, data caching modes, and exception handling modes corresponding to multiple source devices based on at least one collection element and the configuration information of multiple source devices; and generate task templates corresponding to multiple source devices based on the collection trigger conditions, data caching modes, and exception handling modes corresponding to multiple source devices.

[0201] Among them, at least one acquisition element includes at least one of the following: data modality type, acquisition frequency, data precision, and data storage format.

[0202] Acquisition elements are the elements that can be extracted from the acquired data parameters and directly guide the device's acquisition behavior.

[0203] The data acquisition trigger condition can be a trigger rule that initiates the data acquisition operation on the source device.

[0204] Data caching mode can be a way of storing collected data.

[0205] The exception handling mode can be used for standardized handling of anomalies in the collected data.

[0206] For example, suppose the collected data parameters include: ① Temperature and humidity monitoring task: used for environmental control, temperature error ≤0.5℃, humidity error ≤1%, once every 10 minutes, compatible with local database; ② Equipment appearance monitoring task: used for defect identification, once every hour, clear and identifiable images, supports cloud storage; the source device configuration information includes: ① Device 1: Temperature and humidity sensor, ID=TH-S001, protocol=Modbus RTU, serial port=COM5, maximum acquisition frequency=1 time / second, supports periodic triggering, local cache=512MB; ② Device 2: Industrial camera, ID=Cam-S002, protocol=HTTP, IP=192.168.1.103, supports periodic / manual triggering, single image=3MB, local cache=200MB.

[0207] Based on the collected data parameters, at least one data acquisition element is determined. The data acquisition element for Device 1 is: Data modality type = numerical, acquisition frequency = 1 time / 10 minutes, data precision = temperature retained to 1 decimal place, humidity integer, data storage format = CSV. The data acquisition element for Device 2 is: Data modality type = image, acquisition frequency = 1 time / hour, data precision = resolution 1920×1080, data storage format = JPG.

[0208] Based on the configuration information of at least one acquisition element and multiple source devices, acquisition trigger conditions, data caching modes, and anomaly handling modes are generated for each source device. For Device 1, the acquisition trigger condition is periodic triggering with a 10-minute interval; the data caching mode is local caching by date, one CSV file per day; the anomaly handling mode is based on temperature > 50℃ or < -10℃, humidity > 95% or < 10%, indicating an anomaly, triggering a retry of acquisition. For Device 2, the acquisition trigger condition is periodic triggering with a 1-hour interval, pausing acquisition from 00:00 to 06:00 daily; the data caching mode is local temporary caching, retaining the latest 5 images, immediately synchronizing to the cloud after acquisition; the anomaly handling mode is based on image resolution < 1920×1080, triggering camera parameter calibration and retaking the image.

[0209] Based on the above information, task templates for two devices are generated for configuring preset drivers.

[0210] S302. Based on the task templates corresponding to multiple source devices, configure the preset driver to obtain the target driver.

[0211] After the preset driver is started, it reads the task templates corresponding to multiple source devices, automatically parses the core configuration information in each template, loads the corresponding configuration information, and obtains the target driver.

[0212] Optionally, the target driver can be obtained by configuring the preset driver according to the task templates corresponding to the multiple source devices as follows: the task templates corresponding to the multiple source devices are parsed and the driver configuration parameters are extracted; according to the driver configuration parameters, at least one subroutine in the preset driver is assigned parameter values ​​to obtain the target driver.

[0213] Among them, the corresponding parameters can be extracted from the task template of each device according to the functional dimensions of the preset driver subroutine to obtain the driver configuration parameters.

[0214] The driver configuration parameters include at least one of the following: communication protocol type, acquisition trigger condition, data transmission rate, and anomaly handling threshold, corresponding to multiple source devices.

[0215] At least one subroutine includes at least one of the following: a protocol adaptation subroutine, an acquisition control subroutine, a data caching subroutine, and an exception response subroutine.

[0216] It can obtain a mapping table of preset driver configuration parameters and subroutine parameter variables, determine at least one subroutine parameter variable based on the mapping table, and assign parameter values ​​to at least one subroutine in the preset driver based on the at least one subroutine parameter variable to obtain the target driver.

