Robot dexterous hand grasping control method and device, robot, medium and product

By acquiring measurement data of the objects to be operated on the power grid, determining the object type and task type, verifying candidate grasping postures, using image and depth data to determine the position and no-collision areas, and combining tactile and pressure sensors for real-time adjustments, the problem that general robot dexterity hands cannot adapt to objects to be operated on the power grid is solved, and flexible grasping control is achieved.

CN123060698APending Publication Date: 2026-11-10SHENZHEN POWER SUPPLY BUREAU
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611429740.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-15
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

The general robot dexterous hand grasping control method cannot be applied to the objects operated on in the power grid, resulting in inflexible grasping control and an inability to adapt to the differences in different types of objects operated on the power grid.

Method used

By acquiring measurement data of the object to be operated, the object type and task type are determined. Based on the measurement data, candidate gripping postures are verified, the target gripping posture is determined, and gripping control is performed. Image data and depth data are used to determine position information and no-touch areas, and adjustments are made in real time in conjunction with tactile sensors and pressure sensors.

Benefits of technology

It enables the determination of the target gripping posture based on the characteristics of the object to be operated, adapting to different types of objects in the power grid, making gripping control more flexible, and improving the accuracy and stability of gripping.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123060698A_ABST
    Figure CN123060698A_ABST
Patent Text Reader

Abstract

The application relates to a robot dexterous hand grasping control method and device, a robot, a medium and a product. The method comprises the following steps: firstly, obtaining measurement data of an object to be operated, and determining an object type of the object to be operated according to the measurement data; then, determining a candidate grasping posture according to the object type and a task type of the object to be operated; subsequently, verifying the candidate grasping posture based on the measurement data to determine a target grasping posture; and finally, performing grasping control on the object to be operated based on the target grasping posture. The method can determine the target grasping posture according to the characteristics of the object to be operated, instead of performing grasping control through a unified grasping posture, can adapt to different types of operation objects in a power grid, and grasping control is more flexible.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of robot dexterous hand control technology, and in particular to a robot dexterous hand grasping control method, device, robot, medium and product. Background Technology

[0002] In the fields of power production, transmission and distribution, multi-pointed robot dexterous hands can mimic the dexterous operations of human hands such as grasping, rotating and pressing. In high-voltage, narrow or high-altitude work areas, they can replace maintenance personnel in operating objects, reducing the risk of personnel injury or death.

[0003] However, general-purpose robot dexterous hand grasping control methods typically aim to generate grasping postures for arbitrary objects, making them suitable for warehousing or general service scenarios. In contrast, the objects manipulated in power grids exhibit significant differences, requiring different grasping control methods. Therefore, general-purpose robot dexterous hand grasping control methods are not applicable to objects manipulated in power grids. Summary of the Invention

[0004] Therefore, it is necessary to provide a robot dexterous hand grasping control method, device, robot dexterous hand, medium and program product that can be applied to objects operated in the power grid, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a robot dexterous hand grasping control method, the method comprising:

[0006] Obtain the measurement data of the object to be operated on, and determine the object type of the object to be operated on based on the measurement data;

[0007] Based on the object type and the task type of the object to be operated on, determine the candidate gripping postures;

[0008] The candidate gripping posture is verified based on the measurement data to determine the target gripping posture;

[0009] Based on the target's grasping posture, grasping control is performed on the object to be manipulated.

[0010] In one embodiment, the measurement data includes image data and depth data, and the candidate gripping poses include a primary candidate gripping pose and alternative candidate gripping poses. The candidate gripping poses are validated based on the measurement data to determine the target gripping pose, including:

[0011] Based on image and depth data, determine the location information of the object to be operated and the no-touch zone;

[0012] Based on location information and no-touch zones, the primary candidate gripping posture is safely verified.

[0013] If the security check passes, a preliminary gripping test is performed based on the primary candidate gripping posture. If the preliminary gripping test passes, the primary candidate gripping posture is used as the target gripping posture.

[0014] If the initial gripping test fails or the safety verification fails, the alternative gripping posture will be used as the primary gripping posture, and the steps of performing safety verification on the primary gripping posture based on location information and no-touch zones will be repeated.

