Data collection method and device for humanoid robot training ground

Abstract

The embodiment of the application provides a data acquisition method and equipment for a humanoid robot training field, and belongs to the technical field of industrial manufacturing equipment of the humanoid robot training field. The method comprises the following steps: in response to receiving a task instruction, determining a target cleaning position on a to-be-cleaned object; based on a preset weight function, performing trajectory planning according to the target cleaning position to obtain a target trajectory; the preset weight function is associated with a first distance and an arm angle; the first distance is the distance between the target cleaning position and an end effector; performing a cleaning task on the target cleaning position according to the target trajectory; collecting and uploading target data in the task execution process; the target data comprises at least one of the following: the target cleaning position and the target trajectory. The method provided in the embodiment can improve the data acquisition efficiency and data accuracy of the humanoid robot in the training field, and can help improve the work efficiency and safety in the industrial manufacturing industry or feature work scene.

Description

Technical Field

[0001] This application relates to the field of humanoid robot technology, and in particular to a data acquisition method and device for a humanoid robot training field. Background Technology

[0002] Humanoid robots are increasingly used in industrial applications, especially in cleaning tasks requiring high precision, flexibility, and adaptability. To train humanoid robots to perform specific tasks such as pipe sheet cleaning, it is necessary to collect a large amount of data on the robots' task execution processes. This data can then be used to optimize the robot's behavior through machine learning and deep learning algorithms.

[0003] In related technologies, motion capture systems, virtual reality (VR) devices, and teleoperation are commonly used for data acquisition.

[0004] However, in the process of realizing this application, the inventors discovered at least the following problems in the related technology: the efficiency and accuracy of data collection in the above-mentioned methods are low. Summary of the Invention

[0005] This application provides a data acquisition method and device for a humanoid robot training field to improve the efficiency and accuracy of data acquisition.

[0006] In a first aspect, embodiments of this application provide a data acquisition method for a humanoid robot training field, including:

[0007] In response to receiving a task instruction, determine the target cleaning location on the object to be cleaned;

[0008] Based on a preset weighting function, trajectory planning is performed according to the target cleaning location to obtain the target trajectory; the preset weighting function is associated with a first distance and an arm angle, where the first distance is the distance between the target cleaning location and the end effector.

[0009] Perform a cleanup task on the target cleanup location according to the target trajectory;

[0010] During the task execution process, target data is collected and uploaded; the target data includes at least one of the following: the target cleanup location, the target trajectory.

[0011] In one possible design, the weight value calculated by the preset weight function is negatively correlated with the first distance, and the weight value is positively correlated with the arm angle.

[0012] In one possible design, the step of obtaining the target trajectory based on the target cleaning location using a preset weight function includes:

[0013] Based on the inverse kinematics algorithm, multiple candidate solutions are determined according to the target cleaning location;

[0014] For each candidate solution, the weight value of the candidate solution is determined according to the preset weight function;

[0015] The optimal solution is determined based on the weight values ​​corresponding to the multiple candidate solutions.

[0016] The target trajectory is determined based on the optimal solution.

[0017] In one possible design, determining the optimal solution based on the weight values ​​corresponding to the multiple sets of candidate solutions includes:

[0018] When the projection of the robot's hand onto the plane of the chest cavity is within the chest cavity region, the optimal solution is the set of candidate solutions with the largest weight value.

[0019] When the projection of the robot's hand onto the plane containing the chest cavity is outside the chest cavity region, the optimal solution is the set of candidate solutions with the smallest weight value.

[0020] In one possible design, the expression for the preset weighting function is:

[0021]

[0022] Where W is the weight value of the preset weight function, a is the preset weight coefficient of the distance term, d is the distance between the target cleaning position and the end effector, and α is the arm angle. For adaptive weighting coefficients, It is a function of d.

[0023] In one possible design, the preset weighting function includes the product of the arm angle and the corresponding adaptive weighting coefficient, wherein the adaptive weighting coefficient has a natural exponential relationship with the first distance.

[0024] In one possible design, the expression for the adaptive weighting coefficients is:

[0025]

[0026] in, For adaptive weighting coefficients, Based on the weighting coefficient, It is the attenuation coefficient.

[0027] In one possible design, determining the target cleaning location on the object to be cleaned includes:

[0028] Get the first image of the object to be cleaned;

[0029] Based on the first image, determine the first coordinate of the target cleaning location in the first coordinate system of the object to be cleaned;

[0030] Based on the first and second transformation relationships, the second coordinates of the target cleaning position in the robotic arm coordinate system are determined according to the first coordinates; the first transformation relationship is the transformation relationship between the humanoid robot's camera coordinate system and the robotic arm coordinate system; the second transformation relationship is the transformation relationship between the first coordinate system and the camera coordinate system.

[0031] The step of obtaining the target trajectory by trajectory planning based on the target cleaning location according to a preset weight function includes:

[0032] Based on a preset weight function, trajectory planning is performed according to the second coordinate to obtain the target trajectory.