[0217] The implementation details of each step in this application embodiment can be found in the description of the corresponding steps or operations in the above method embodiments; repeated content will not be repeated.

[0218] This application provides a data acquisition method for an embodied intelligent training field. It generates task templates corresponding to multiple source devices based on the acquired data parameters and configuration information of multiple source devices. Then, based on the task templates corresponding to each source device, it configures a preset driver to obtain a target driver. This achieves precise matching between the acquired data parameters and the configurations of multiple source devices, quickly customizing a general preset driver to a target driver adapted to each source device, ensuring accurate and efficient execution of the data acquisition task, and improving data acquisition efficiency.

[0219] Below, in conjunction with Figure 4 This section explains the process of converting the format of the collected data and labeling the data based on the output data parameters to generate the output data corresponding to the collection task.

[0220] Figure 4 A flowchart illustrating a data acquisition method for an embodied intelligence training field provided in this application. Figure 3 ,like Figure 4 As shown, in this embodiment... Figure 2 or Figure 3 Based on the embodiments, a data acquisition method for the embodied intelligence training field is described in detail, the method including:

[0221] S401. Determine the target data format and annotation parameters based on the output data parameters.

[0222] The target data format can be a standardized data format determined based on the output data parameters and conforming to the requirements of subsequent application scenarios or receiving ends.

[0223] Labeling parameters can be a set of rules that guide data labeling operations, determined based on output data parameters.

[0224] The output data parameters can be parsed and processed to determine the target data format and annotation parameters.

[0225] S402. Based on the target data format, perform format conversion processing on the collected data to obtain the second intermediate data.

[0226] The second intermediate data includes structured data and / or unstructured data.

[0227] The second intermediate data can be intermediate data obtained after the collected data has undergone format conversion processing.

[0228] Structured data can be data with a fixed data structure, clearly defined fields, and can be directly parsed and processed by database or spreadsheet tools.

[0229] Unstructured data can be data without a fixed data structure, which requires specific parsing tools to extract information.

[0230] Format conversion processing can include data structure reorganization, encoding conversion, field mapping, and removal of redundant information.

[0231] The collected data can be preprocessed, and the format of the preprocessed collected data can be converted according to the target data format to obtain the second intermediate data.

[0232] The preprocessed acquisition data may include structured acquisition data and / or unstructured acquisition data.

[0233] Format conversion processing for structured acquired data can be performed according to the first conversion format rule. For example, binary acquired data can be converted to decimal values.

[0234] For unstructured acquired data, format conversion can be performed according to the second conversion format rules. For example, RAW format images can be converted to JPG format images.

[0235] S403. Based on the annotation parameters, perform annotation processing on structured data and / or unstructured data to generate output data corresponding to the data acquisition task.

[0236] Labeling can be the process of adding semantic labels, attribute information, or spatial / temporal tags to second intermediate data according to the rules and requirements of the labeling parameters.

[0237] Based on the annotation parameters, an automatic annotation model can be used to annotate structured and / or unstructured data, generating output data corresponding to the data collection task.

[0238] Based on the annotation parameters, an automatic annotation model and a verification model can be used to annotate structured and / or unstructured data, generating output data corresponding to the data collection task.

[0239] Optionally, structured and unstructured data can be labeled according to the labeling parameters to generate output data corresponding to the data collection task in the following manner: determine the first labeling model and the second labeling model according to the labeling parameters; perform the first labeling process on the structured data according to the first labeling model to obtain the first output data; and perform the second labeling process on the unstructured data according to the second labeling model to obtain the second output data.

[0240] The first annotation model is used to annotate structured data.

[0241] The first annotation model can be a mapping rule engine or a machine learning model. After training with labeled samples, the model weights and decision thresholds are configured to achieve multi-field association annotation. No specific limitation is made here.

[0242] The second annotation model is used to annotate unstructured data.

[0243] Based on the annotation configuration of the unstructured data, select the appropriate second annotation model, and select a dedicated deep learning model based on the data modality type.