[0015] In one embodiment, a security check is performed on the primary candidate gripping posture based on location information and no-touch zones, including:

[0016] Transform the primary candidate grasping posture to the robot coordinate system and determine the target contact point;

[0017] If the target contact point does not fall into the no-touch zone, or if the distance between the target contact point and the adjacent object of the object to be operated is greater than the preset distance threshold, then the safety verification of the primary candidate gripping posture is determined to be passed.

[0018] In one embodiment, preliminary grip detection is performed based on the primary candidate grip posture, including:

[0019] Determine the preset contact area of ​​the object to be operated on;

[0020] Based on the pre-forming angle corresponding to the main candidate gripping posture, low-force contact with the object to be operated is achieved;

[0021] The actual contact point detected by the tactile sensor is obtained. If the actual contact point is located within the preset contact area, the initial grasp test is deemed qualified.

[0022] In one embodiment, grasping control of the object to be manipulated is performed based on the target grasping posture, including:

[0023] Determine the target gripping force range based on the target's gripping posture;

[0024] If the actual contact point is located within the preset contact area, increase the gripping force on the object to be operated to the target gripping force range;

[0025] Acquire the pressure distribution of the object to be operated as detected by the pressure sensor;

[0026] Adjust the gripping force of the object to be operated on based on pressure distribution.

[0027] In one embodiment, determining the object type of the object to be operated on based on measurement data includes:

[0028] Feature extraction is performed based on the measurement data to obtain attribute feature vectors;

[0029] Classification and detection are performed based on attribute feature vectors to determine the object type.

[0030] Secondly, this application also provides a robot dexterous hand grasping control device, the device comprising:

[0031] The acquisition module is used to acquire measurement data of the object to be operated on, and determine the object type of the object to be operated on based on the measurement data;

[0032] The first determination module is used to determine candidate gripping postures based on the object type and the task type of the object to be operated on.

[0033] The second determination module is used to verify the candidate gripping posture based on the measurement data and determine the target gripping posture.

[0034] The control module is used to control the grasping of the object to be operated based on the target grasping posture.

[0035] Thirdly, this application also provides a robot, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any one of the first aspects above.

[0036] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects above.

[0037] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any one of the first aspects above.

[0038] The aforementioned robot dexterous hand grasping control method, device, robot, medium, and product first acquire measurement data of the object to be manipulated, and then determine the object type based on the measurement data. Next, based on the object type and the task type of the object, candidate grasping postures are determined. Then, the candidate grasping postures are verified based on the measurement data to determine the target grasping posture. Finally, based on the target grasping posture, grasping control is performed on the object to be manipulated. In this way, by determining the object type based on the measurement data of the object to be manipulated, and then determining the target grasping posture based on the object type and task type to perform grasping control, the target grasping posture is determined according to the characteristics of the object to be manipulated, rather than using a uniform grasping posture for grasping control. This allows for adaptation to different types of objects in the power grid, making grasping control more flexible. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating a robot dexterous hand grasping control method in one embodiment;

[0041] Figure 2 This is a flowchart illustrating the target grasping posture determination steps in one embodiment;

[0042] Figure 3 This is a flowchart illustrating the security verification steps for grip posture in one embodiment;

[0043] Figure 4 This is a flowchart illustrating the preliminary detection steps for grip posture in one embodiment;

[0044] Figure 5 This is a flowchart illustrating a robot dexterous hand grasping control method in another embodiment;

[0045] Figure 6 This is a flowchart illustrating a robot dexterous hand grasping control method in another embodiment;

[0046] Figure 7 This is a schematic diagram of the gripping posture of a cap-screwing operation object in one embodiment;

[0047] Figure 8 This is a schematic diagram of the gripping posture of a rod-shaped probe-type manipulation object in one embodiment;

[0048] Figure 9 This is a schematic diagram of the gripping posture of a sheet-like pressure plate-type operation object in one embodiment;

[0049] Figure 10 This is a structural block diagram of a robot dexterous hand grasping control device in one embodiment;

[0050] Figure 11 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0052] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0053] In one exemplary embodiment, such as Figure 1 As shown, a robot dexterous hand grasping control method is provided. Taking the application of this method to a terminal as an example, the terminal can be a robot, which can be an industrial and collaborative robot, a humanoid and bionic robot, a special-purpose robot, a medical and rehabilitation robot, etc., including the following steps 101 to 104. Wherein:

[0054] Step 101: Obtain the measurement data of the object to be operated on, and determine the object type of the object to be operated on based on the measurement data.