[0033] In one possible design, before determining the target cleaning location on the object to be cleaned, the process further includes:

[0034] Get the second image of the object to be cleaned;

[0035] Determine the first cleanliness level of the object to be cleaned based on the second image;

[0036] If the first cleanliness level is less than the first preset level, then the target cleaning location on the object to be cleaned is determined;

[0037] After performing the cleanup task on the target cleanup location according to the target trajectory, the method further includes:

[0038] Obtain the third image of the object to be cleaned;

[0039] The third image is input into the target model to obtain the second cleanliness level;

[0040] If the second cleanliness level is greater than the second preset level, the cleaning task is determined to be completed; the second preset level is greater than or equal to the first preset level.

[0041] Secondly, embodiments of this application provide a data acquisition device for a humanoid robot training field, comprising:

[0042] The process control module is used to determine the target cleaning location on the object to be cleaned in response to the received task instruction.

[0043] The motion control module is used to perform trajectory planning based on the target cleaning position according to a preset weight function to obtain the target trajectory; the preset weight function is associated with a first distance and an arm angle, wherein the first distance is the distance between the target cleaning position and the end effector.

[0044] The driving module is used to perform a cleaning task on the target cleaning location according to the target trajectory.

[0045] The data acquisition module is used to collect and upload target data during task execution; the target data includes at least one of the following: the target cleaning location, the target trajectory.

[0046] Thirdly, embodiments of this application provide a data acquisition device for a humanoid robot training field, comprising: at least one processor and a memory;

[0047] The memory stores computer-executed instructions;

[0048] The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method described in the first aspect above and various possible designs of the first aspect.

[0049] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the method described in the first aspect and various possible designs of the first aspect.

[0050] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method described in the first aspect and various possible designs of the first aspect.

[0051] This embodiment provides a data acquisition method and device for a humanoid robot training field. The method includes, in response to receiving a task instruction, determining the target cleaning position on the object to be cleaned; performing trajectory planning based on the target cleaning position using a preset weight function to obtain a target trajectory; the preset weight function being associated with a first distance and an arm angle; the first distance being the distance between the target cleaning position and the end effector; performing a cleaning task on the target cleaning position according to the target trajectory; and collecting and uploading target data during task execution. The target data includes at least one of the following: target cleaning position and target trajectory. The method provided in this embodiment enables a humanoid robot to control itself to perform tasks such as tube sheet cleaning and collects and uploads data during the task, thereby avoiding excessive human intervention, improving collection efficiency and data accuracy, not being limited to a single environment, and further enhancing the richness of the data by perturbing the arm angle, thus improving the generalization ability of the humanoid robot. Attached Figure Description

[0052] 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.

[0053] Figure 1 A schematic diagram illustrating the principle of a data acquisition method for a humanoid robot training field provided in an embodiment of this application;

[0054] Figure 2 A flowchart illustrating the data acquisition method for a humanoid robot training field provided in this application embodiment. Figure 1 ;

[0055] Figure 3 A schematic diagram illustrating the principle of coordinate transformation provided in the embodiments of this application;

[0056] Figure 4 A comparative diagram showing the arm angle optimization strategy provided in this application before and after its application;

[0057] Figure 5 A flowchart illustrating the data acquisition method for a humanoid robot training field provided in this application embodiment. Figure 2 ;

[0058] Figure 6 This is a schematic diagram of an application scenario with added perturbation provided in an embodiment of this application;

[0059] Figure 7 This is a schematic diagram of the structure of a data acquisition device for a humanoid robot training field provided in an embodiment of this application;

[0060] Figure 8 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application.

[0061] 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

[0062] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] It should be noted that the data acquisition method and equipment for humanoid robot training fields provided in this application can be used in the field of humanoid robot technology, or in any field other than the field of humanoid robot technology. The application field of the data acquisition method and equipment for humanoid robot training fields provided in this application is not limited.

[0064] With the continuous advancement of artificial intelligence and robotics, humanoid robots are increasingly widely used in industrial fields, especially in cleaning tasks requiring high precision, flexibility, and adaptability. For example, pipe sheet cleaning operations, due to their complexity and variability, require humanoid robots to possess a high degree of coordination and precise movement capabilities. To train humanoid robots in specific tasks such as pipe sheet cleaning (e.g., grasping, sweeping, tool adjustment), it is essential to collect a large amount of data during the robot's task execution. This data includes the robot's joint poses, joint torques, joint accelerations, visual information, and sensor data, which can then be used to optimize the humanoid robot's behavior through machine learning and deep learning algorithms.

[0065] In related technologies, motion capture systems, virtual reality (VR) devices, and teleoperation are commonly used for data acquisition.

[0066] However, most of the aforementioned data collection methods rely on manual demonstration or control, requiring frequent intervention from operators. This results in a high manpower investment in the data collection process, making it difficult to guarantee an efficient workflow and data quality. Furthermore, manual intervention introduces errors, affecting data accuracy, especially during precise operations where it's difficult to capture subtle differences in the humanoid robot's movements. Additionally, these methods are typically conducted under limited environmental conditions, leading to insufficient data diversity and impacting the humanoid robot's generalization capabilities.