[0244] Data modal types can include text, images, audio, etc.

[0245] The output data includes the first output data and the second output data.

[0246] The implementation details of each step in this application embodiment can be found in the description of the corresponding steps or operations in the above method embodiments; repeated content will not be repeated.

[0247] This application provides a data acquisition method for an embodied intelligence training field. The method involves determining the target data format and annotation parameters based on output data parameters; performing format conversion on the acquired data according to the target data format to obtain second intermediate data, which includes structured and / or unstructured data; and then annotating the structured and / or unstructured data according to the annotation parameters to generate output data corresponding to the acquisition task. In this way, through format conversion and standardized annotation, the acquired data accurately matches the output data parameter requirements, generating standardized annotated output data of structured and / or unstructured data. This provides high-quality data support for subsequent model training, business analysis, and other scenarios, improving data acquisition efficiency.

[0248] Figure 5 This application provides a schematic diagram of the structure of a data acquisition device for an embodied intelligent training field, as shown below. Figure 5 As shown, the data acquisition device 500 for embodied intelligent training field provided in this embodiment includes a first determining module 501, a second determining module 502, a generating module 503, a running module 504, and a processing module 505.

[0249] The first determining module 501 is used to determine the acquisition data parameters and output data parameters corresponding to the acquisition task in response to the received acquisition task.

[0250] The second determining module 502 is used to determine multiple source devices and their respective configuration information based on the collected data parameters.

[0251] The generation module 503 is used to generate a target driver based on the collected data parameters and the configuration information of multiple source devices;

[0252] The execution module 504 is used to run the target driver program to obtain the acquisition data corresponding to the acquisition task.

[0253] The processing module 505 is used to perform format conversion and data annotation processing on the collected data according to the output data parameters, and generate the output data corresponding to the collection task.

[0254] In one possible implementation, the data acquisition task includes acquiring quality requirement information and acquisition time limit requirement information, and the data acquisition parameters include acquisition priority; the first determining module 501 is specifically used for:

[0255] Based on the information on the required collection quality and the required collection time, determine the collection priority for each collection task.

[0256] In one possible implementation, the information required to collect quality data includes the scarcity of the data to be collected, current collection environment information, and collection cost; the first determining module 501 is specifically used for:

[0257] The expected information to be collected is determined based on the scarcity of the data to be collected, the current collection environment, and the collection cost.

[0258] Based on the expected data collection information and the time limit requirements for data collection, determine the collection priority corresponding to the collection task.

[0259] In one possible implementation, the generation module 503 is specifically used for:

[0260] Based on the collected data parameters and the configuration information of multiple source devices, generate task templates corresponding to each of the multiple source devices;

[0261] Based on the task templates corresponding to multiple source devices, the preset driver is configured to obtain the target driver.

[0262] In one possible implementation, the generation module 503 is specifically used for:

[0263] Based on the collected data parameters, at least one acquisition element is determined. The at least one acquisition element includes at least one of the following: data modality type, acquisition frequency, data precision, and data storage format.

[0264] Based on the configuration information of at least one acquisition element and multiple source devices, generate acquisition trigger conditions, data caching modes and exception handling modes corresponding to multiple source devices respectively;

[0265] Based on the acquisition trigger conditions, data caching mode, and exception handling mode corresponding to each of the multiple source devices, generate task templates corresponding to each of the multiple source devices.

[0266] In one possible implementation, the generation module 503 is specifically used for:

[0267] The task templates corresponding to multiple source devices are parsed and processed to extract driver configuration parameters. The driver configuration parameters include at least one of the communication protocol type, acquisition trigger condition, data transmission rate and anomaly handling threshold corresponding to multiple source devices.

[0268] Based on the driver configuration parameters, parameters are assigned to at least one subroutine in the preset driver to obtain the target driver. The at least one subroutine includes at least one of the following: protocol adaptation subroutine, acquisition control subroutine, data caching subroutine, and exception response subroutine.