[0055] The object to be operated is the object that the robot's dexterous hand needs to operate and control. This application uses objects operated in power grid operations as examples, such as caps, knobs, valves, terminal screws, pressure plates, buttons, wires, probes, tool handles, etc.

[0056] The measurement data of the object to be operated on can be image data and depth data of the object. The image data and depth data can be two-dimensional images, depth maps, or point cloud data acquired by vision sensors, depth cameras, or depth sensors. Information such as the length, width, diameter, curvature, principal axis direction, surface flatness, grippable area, and minimum distance from the no-touch zone of the object to be operated on can be extracted from the image data and depth data.

[0057] For example, by performing planar segmentation and Euclidean clustering on the point cloud data in the measurement data, the object to be operated on can be separated from the background. By calculating the minimum bounding box of the clustered point cloud, the length, width, and height can be directly read. For cylinders, a spatial cylinder model is fitted to determine the diameter and axial length.

[0058] The main axis direction can be calculated in different ways depending on the shape of the object to be operated on. For example, for a slender object to be operated on, it can be obtained by fitting the main axis of the contour or the main direction of the point cloud. For a near-circular object to be operated on, the rotation axis can be determined by the outer edge of the circle, the center point and the surface normal. For a rectangular object to be operated on, its operation direction can be determined by the edge line, the direction of the long side and the position of the rotation axis.

[0059] Curvature can be obtained by fitting a local plane to the point cloud of each point on the surface of the object to be operated on, and calculating the rate of change of the normal vector at that point. Surface smoothness is calculated by projecting all points on the surface of the object to be operated on onto the fitted plane and calculating the root mean square error of each point to that plane. The smaller the root mean square error value, the smoother the surface; the larger the root mean square error value, the more uneven the surface.

[0060] The grippable area can be determined by semantically segmenting the image data to mark the permissible contact pixel regions such as "insulation layer" or "handle". Then, the depth values ​​of these pixel regions are mapped to a 3D image including depth data, and the convex hull area of ​​the pixel regions is calculated to obtain the effective grippable area.

[0061] The minimum distance from the no-touch zone can be obtained from the distance between the object to be operated and surrounding obstacles identified by image data and depth data. Surrounding obstacles include non-target objects such as adjacent wires, terminal blocks, housing edges, and adjacent buttons around the object to be operated.

[0062] Measurement data can also include surface characteristic data, such as surface hardness and surface friction properties. For objects that can be touched, surface hardness can be measured using tactile sensors. For objects whose intimacy is uncertain, surface hardness can be assessed indirectly, such as through pneumatic jetting, where a pulsed airflow is injected into the object to be manipulated, and the hardness is assessed by analyzing its surface vibration response. For objects that can be touched, surface friction properties can be measured using tactile sensors. For objects whose intimacy is uncertain, surface friction properties can be obtained through image data analysis.

[0063] Optionally, feature extraction is performed based on the measurement data to obtain attribute feature vectors. Based on the aforementioned measurement data, attribute feature vectors are obtained, including the length, width, diameter, curvature, principal axis direction, surface flatness, grippable area, minimum distance from the no-collision zone, surface hardness, and surface friction characteristics of the object to be manipulated. The attribute feature vectors may also include the task type, which can be obtained based on job planning instructions, manual task templates, job ticket information, the robot's current job flow node, or control system input.