[0067] To address the aforementioned technical problems, the inventors of this application have discovered that a dedicated data acquisition system can be installed on a humanoid robot to control it to perform tasks such as tube sheet cleaning, and to collect and upload data during the task process. This avoids excessive human intervention, improves acquisition efficiency and data accuracy, is not limited to a single environment, and further enhances data richness by perturbing the arm angle, thereby improving the generalization ability of the humanoid robot. Based on this, this application provides a data acquisition method for a humanoid robot training field.

[0068] Figure 1 This is a schematic diagram illustrating the principle of a data acquisition method for a humanoid robot training field provided in an embodiment of this application. Figure 1As shown. The data acquisition system includes a process control module, a perception system module, a motion control algorithm module, a drive module, and a data module. The process control module is used for process planning of the tube sheet cleaning task; specifically, it receives task start commands, performs module scheduling, and handles anomalies. It is the top layer of the entire system. The perception system module is used to acquire visual sensor data and force control sensor data. The motion control algorithm module is used for path planning (e.g., path planning for a humanoid robot to move to the work position using its lower limbs), trajectory planning (e.g., trajectory planning for a humanoid robot to move to the work position and perform specific cleaning tasks using its upper limbs), and admittance control. The drive module is used to drive the joints of the humanoid robot and acquire motor data based on bus control. The data module is used to process the data collected during task execution, such as data filtering, data storage, and data uploading.

[0069] In the specific implementation process, the process control module receives the task start command, performs task planning, and generates the work position. The motion control module receives the work position sent by the process control module, performs path planning based on the work position, and calls the drive control module to control the robot to reach the work position. The perception module processes the images acquired by the vision sensor to obtain positioning information and transmits the positioning information to the process control module. The process control module calls the motion control module to perform trajectory planning based on the positioning information and a preset weight function, and executes the cleaning task. The data acquisition module performs data acquisition, filtering, and uploading during task execution. The data acquisition method for humanoid robot training fields provided in this application embodiment enables the humanoid robot to control itself to perform tasks such as tube sheet cleaning, and to collect and upload data during the task process, thereby avoiding excessive human intervention, improving acquisition efficiency and data accuracy, not being limited to a single environment, and also improving the richness of data by perturbing the arm angle, thus improving the generalization ability of the humanoid robot.

[0070] Optionally, after the cleaning operation is completed, the process control module calls the motion control module to perform path planning to exit the work area and return to the waiting area.

[0071] It should be noted that, Figure 1 The schematic diagram shown is merely an example. The data acquisition method and scenario for humanoid robot training field described in this application are for the purpose of more clearly illustrating the technical solutions of this application, and do not constitute a limitation on the technical solutions provided in this application. As those skilled in the art will know, with the evolution of the system and the emergence of new business scenarios, the technical solutions provided in this application are also applicable to similar technical problems.

[0072] The technical solutions of this application will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0073] Figure 2 A flowchart illustrating the data acquisition method for a humanoid robot training field provided in this application embodiment. Figure 1 .like Figure 2 As shown, the method includes:

[0074] 201. In response to receiving the task instruction, determine the target cleaning location on the object to be cleaned.

[0075] The executing entity in this embodiment is a humanoid robot.

[0076] In this embodiment, task instructions can be input by the user through a user interface or automatically generated by the system. Task instructions can be transmitted to the humanoid robot via wired or wireless communication. Task instructions may include specific cleaning task parameters, such as the type of cleaning, priority, and time limit. The object to be cleaned can be a tube sheet.

[0077] Specifically, after receiving the task instruction, the target cleaning location on the object to be cleaned (e.g., a hole on the pipe plate to be cleaned) can be determined by the humanoid robot's sensors (e.g., image sensors).

[0078] In some embodiments, determining the target cleaning position on the object to be cleaned may include: acquiring a first image of the object to be cleaned; determining a first coordinate of the target cleaning position in a first coordinate system of the object to be cleaned based on the first image; determining a second coordinate of the target cleaning position in a robotic arm coordinate system based on the first coordinates according to a second transformation relationship and a first transformation relationship; the first transformation relationship is a transformation relationship between the camera coordinate system of the humanoid robot and the robotic arm coordinate system; the second transformation relationship is a transformation relationship between the first coordinate system and the camera coordinate system; and performing trajectory planning based on a preset weight function to obtain a target trajectory may include: performing trajectory planning based on the second coordinates according to a preset weight function to obtain a target trajectory. The target cleaning position determination method for a humanoid robot provided in this disclosure, by acquiring an image of the object to be cleaned and performing coordinate transformation, can accurately locate the target cleaning position between different coordinate systems. This multi-coordinate system transformation mechanism ensures that the robotic arm can accurately identify and reach the target position. By performing trajectory planning based on a preset weight function, the arm angle of the robotic arm can be perturbed, improving the diversity of the collected data.

[0079] In this embodiment, the first coordinate system is a coordinate system established based on the object to be cleaned, such as a coordinate system established based on the holes on the tube sheet.