[0269] In one possible implementation, the target driver includes a protocol adaptation subroutine, an acquisition control subroutine, a data buffering subroutine, and an exception response subroutine. The execution module 504 is specifically used for:

[0270] Run the protocol adaptation subroutine to establish communication links with multiple source devices;

[0271] Run the acquisition control subroutine to send acquisition commands to multiple source devices respectively. The acquisition commands include acquisition trigger conditions, acquisition frequency and data precision.

[0272] After receiving the first intermediate data sent by each source device, the data caching subroutine is run to store the first intermediate data corresponding to each source device in real time according to the data caching mode corresponding to each source device.

[0273] Run the exception response subroutine to perform exception detection processing on the first intermediate data corresponding to each source device, and obtain the detection result corresponding to each first intermediate data; the detection result is either detection passed or detection failed.

[0274] Multiple first intermediate data points that passed the test were identified as collected data.

[0275] In one possible implementation, the processing module 505 is specifically used for:

[0276] Based on the output data parameters, determine the target data format and annotation parameters;

[0277] According to the target data format, the collected data is converted to obtain the second intermediate data, which includes structured data and / or unstructured data.

[0278] Based on the annotation parameters, the structured and / or unstructured data are annotated to generate the output data corresponding to the data acquisition task.

[0279] In one possible implementation, the processing module 505 is specifically used for:

[0280] Based on the annotation parameters, a first annotation model and a second annotation model are determined. The first annotation model is used to annotate structured data, and the second annotation model is used to annotate unstructured data.

[0281] Based on the first annotation model, the structured data is subjected to the first annotation process to obtain the first output data;

[0282] Based on the second annotation model, the unstructured data is processed with a second annotation to obtain the second output data;

[0283] The output data includes first output data and second output data.

[0284] This embodiment provides a data acquisition device for a embodied intelligent training field, which can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0285] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Please refer to... Figure 6 The electronic device 600 may include: a memory 601, a processor 602, and a transceiver 603.

[0286] Memory 601 is used to store program instructions;

[0287] The processor 602 is used to execute the program instructions stored in the memory so that the electronic device 600 performs the above-described method.

[0288] Transceiver 603 may include a transmitter and / or a receiver. The transmitter may also be referred to as a transmitter, transmitter port, or transmitter interface, and the receiver may also be referred to as a receiver port, receiver interface, or similar descriptions. Exemplarily, memory 601, processor 602, and transceiver 603 are interconnected via bus 604.

[0289] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0290] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0291] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0292] This application also provides a chip on which a computer program is stored. When the computer program is executed by the chip, it implements the above-mentioned data acquisition method for the embodied intelligent training field. The corresponding content and effects can be referred to the method embodiment section, and will not be repeated here.

[0293] This application also provides a chip module, on which a computer program is stored. When the computer program is executed by the chip module, it implements the above-mentioned data acquisition method for the embodied intelligent training field. The corresponding content and effects can be referred to the method embodiment section, and will not be repeated here.

[0294] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described data acquisition method for an embodied intelligence training field.

[0295] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the aforementioned data acquisition method for an embodied intelligent training field.

[0296] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0297] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0298] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0299] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0300] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0301] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0302] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0303] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A data acquisition method for an embodied intelligent training field, characterized in that, The method includes: In response to a received acquisition task, determine the acquisition data parameters and output data parameters corresponding to the acquisition task; Based on the collected data parameters, determine multiple source devices and their corresponding configuration information; Based on the collected data parameters and the configuration information of the multiple source devices, a target driver is generated; Run the target driver program to obtain the acquisition data corresponding to the acquisition task; Based on the output data parameters, the collected data undergoes format conversion and data annotation to generate the output data corresponding to the collection task.

2. The method according to claim 1, characterized in that, The data acquisition task includes data acquisition quality requirements and data acquisition time limit requirements, and the data acquisition parameters include data acquisition priority; in response to the received data acquisition task, the data acquisition parameters and output data parameters corresponding to the data acquisition task are determined, including: Based on the acquisition quality requirements and acquisition time limit requirements, the acquisition priority corresponding to the acquisition task is determined.