[0064] Object type is determined by classifying and detecting objects based on attribute feature vectors. Optionally, the classification and detection methods can include rule-based classification, machine learning classification, visual language model-assisted classification, or manual task template specification. For example, the attribute feature vectors are input into a pre-defined classification model, which outputs the object category and classification confidence score. For instance, a near-circular outer edge with a screw-on task can be classified as a screw cap / knob; a slender spindle requiring axial orientation can be classified as a rod-shaped probe / tool; and a sheet-like edge requiring movement in a fixed direction can be classified as a sheet-like pressure plate / paddle. The pre-defined classification model can be obtained by training an initial classification model using sample data. The sample data can include attribute feature vectors of different categories of objects to be operated on and pre-labeled object types. The initial classification model is trained using sample data until training stops (e.g., model convergence or the number of training iterations reaches a preset value), thus enabling the pre-defined classification model to stably and accurately classify objects based on their attribute feature vectors.

[0065] For example, the objects to be operated in power grid operations can be categorized into the following types based on common equipment types, as shown below:

[0066] 1. Screw caps / knobs: These have a cylindrical or near-circular outer edge and require rotation around a central axis.

[0067] 2. Valves / handwheels: These have rotatable handles or rims and require a large output torque.

[0068] 3. Terminal screws: These are small in size and require tools or fingertips to hold the axis and apply torque.

[0069] 4. Sheet-shaped pressure plates / paddles: These have edges and a rotation axis, and need to be flicked in a specific direction.

[0070] 5. Rod-shaped probes / tools: Slender objects that require stable clamping and axial orientation.

[0071] 6. Wires / Cables: Flexible objects that require low-force clamping and should be kept away from pulling.

[0072] 7. Button / Switch type: Localized click objects that require restrictions on travel distance and force.

[0073] 8. Fragile / Non-touchable items: Transparent oil cups, brittle shells, nearby electrical components, etc., which need to be restricted or prohibited from being gripped.

[0074] Step 102: Determine candidate gripping postures based on the object type and the task type of the object to be operated on.

[0075] Based on the determined object type and task type of the object to be operated, a preset power grid object classification library is queried. This library can be stored in the robot control system, host computer, or job controller. It can include typical object categories, geometric features, permissible contact areas, prohibited contact areas, recommended gripping postures, force limits, safety distances, and the mapping relationship between object categories, task types, and gripping postures. Candidate gripping postures are determined based on this mapping relationship. Each object type and its corresponding task type can correspond to at least one gripping posture, i.e., a candidate gripping posture. For example, for screw-cap objects, the task type is rotation, corresponding to an envelope gripping posture with fingertips evenly distributed along the outer edge; for rod-shaped probe objects, the task type is clamping, corresponding to a three-finger clamping posture with the three fingers arranged circumferentially along the rod axis; for sheet-shaped pressure plate objects, the task type is undulation, corresponding to a lateral flicking posture where fingertips can flick the edge of the pressure plate; for button objects, the task type is pressing, corresponding to a single-finger pressing posture; and for wire objects, the task type is clamping, corresponding to a low-force two-finger clamping posture.

[0076] Optionally, based on the candidate grip postures and a preset grip posture library, parameter information for the candidate grip postures can be determined. The grip posture library can include different postures such as envelope grip, three-finger grip, two-finger grip, lateral flicking, single-finger pressing, supporting grip, and low-force guidance. Each candidate grip posture includes the pre-shaped angle of the knuckles, and may also include parameters such as contact point, grip force range, permissible operating direction, maximum torque, and recommended resistance.

[0077] For example, the power grid object classification library and the gripping posture library can be pre-configured manually or optimized from historical operation data, and updated according to the expansion of actual power grid object categories. This application embodiment does not limit this.

[0078] Step 103: Verify the candidate gripping posture based on the measurement data to determine the target gripping posture.

[0079] Since the candidate grasping postures include at least one, for example, a task category for each object type can correspond to a primary candidate grasping posture and multiple alternative candidate grasping postures, in order to improve the accuracy of grasping control, the candidate grasping postures are further verified based on the measurement data acquired in real time during the robot's dexterous hand movements to determine the matching target grasping posture.

[0080] Step 104: Based on the target gripping posture, perform gripping control on the object to be operated on.