[0080] For example, such as Figure 3 As shown, the humanoid robot's hand-eye calibration can be performed in advance, calibrating the robot's camera and arms in the same coordinate system. Then, the process control module schedules the process. When a task instruction is received, the vision system module can be invoked to accurately locate the holes to be cleaned, and the hole coordinate system (obj, i.e., the first coordinate system) is transformed into the robotic arm coordinate system (base) through pose transformation. The first transformation relationship between the camera coordinate system and the robotic arm coordinate system is shown. The second transformation relationship between the camera coordinate system and the first coordinate system obtained from the calibration. This was calculated using a visual algorithm. Since this is a fixed coordinate transformation, the crawling process involves aligning the coordinate system of the tip with that of the obj, thus leading to the following derivation formula:

[0081] (1)

[0082] in, This represents the first transformation relationship between the camera coordinate system and the robot arm coordinate system. This represents the second transformation relationship between the camera coordinate system and the first coordinate system. This represents the transformation relationship between the coordinate system of the end effector and the robot arm; it is a fixed transformation relationship. This represents the transformation relationship between the cleaning tool coordinate system and the hole coordinate system.

[0083] It can be obtained based on formula (1): The calculated end-effector pose of the humanoid robot is then determined. Afterwards, the process control module can call the motion control algorithm module to control the humanoid robot to perform trajectory planning and admittance control, and call the data acquisition module to collect information such as joint angles, torque, and images of the humanoid robot. At this time, the humanoid robot is already performing its work.

[0084] 202. Based on a preset weight function, trajectory planning is performed according to the target cleaning position to obtain the target trajectory; the preset weight function is related to a first distance and the arm angle of the robotic arm, and the first distance is the distance between the target cleaning position and the end effector.

[0085] In this embodiment, the preset weight function includes a distance term and an arm angle index term; the distance term includes the product of the reciprocal of a first distance and the corresponding preset weight coefficient, and the arm angle index term includes the product of the arm angle of the robotic arm and the corresponding adaptive weight coefficient; the first distance is the distance between the target cleaning position and the end effector of the robotic arm; the adaptive weight coefficient is related to the first distance.

[0086] Specifically, in solving the inverse kinematics of multi-DOF (e.g., 7-DOF) humanoid robots, solutions with joint constraints are typically selected after eliminating those with joint limitations. However, this approach can lead to suboptimal robot motion paths, especially in scenarios requiring high flexibility, and also hinders the diversity of data acquisition. Therefore, this embodiment employs a weighted selection strategy based on end-effector position and arm angle to construct a more reasonable inverse kinematics algorithm. This involves integrating a preset weight function into the inverse kinematics algorithm, comprehensively considering the distance between the target cleaning position and the end effector, as well as the arm angle, to prioritize solutions that meet specific task requirements. Specifically, when the robot's end effector approaches the target position, the arm angle receives greater weight, allowing for more consideration of the arm angle and enabling the robot to complete the task with a more reasonable posture while maintaining high-precision positioning. This optimization strategy is particularly suitable for applications requiring flexible posture adjustments, such as tube sheet cleaning. Furthermore, it allows for greater diversity in the collected data.

[0087] In some embodiments, the weight value calculated by the preset weighting function is negatively correlated with the first distance and positively correlated with the arm angle. By making the weight value negatively correlated with the first distance and positively correlated with the arm angle, this application embodiment can prioritize candidate solutions during trajectory planning that bring the robotic arm's end effector closer to the target position (smaller first distance) and result in a better robotic arm posture and more distributed joint load (larger arm angle). This trade-off mechanism effectively avoids the robotic arm falling into a suboptimal configuration with extreme joint angles and approaching a kinematic singularity in blindly pursuing the shortest movement distance. Therefore, while ensuring cleaning efficiency, it significantly improves the stability, safety, and lifespan of the robotic arm during the movement process.

[0088] In some embodiments, trajectory planning based on the target cleaning position using a preset weight function to obtain the target trajectory may include: determining multiple candidate solutions based on the target cleaning position using a preset inverse kinematics algorithm; determining the weight value of each candidate solution according to the preset weight function; determining an optimal solution based on the weight values ​​corresponding to the multiple candidate solutions; and determining the target trajectory based on the optimal solution. Each joint corresponds to an optimal solution, and then the optimal solutions corresponding to each joint are combined in an orderly manner according to the joint sequence to form the target trajectory. In this embodiment, by obtaining multiple candidate solutions based on inverse kinematics and selecting the best one using a preset weight function, the quality and intelligence level of trajectory planning can be significantly improved.

[0089] The trajectory planning method for humanoid robots provided in this disclosure determines multiple candidate solutions based on the target cleaning location and calculates the weight value of a preset weight function for each candidate solution. By combining distance and arm angle indices, the solution that best meets the task requirements can be selected from multiple candidate solutions, thereby determining the optimal trajectory. This not only ensures that the robot's end effector can approach the target location with high precision but also allows for adjustments to the robotic arm's posture in scenarios requiring high flexibility. This optimization strategy is particularly suitable for complex task environments, such as pipe cleaning, and can improve the robot's adaptability and task execution efficiency in diverse environments. By introducing diverse solution selections into trajectory planning, the collected data becomes more diverse, thereby enhancing the generalization ability and learning effect of the humanoid robot.