3. The method according to claim 2, characterized in that, The information on the quality requirements for data collection includes the scarcity of the data to be collected, the current collection environment information, and the collection cost. Based on the data acquisition quality requirements and time limit requirements, the acquisition priority corresponding to the acquisition task is determined, including: Based on the scarcity of the data to be collected, the current collection environment information, and the collection cost, the expected information for data collection is determined; Based on the expected data collection information and the time limit requirements for data collection, the collection priority corresponding to the collection task is determined.

4. The method according to any one of claims 1-3, characterized in that, Based on the collected data parameters and the configuration information of the multiple source devices, a target driver is generated, including: Based on the collected data parameters and the configuration information of the multiple source devices, generate task templates corresponding to the multiple source devices respectively; Based on the task templates corresponding to the multiple source devices, the preset driver is configured to obtain the target driver.

5. The method according to claim 4, characterized in that, Based on the collected data parameters and the configuration information of the multiple source devices, task templates corresponding to the multiple source devices are generated respectively, including: Based on the collected data parameters, at least one acquisition element is determined, and the at least one acquisition element includes at least one of data modality type, acquisition frequency, data precision, and data storage format; Based on the configuration information of the at least one acquisition element and the plurality of source devices, generate acquisition trigger conditions, data caching modes and exception handling modes corresponding to the plurality of source devices respectively; Based on the acquisition trigger conditions, data caching modes, and exception handling modes corresponding to the multiple source devices, task templates corresponding to the multiple source devices are generated.

6. The method according to claim 4, characterized in that, Based on the task templates corresponding to the multiple source devices, the preset driver is configured to obtain the target driver, including: The task templates corresponding to the multiple source devices are parsed and processed to extract the driver configuration parameters. The driver configuration parameters include at least one of the communication protocol type, acquisition trigger condition, data transmission rate and anomaly handling threshold corresponding to the multiple source devices. Based on the driver configuration parameters, parameters are assigned to at least one subroutine in the preset driver program to obtain the target driver program. The at least one subroutine includes at least one of the following: protocol adaptation subroutine, acquisition control subroutine, data caching subroutine, and exception response subroutine.

7. The method according to any one of claims 1-3, characterized in that, The target driver program includes a protocol adaptation subroutine, a data acquisition control subroutine, a data caching subroutine, and an exception response subroutine. Running the target driver program allows the acquisition of data corresponding to the acquisition task, including: Run the protocol adaptation subroutine to establish communication links with multiple source devices; The acquisition control subroutine is run to send acquisition instructions to the multiple source devices respectively. The acquisition instructions include acquisition trigger conditions, acquisition frequency and data precision. After receiving the first intermediate data sent by each source device, the data caching subroutine is run to store the first intermediate data corresponding to each source device in real time according to the data caching mode corresponding to each source device. The exception response subroutine is run to perform exception detection processing on the first intermediate data corresponding to each source device, and to obtain the detection result corresponding to each first intermediate data; the detection result is either detection passed or detection failed. The first intermediate data points that show the detection result as passed are identified as the collected data.

8. The method according to any one of claims 1-3, characterized in that, Based on the output data parameters, the collected data undergoes format conversion and data annotation processing to generate the output data corresponding to the collection task, including: Based on the output data parameters, determine the target data format and annotation parameters; According to the target data format, the collected data is converted to obtain second intermediate data, which includes structured data and / or unstructured data. Based on the annotation parameters, the structured data and / or the unstructured data are annotated to generate the output data corresponding to the acquisition task.

9. The method according to claim 8, characterized in that, Based on the annotation parameters, the structured data and the unstructured data are annotated to generate the output data corresponding to the data acquisition task, including: Based on the annotation parameters, a first annotation model and a second annotation model are determined. The first annotation model is used to annotate structured data, and the second annotation model is used to annotate unstructured data. Based on the first annotation model, the structured data is subjected to a first annotation process to obtain the first output data; According to the second annotation model, the unstructured data is subjected to a second annotation process to obtain the second output data; The output data includes first output data and second output data.

10. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 9.

Images