[0081] By controlling components such as finger drive motors, joints, and actuators of the robotic arm, the gripping control is performed on the object to be manipulated based on parameters such as the pre-forming angle of the finger joints, contact point, gripping force range, allowable operating direction, maximum torque, and recommended impedance of the target gripping posture.

[0082] Optionally, during the gripping control process, by acquiring the measurement values ​​of the tactile sensors and torque / current feedback components set on the robot's dexterous hand in real time, slippage detection can be performed. Based on the slippage detection results, the gripping force, impedance and other parameters of the target gripping posture can be further adjusted.

[0083] In the above embodiments, firstly, measurement data of the object to be operated is acquired, and the object type of the object to be operated is determined based on the measurement data. Then, candidate gripping postures are determined based on the object type and the task type of the object to be operated. Next, the candidate gripping postures are verified based on the measurement data to determine the target gripping posture. Finally, gripping control is performed on the object to be operated based on the target gripping posture. In this way, the object type is determined based on the measurement data of the object to be operated, and then the target gripping posture is determined based on the object type and the task type to perform gripping control on the object to be operated. This achieves the determination of the target gripping posture based on the characteristics of the object to be operated, rather than using a uniform gripping posture for gripping control. This can adapt to different types of objects to be operated in the power grid, and the gripping control is more flexible.

[0084] In one embodiment, the measurement data includes image data and depth data, and the candidate grasping poses include a primary candidate grasping pose and alternative candidate grasping poses. The candidate grasping poses are validated based on the measurement data to determine the target grasping pose. Figure 2 As shown, it includes:

[0085] Step 201: Based on image data and depth data, determine the location information and no-touch zone of the object to be operated.

[0086] Based on image and depth data, target detection algorithms (such as YOLO) are used to locate the pixel region of the object to be operated on in the image. Combined with depth information, the three-dimensional centroid coordinates and spatial pose of the object in the robot coordinate system are calculated through camera intrinsic parameter transformation, thus obtaining the position information of the object to be operated on. No-touch zones can include the following situations: First, pre-configured safety constraints in the power grid object classification library, such as live adjacent parts, terminal blocks, transparent oil cups, brittle shells, etc. After determining the object type of the object to be operated on, the no-touch zone is determined according to the pre-configured safety constraints of the object type. Second, obstacles around non-target objects, such as adjacent wires, terminal blocks, shell edges, and adjacent buttons, identified based on image and depth data, are designated as no-touch zones. Third, prohibited contact zones are specified by the task template or manually defined rules, such as "no contact with the wire side" or "no insertion into the terminal block gap."

[0087] Step 202: Based on location information and no-touch zones, perform a safety verification of the primary candidate gripping posture.

[0088] The candidate gripping posture includes a primary candidate gripping posture and alternative candidate gripping postures. First, the primary candidate gripping posture undergoes a security check. If the primary candidate gripping posture fails the security check, one of the alternative candidate gripping postures is adopted as the primary candidate gripping posture and the security check continues until the security check passes, proceeding to the next step. Optional security check methods include... Figure 3 As shown, it includes:

[0089] Step 301: Transform the primary candidate gripping posture to the robot coordinate system and determine the target contact point.

[0090] The primary candidate grasping posture is defined in the camera coordinate system (i.e., the pose of the object to be manipulated as seen by the camera). It needs to be transformed to the robot's base coordinate system. Based on the hand-eye calibration relationship between the camera and the robot's dexterous hand, the primary candidate grasping posture is transformed from the camera coordinate system to the robot's base coordinate system. The rotation matrix and translation vector between the camera coordinate system and the robot's base coordinate system are obtained through hand-eye calibration, and the coordinate transformation is then performed based on these rotation matrix and translation vectors.

[0091] The parameters of the primary candidate grasping posture include the contact point. After coordinate transformation, the contact point becomes the target contact point, which is the candidate fingertip position of the robot's dexterous hand.

[0092] Step 302: If the target contact point does not fall into the no-touch zone, or if the distance between the target contact point and the adjacent object of the object to be operated is greater than the preset distance threshold, then the safety verification of the primary candidate gripping posture is determined to be passed.