[0090] In some embodiments, the candidate solution with the highest weight value is selected as the optimal solution. This allows the robot to plan its path more effectively in complex task environments, achieve more comprehensive coverage of the cleaning area of ​​the object to be cleaned, and thus make the collected data more diverse.

[0091] In some embodiments, when the projection of the robot's hand onto the plane containing the chest cavity is within the chest cavity region, i.e. within the chest cavity interference zone, the above-mentioned optimal solution is the set of candidate solutions with the largest weight value.

[0092] When the projection of the robot's hand onto the plane containing the chest cavity is outside the chest cavity region, i.e. outside the chest cavity interference zone, the above optimal solution is the set of candidate solutions with the smallest weight value.

[0093] This approach avoids interference between the arm and the robot body or chest cavity during the cleaning process, ensuring a large range of motion and thus a wider cleaning area. This results in more comprehensive coverage of the area to be cleaned, leading to more diverse data collection.

[0094] In some embodiments, the expression of the preset weighting function is:

[0095] (2)

[0096] Where W is the weight value of the weighting function, a is the preset weight coefficient of the distance term, d is the distance between the target cleaning position and the end effector, and α is the arm angle. Let be the adaptive weighting coefficient, and be a function of d.

[0097] The trajectory planning method for humanoid robots provided in this disclosure optimizes the robot's path selection during task execution by combining a weighting function with a distance term and an arm angle index term. The distance term prioritizes precise positioning when the end effector approaches the target location, thereby improving task completion accuracy. Simultaneously, the arm angle index term dynamically adjusts the influence of the arm angle through adaptive weighting coefficients, enabling the robot to flexibly adjust its posture at different distances. This not only improves the robot's performance in high-precision tasks but also enhances its adaptability in complex environments. Through this optimization strategy, the system can achieve efficient trajectory planning in various task scenarios, improving the diversity and quality of data acquisition, thereby enhancing the generalization ability and intelligence level of the humanoid robot.

[0098] In some embodiments, the adaptive weighting coefficient and the first distance have a natural exponential relationship. In this embodiment, by setting the adaptive weighting coefficient and the first distance to a natural exponential relationship, more dynamic and precise control can be achieved in trajectory planning. This allows the weighting coefficient to be non-linearly adjusted based on the distance between the end effector and the target cleaning position, thereby optimizing the robot's motion strategy at different distances. When the distance is large, the adaptive weighting coefficient can reduce the influence of the arm angle, prioritizing rapid approach to the target position; while when the distance decreases, the weighting coefficient enhances the influence of the arm angle to ensure posture adjustment and accuracy when approaching the target. This natural exponential relationship provides a smooth weight adjustment mechanism, avoiding abrupt weight changes and improving the stability and continuity of trajectory planning. Through this method, the robot can better balance speed and accuracy when performing complex tasks, improving overall task execution efficiency and data acquisition diversity, thereby enhancing the adaptability and intelligence level of the humanoid robot.

[0099] In some embodiments, the expression for the adaptive weighting coefficient is:

[0100] (3)

[0101] in, For adaptive weighting coefficients, Based on the weighting coefficient, It is the attenuation coefficient.

[0102] In this embodiment, the expression for the adaptive weighting coefficients incorporates a natural exponential decay relationship to dynamically adjust the influence of the arm angle in the robot's trajectory planning. The base weighting coefficients set the initial influence, while the decay coefficients control the rate at which this influence decays with distance. When the end effector is far from the target position, exponential decay weakens the influence of the arm angle, prioritizing path planning that quickly approaches the target position, thus improving task execution efficiency. As the actuator gradually approaches the target position, the influence of the weighting coefficients gradually increases, ensuring fine-grained posture adjustments and precise positioning upon approaching the target. This mechanism provides a smooth and continuous weighting adjustment method, avoiding abrupt weight changes and improving the stability and continuity of trajectory planning. Through this method, the system can achieve more flexible and efficient trajectory planning in different task scenarios, improving the diversity and quality of data acquisition, thereby enhancing the adaptability and intelligence level of the humanoid robot.

[0103] Specifically, the first step is to determine the initial distance between the target cleaning location and the humanoid robot's end effector: Let the current position of the robot's end effector be... ,in, , and , where is the position coordinate of the end effector in the robot arm coordinate system. The target cleaning location is . , , and Let be the position coordinates of the target cleaning location (e.g., the hole to be cleaned) in the robot arm coordinate system. Then the first distance d is defined as:

[0104] Arm shape angles reflect the distribution of joint angles in a robot and can be quantified by the variance or dispersion of joint angles. For example, the arm shape angles of a 7-DOF robotic arm are determined by a set of joint angles. ,in, For the angle of the first joint, For the angle of the second joint, The angle of the seventh joint, arm angle It can be defined as: , where, Let i be the angle of the i-th joint. is a preset value, representing the maximum value of the angle between the plane formed by the robot arm and the Z-axis normal plane.

[0105] To achieve the characteristic that the closer the end effector is to the target cleaning location, the greater the arm angle weight, this embodiment employs an adaptive weighting coefficient. It is the end distance. Functions:

[0106] (3)

[0107] in, These are the basic weighting coefficients. It is the attenuation coefficient, which controls the rate at which the weight changes with distance. When When decreasing, The increase makes the effect of the arm angle more significant when approaching the target.

[0108] Preset weight function Combined with the first distance The reciprocal of the distance (the closer the distance, the greater the weight) and the arm angle Weighted values:

[0109] (2)

[0110] in, This is the weighting coefficient for the distance term. By adjusting... , and The influence of distance and arm angle can be balanced according to specific task requirements.

[0111] Specifically, parameter adjustment methods can include: increasing This will make the algorithm focus more on the end-point distance, making it suitable for tasks requiring high positioning accuracy. Increase This will enhance the effect of the arm angle, making it suitable for tasks requiring a larger operating space. Increasing... This will cause the arm angle weight to increase more rapidly as the distance decreases, making it suitable for close-range operations where high posture adjustment is required.

[0112] like Figure 4 As shown, by adding a preset weight function for trajectory planning, the arm angle can be adjusted to improve data diversity and adapt to specific task requirements.

[0113] 203. Perform a cleanup task on the target cleanup location according to the target trajectory.

[0114] Specifically, during the cleaning task at the target location based on the target trajectory, the humanoid robot moves its end effector along a precisely calculated path to reach and act on the designated cleaning position. By utilizing a motion control module, the humanoid robot can coordinate the movement of its joints, ensuring it maintains appropriate posture and force upon reaching the target location, thereby effectively performing the cleaning operation. This process relies not only on precise trajectory planning but also on real-time sensor feedback to monitor and adjust the cleaning actions, ensuring efficient and accurate task completion.

[0115] 204. Collect and upload target data during task execution; the target data includes at least one of the following: the target cleanup location, the target trajectory.

[0116] Specifically, during task execution, target data is collected and uploaded, with a focus on key information such as the target clearing location and trajectory. High-precision sensors and data loggers can capture the robot's actual position and movement path in real time. This data can be uploaded to a central system or the cloud for real-time monitoring and subsequent analysis. Real-time data uploads provide immediate feedback, helping operators quickly identify and resolve potential problems. Simultaneously, the stored data provides a foundation for training machine learning models, enabling the robot to improve its decision-making capabilities and execution efficiency in future tasks. Through this data-driven approach, the system can continuously optimize robot performance, enhancing its adaptability and intelligence in complex environments.

[0117] The data acquisition method for humanoid robot training grounds provided in this embodiment enables the humanoid robot to control itself to perform tasks such as tube sheet cleaning, and to collect and upload data during the task process. This avoids excessive human intervention, improves acquisition efficiency and data accuracy, is not limited to a single environment, and can also improve the richness of data by perturbing the arm angle, thereby improving the generalization ability of the humanoid robot.

[0118] In some embodiments, before determining the target cleaning location on the object to be cleaned, the method may further include: acquiring a second image of the object to be cleaned; determining a first cleanliness level of the object to be cleaned based on the second image; if the first cleanliness level is less than a first preset level, then determining the target cleaning location on the object to be cleaned; after performing a cleaning task on the target cleaning location according to the target trajectory, the method may further include: acquiring a third image of the object to be cleaned; inputting the third image into a target model to obtain a second cleanliness level; if the second cleanliness level is greater than a second preset level, then determining that the cleaning task is completed; the second preset level is greater than or equal to the first preset level. The data acquisition method for a humanoid robot training field provided in this disclosure, by acquiring images of the object to be cleaned before and after the cleaning task is executed and performing cleanliness assessments, can effectively improve the efficiency and quality of the cleaning task. Before the task begins, determining the first cleanliness level of the object by analyzing the second image allows cleaning to be performed only when the cleanliness is lower than a preset standard, thereby avoiding unnecessary operations and saving resources. After cleaning is completed, the second cleanliness level is assessed using the third image and the target model to ensure that the cleaning effect meets or exceeds the expected standard. This approach not only optimizes the cleanup process, reducing over- or under-cleaning, but also lowers the need for human intervention through an automated evaluation mechanism, enhancing the system's autonomy and intelligence. Overall, the method provided in this embodiment improves resource utilization and enhances the reliability and effectiveness of task execution.

[0119] Figure 5 A flowchart illustrating the data acquisition method for a humanoid robot training field provided in this application embodiment. Figure 2 .like Figure 5 As shown, in this embodiment, perturbations are added to the work location and waiting location to improve data diversity. Specifically, the method includes:

[0120] 501. In response to receiving a task instruction, randomly determine a target work location from multiple work locations in the work area of ​​the object to be cleaned; perform path planning based on the target work location to obtain a first target path; and move to the target work location based on the first target path.

[0121] 502. Determine the target cleaning location on the object to be cleaned.

[0122] 503. Based on the preset weight function, trajectory planning is performed according to the target cleaning position to obtain the target trajectory; the preset weight function is related to the first distance and the arm angle, and the first distance is the distance between the target cleaning position and the end effector.

[0123] 504. Perform a cleanup task on the target cleanup location according to the target trajectory.