[0093] If the target contact point falls into the no-touch zone, or the distance between the target contact point and the adjacent object of the object to be operated is less than or equal to the preset distance threshold, that is, the distance between the target contact point and the adjacent object of the object to be operated is too small, then the safety verification of the primary candidate gripping posture fails, and the primary candidate gripping posture is downgraded to the alternative candidate gripping posture.

[0094] Step 203: If the security check passes, perform a preliminary gripping test based on the primary candidate gripping posture. If the preliminary gripping test is qualified, use the primary candidate gripping posture as the target gripping posture.

[0095] The initial detection steps can be as follows: Figure 4 As shown, it includes:

[0096] Step 401: Determine the preset contact area of ​​the object to be operated on.

[0097] The preset contact area can be pre-set based on the object category, task type, and candidate gripping posture of the object to be operated on. It is a specific fingertip contact sub-region selected from the accessible area of ​​the object to be operated on. The accessible area is the area on the surface of the object to be operated on that is allowed to be touched. It can be determined based on the object category or by identification based on image data. The preset contact area belongs to the accessible area. Therefore, the preset contact area can be understood as the target contact area in the accessible area used for this gripping execution. After determining the object category, task category, and primary candidate gripping posture of the object to be operated on, the preset contact area can be determined by matching according to the preset correspondence.

[0098] Step 402: Based on the pre-forming angle corresponding to the primary candidate gripping posture, make low-force contact with the object to be operated.

[0099] The robot's dexterous hand approaches the object to be manipulated based on the pre-formed angle of the finger joints corresponding to the main candidate grasping posture, and enters the low-force contact stage.

[0100] Step 403: Obtain the actual contact point detected by the tactile sensor. If the actual contact point is located within the preset contact area, the initial grasp test is deemed qualified.

[0101] The robot obtains the actual contact point of the robot's dexterous hand detected by the tactile sensor and determines whether the actual contact point is within the preset contact area. If the actual contact point is within the preset contact area, the initial grasping test is deemed qualified. If the actual contact point is not within the preset area, the contact offset is confirmed. At this time, the pre-forming angle of the robot's dexterous hand's finger joints can be adjusted according to the positional deviation between the actual contact point and the preset contact area to adjust the fingertip position. If the adjusted actual contact point is still not within the preset area, the initial grasping test is deemed unqualified.

[0102] Step 204: If the initial gripping test fails or the safety verification fails, the alternative candidate gripping posture is used as the primary candidate gripping posture, and the step of performing safety verification on the primary candidate gripping posture based on location information and no-touch zone is repeated.

[0103] If the initial gripping test fails or the safety verification fails, the alternative gripping posture is used as the primary candidate gripping posture, and the safety verification steps are repeated. Optionally, if there are multiple alternative gripping postures, the primary candidate gripping posture for the current round can be determined according to the priority of each alternative gripping posture from high to low. Safety verification and initial gripping test are performed on each alternative gripping posture in turn until the target gripping posture is determined. If all alternative gripping postures fail the initial gripping test or the safety verification fails, the step of obtaining the measurement data of the object to be operated on is repeated, or the abnormal information is reported for maintenance personnel to handle.

[0104] If the security check passes and the initial gripping test is qualified, the primary candidate gripping posture for the current round will be taken as the target gripping posture.

[0105] In the above embodiments, candidate gripping postures are selected based on object category, object pose, and task type. Based on the actual perception results, each candidate gripping posture is subjected to security verification and preliminary gripping detection to determine the target gripping posture, so as to better adapt to various types of operation objects.

[0106] In one embodiment, grasping control of the object to be manipulated is performed based on the target grasping posture, such as... Figure 5 As shown, it includes:

[0107] Step 501: Determine the target gripping force range based on the target gripping posture.