[0124] 505. During the task execution process, target data is collected and uploaded; the target data includes at least one of the following: the target cleanup location, the target trajectory.

[0125] Steps 502 to 505 in this embodiment are similar to steps 201 to 204 in the above embodiment, and will not be described again here.

[0126] 506. After the task is completed, a target waiting position is randomly determined from multiple waiting positions in the task waiting area; a path is planned based on the target waiting position to obtain a second target path; and the user moves to the target waiting position according to the second target path.

[0127] Specifically, such as Figure 6 As shown, to avoid data duplication due to overly uniform environments during robot operation, a perturbation module can be set up to randomly assign a work position in the robot's work area. When the robot receives a start command from the work completion waiting area, the process control module will call the perturbation function to randomly assign a work position from the work area, allowing the robot to move to different work positions, increasing the randomness of the task at the task level. Then, the robot moves to the work position in the work area to perform the task. After the robot completes the task, the perturbation function module will randomly assign a waiting position in the work waiting area, and the robot will automatically move to the waiting position after completion, ensuring that the robot can return to different positions in the waiting area after completing the task.

[0128] The data acquisition method for humanoid robot training fields provided in this embodiment significantly improves the diversity of acquired data and the flexibility of task execution by introducing perturbations at the work and waiting positions. Specifically, after receiving a task instruction, the target work position is randomly determined from multiple work positions, and a waiting position is randomly assigned after the work is completed. This randomness not only avoids data duplication caused by an overly uniform environment but also increases the randomness of the task, allowing the robot to perform tasks in different positions and enriching the data sample. Trajectory planning based on a preset weight function ensures the flexibility of robot movement and scene adaptability. During task execution, the system collects and uploads target data, providing a foundation for subsequent analysis and optimization. Overall, this method improves the robot's adaptability and intelligence level, and enhances the efficiency and diversity of data acquisition by introducing random perturbations and a preset weight function.

[0129] Figure 7 This is a schematic diagram of the structure of a data acquisition device for a humanoid robot training field provided in an embodiment of this application. Figure 7As shown, the data acquisition device 70 for humanoid robot training field includes: a process control module 701, a motion control module 702, a drive module 703, and a data acquisition module 704.

[0130] The process control module 701 is used to determine the target cleaning location on the object to be cleaned in response to receiving a task instruction.

[0131] The motion control module 702 is used to perform trajectory planning based on the target cleaning position according to a preset weight function to obtain the target trajectory; the preset weight function is associated with a first distance and an arm angle, wherein the first distance is the distance between the target cleaning position and the end effector.

[0132] The drive module 703 is used to perform a cleaning task on the target cleaning location according to the target trajectory.

[0133] The data acquisition module 704 is used to collect and upload target data during task execution; the target data includes at least one of the following: the target cleaning location, the target trajectory.

[0134] The data acquisition device for a humanoid robot training field provided in this application embodiment enables the humanoid robot to control itself to perform tasks such as tube sheet cleaning, and to collect and upload data during the task process. This avoids excessive human intervention, improves acquisition efficiency and data accuracy, is not limited to a single environment, and can also improve the richness of data by perturbing the arm angle, thereby improving the generalization ability of the humanoid robot.

[0135] In some embodiments, the motion control module 702 is specifically used to: determine multiple candidate solutions based on the target cleaning position using an inverse kinematics algorithm; determine the weight value of each candidate solution according to the preset weight function; determine the optimal solution based on the weight values ​​corresponding to the multiple candidate solutions; and determine the target trajectory based on the optimal solution.

[0136] In some embodiments, the expression of the preset weighting function is:

[0137]

[0138] Where W is the weight value of the weighting function, a is the weight coefficient of the distance term, d is the distance between the target cleaning position and the end effector, and α is the arm angle. Let be the adaptive weighting coefficient, and be a function of d.

[0139] In some embodiments, the adaptive weighting coefficients and the first distance have a natural exponential relationship.

[0140] In some embodiments, the expression for the adaptive weighting coefficient is:

[0141]

[0142] in, For adaptive weighting coefficients, Based on the weighting coefficient, It is the attenuation coefficient.

[0143] In some embodiments, the process control module 701 is further configured to: randomly determine a target job location from multiple job locations in the job area of ​​the object to be cleaned; perform path planning based on the target job location to obtain a first target path; and move to the target job location based on the first target path.

[0144] In some embodiments, the process control module 701 is further configured to: after a task is completed, randomly determine a target waiting position from multiple waiting positions in the job waiting area; perform path planning based on the target waiting position to obtain a second target path; and move to the target waiting position based on the second target path.

[0145] In some embodiments, the process control module 701 is specifically used to: acquire a first image of the object to be cleaned; determine the first coordinates of the target cleaning position in the first coordinate system of the object to be cleaned based on the first image; determine the second coordinates of the target cleaning position in the coordinate system of the robotic arm based on the first coordinates according to the first transformation relationship and the second transformation relationship; the first transformation relationship is the transformation relationship between the camera coordinate system of the humanoid robot and the coordinate system of the robotic arm; the second transformation relationship is the transformation relationship between the first coordinate system and the camera coordinate system; the motion control module 702 is specifically used to: perform trajectory planning based on the second coordinates according to a preset weight function to obtain the target trajectory.

[0146] In some embodiments, the process control module 701 is further configured to: acquire a second image of the object to be cleaned; determine a first cleanliness level of the object to be cleaned based on the second image; if the first cleanliness level is less than a first preset level, determine a target cleaning location on the object to be cleaned; acquire a third image of the object to be cleaned; input the third image into a target model to obtain a second cleanliness level; if the second cleanliness level is greater than a second preset level, determine that the cleaning task is completed; the second preset level is greater than or equal to the first preset level.

[0147] The data acquisition device for humanoid robot training field provided in this application embodiment can be used to execute the above method embodiment. Its implementation principle and technical effect are similar, and will not be described again here.

[0148] Figure 8 A schematic diagram of the structure of the electronic device provided in this application. Figure 8 As shown, the electronic device 80 provided in this embodiment includes at least one processor 801 and a memory 802. Optionally, the electronic device 80 further includes a communication component 803. The processor 801, memory 802, and communication component 803 are connected via a bus.

[0149] In a specific implementation, at least one processor 801 executes computer execution instructions stored in memory 802, causing at least one processor 801 to perform the above-described method.

[0150] The specific implementation process of processor 801 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0151] 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.

[0152] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0153] 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.

[0154] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0155] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0156] 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.

[0157] 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.

[0158] 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.

[0159] 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.

[0160] 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.

[0161] 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.

[0162] 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.

[0163] 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 collection method for a humanoid robot training field, characterized by, Comprising: in response to receiving a task instruction, determining a target cleaning position on the object to be cleaned; based on a preset weight function, performing trajectory planning according to the target cleaning position to obtain a target trajectory; the preset weight function is associated with a first distance and an arm type angle, and the first distance is the distance between the target cleaning position and the end effector; performing a cleaning task on the target cleaning position according to the target trajectory; collecting and uploading target data during task execution; the target data includes at least one of the target cleaning position and the target trajectory; wherein the expression of the preset weight function is: ; wherein ; Wherein, W is the weight value of the preset weight function, a is the weight coefficient of the preset distance term, d is the distance between the target cleaning position and the end effector, and a is the arm angle, is a function of d, the is an adaptive weight coefficient, is a basic weight coefficient, is a decay coefficient.

2. The method of claim 1, wherein, the weight value calculated by the preset weight function is negatively correlated with the first distance, and the weight value is positively correlated with the arm type angle.

3. The method of claim 1, wherein, based on a preset weight function, performing trajectory planning according to the target cleaning position to obtain a target trajectory, comprising: based on an inverse kinematics solving algorithm, determining a plurality of candidate solutions according to the target cleaning position; for each candidate solution, determining a weight value of the candidate solution according to the preset weight function; determining an optimal solution according to the weight values corresponding to the plurality of candidate solutions respectively; determining a target trajectory according to the optimal solution.

4. The method of claim 3, wherein, determining an optimal solution according to the weight values corresponding to the plurality of candidate solutions respectively, comprising: when the projection of the hand of the robot on the plane where the chest cavity is located is located within the chest cavity region, the optimal solution is the candidate solution with the maximum weight value; when the projection of the hand of the robot on the plane where the chest cavity is located is located outside the chest cavity region, the optimal solution is the candidate solution with the minimum weight value.

5. The method according to any one of claims 1 to 4, characterized in that, determining a target cleaning position on the object to be cleaned, comprising: obtaining a first image of the object to be cleaned; determining a first coordinate of the target cleaning position in a first coordinate system of the object to be cleaned according to the first image; based on a first conversion relationship and a second conversion relationship, determining a second coordinate of the target cleaning position in a robot arm coordinate system according to the first coordinate; the first conversion relationship is the conversion relationship between the camera coordinate system of the humanoid robot and the robot arm coordinate system; the second conversion relationship is the conversion relationship between the first coordinate system and the camera coordinate system; based on a preset weight function, performing trajectory planning according to the target cleaning position to obtain a target trajectory, comprising: based on a preset weight function, performing trajectory planning according to the second coordinate to obtain a target trajectory.

6. The method according to any one of claims 1 to 4, characterized in that, before determining a target cleaning position on the object to be cleaned, further comprising: obtaining a second image of the object to be cleaned; determining a first cleanliness level of the object to be cleaned according to the second image; if the first cleanliness level is less than a first preset level, determining a target cleaning position on the object to be cleaned; after performing a cleaning task on the target cleaning position according to the target trajectory, further comprising: obtaining a third image of the object to be cleaned; inputting the third image into a target model to obtain a second cleanliness level; if the second cleanliness level is greater than a second preset level, determining that the cleaning task is completed; the second preset level is greater than or equal to the first preset level.

7. A data acquisition device for a humanoid robot training arena, characterized by, Comprising: at least one processor and a memory; the memory stores computer-executed instructions; the at least one processor executes the computer-executed instructions stored in the memory, so that the at least one processor executes the data collection method for the humanoid robot training field according to any one of claims 1 to 6.

Images