[0108] Each candidate gripping posture in the gripping posture library includes parameters such as gripping force range, permissible operating direction, maximum torque, and recommended impedance. The target gripping force range can be determined based on the target gripping posture. The target gripping force range can also be dynamically adjusted according to the surface hardness, surface friction characteristics of the object being manipulated, and the task type. For example, when the task type is capping or valve tightening, the target gripping force range can be appropriately increased to prevent slippage, but it cannot exceed the axial pressure or maximum torque limit; when the task type is wire clamping, the target gripping force range can be appropriately decreased to keep it within a low force range and avoid excessive pulling on the wire.

[0109] Step 502: If the actual contact point is located within the preset contact area, increase the gripping force on the object to be operated to the target gripping force range.

[0110] Based on the target grip posture, the object to be operated is operated. If the actual contact point is located within the preset contact area, the grip force is gradually increased to the target grip force range.

[0111] Step 503: Obtain the pressure distribution of the object to be operated detected by the pressure sensor.

[0112] Step 504: Adjust the gripping force of the object to be operated on based on pressure distribution.

[0113] During the target gripping posture operation, slippage detection is performed, and the gripping force is dynamically adjusted based on the slippage index obtained from the slippage detection. For caps or valves, if slippage occurs during rotation, the normal gripping force is increased or the fingertip distribution is adjusted. For probes, if the axial deviation increases, the advancement speed is reduced and the clamping posture is corrected. For wires, if the tension exceeds the threshold, the clamping force is reduced and the movement is stopped.

[0114] The slippage index can be obtained by combining tactile pressure distribution, force feedback, and visual motion feedback. Specifically, it can include: the amount or speed of pressure center shift over time exceeding a preset shift threshold; a sudden decrease in contact area; a drop in local pressure; an abnormal ratio of tangential to normal force; micro-vibrations or high-frequency components in the tactile signal; and inconsistencies between object pose changes and finger movements. For example, in capping operations, if the finger has already rotated but the capping angle change is insufficient, and the pressure center continues to shift or there is a sudden drop in local pressure, then a slippage trend can be identified.

[0115] Optionally, when gripping fails, sliding continues, the object does not move as expected, or the contact force exceeds the limit, alternative gripping postures for similar objects can be selected based on the cause of failure. For example, when gripping and sliding with a screw cap envelope, switch to a three-finger envelope with a larger contact area; when the rod-shaped probe deflects, switch to a gripping point closer to the center of gravity; and when the pressure plate fails to move, switch to a single-finger pressing or lateral gripping mode.

[0116] Grasping failure can be detected by comprehensively considering contact status, gripping stability, task execution results, and safety thresholds. Contact failure includes failure to detect the expected tactile pressure or contact point deviating from the preset contact area; stability failure includes multi-finger pressure difference exceeding the threshold, continuous shift of the pressure center, and slippage exceeding the threshold; task failure includes the cap not reaching the target rotation angle, the pressure plate not being moved to the target position, and probe axis deviation exceeding the threshold; safety failure includes contact force, joint torque, motor current, and wire tension exceeding the threshold; and recognition failure includes low confidence in the object category or being too close to a no-touch area. When any of the above situations are detected, the current grip or operation is determined to have failed, and the reason for the failure is displayed, or the system switches to other alternative gripping postures for further testing.

[0117] In the embodiments of this application, such as Figure 6 The present invention provides a robot dexterous hand grasping control method, comprising:

[0118] Step 601: Obtain the measurement data of the object to be operated on, and determine the object type of the object to be operated on based on the measurement data.

[0119] Step 602: Determine candidate gripping postures based on the object type and the task type of the object to be operated on.

[0120] Step 603: Based on image data and depth data, determine the location information and no-touch zone of the object to be operated.

[0121] Step 604: Based on location information and no-touch zones, perform a safety verification of the primary candidate gripping posture.

[0122] Step 605: If the security check passes, perform a preliminary gripping test based on the primary candidate gripping posture. If the preliminary gripping test is qualified, use the primary candidate gripping posture as the target gripping posture.

[0123] Step 606: If the initial gripping test fails or the safety verification fails, the alternative candidate gripping posture is used as the primary candidate gripping posture, and the step of performing safety verification on the primary candidate gripping posture based on location information and no-touch zone is repeated.

[0124] Step 607: Based on the target gripping posture, perform gripping control on the object to be operated on.

[0125] In the embodiments of this application, the robot dexterous hand grasping control method includes a robot dexterous hand grasping control system that can be applied to a robot. The robot dexterous hand grasping control system includes an object perception component, a tactile and hardness estimation component, an electric field object classification library, a grasping posture library, a posture matching and correction unit, and a grasping execution and feedback correction unit.

[0126] The object perception component may include a head camera, wrist camera, depth sensor, or other vision / distance sensor mounted on the robot. It processes the measurement data through image processing, point cloud processing, object detection, contour extraction, size calculation, spindle estimation, object pose calculation, accessible area recognition, and obstacle recognition. The tactile and hardness estimation component uses tactile array sensors, fingertip force sensors, joint torque sensors, or motor current detection units used to indirectly estimate contact forces, mounted on the robot's dexterous fingertips, finger pads, or contact pads, to acquire the surface hardness and surface friction characteristics of the object to be manipulated. These, along with the aforementioned measurement data, form the attribute feature vector of the object to be manipulated, thereby determining the object type.

[0127] Candidate gripping postures are determined based on the object type and task type using a power grid object classification library, and the parameters of each candidate gripping posture are determined using a gripping posture library. Parameters may include knuckle pre-forming angle, contact point or contact area, approach direction, gripping force range, permissible operating direction, maximum torque, recommended impedance parameters, and candidate posture priority.

[0128] The posture matching and correction unit performs safety verification and preliminary gripping detection on each candidate gripping posture based on the object category, task type, actual object pose, accessible area and prohibited area, and determines the target gripping posture.

[0129] The gripping execution and feedback correction unit includes a robotic arm, a dexterous hand, finger drive motors, joints, drivers, controllers, tactile sensors, and torque / current feedback components. It performs gripping control on the object to be manipulated according to the target gripping posture, acquires the contact stability and slippage trend during the action execution process in real time, and determines whether it is necessary to adjust the parameters of the target gripping posture or switch the candidate gripping posture.

[0130] In the above embodiments, a power grid object classification library is established based on the characteristics of the objects to be operated on in the power grid. This avoids the uncertainty of general grasping algorithms in fine-grained operations of the power grid. Each grasping posture in the grasping posture library is linked to the grasping force limit, no-touch area, and impedance parameters, enabling the grasping action to serve subsequent tasks such as screwing, inserting, turning, or pressing. Furthermore, tactile feedback is obtained through low-force contact for slip detection, adjusting the grasping force and contact point to improve the grasping stability of different objects such as caps, probes, pressure plates, and wires. At the same time, if a grasping fails, the reason for the failure is returned, and other alternative candidate grasping postures are switched for re-detection, improving the continuity of operations in complex field conditions.

[0131] In the embodiments of this application, such as Figure 7 As shown, taking the grasping of a screw cap as an example, the robot's wrist camera identifies the outer edge and central axis of the screw cap. The robot's dexterous hand grasping control system determines its classification as a screw cap / knob object by acquiring measurement data, and identifies the corresponding candidate grasping postures, which are either three-finger envelope grasping or multi-finger envelope grasping. The parameters for three-finger envelope grasping or multi-finger envelope grasping are obtained from the grasping posture library to determine the contact point, with fingertips evenly distributed along the outer circumference of the screw cap. The three-finger envelope grasping passes safety verification and preliminary grasping detection, thus confirming it as the target grasping posture. The robot then uses this target grasping posture to control the grasping of the object, entering a low-force contact stage. The tactile array sensor confirms whether the pressure of each finger is balanced. If the pressure of one finger is significantly lower than the others, it indicates grasping eccentricity and low contact stability; the contact point position of that finger is then adjusted. After the grasping is confirmed to be stable, a rotational torque is applied along the screw cap axis to the target grasping force range, and axial pressure is limited, thus achieving grasping control of the object.

[0132] In the embodiments of this application, such as Figure 8As shown, the grasping of a rod-shaped probe-like object is illustrated using this example. Measurement data is used to classify the object as a rod-shaped probe, determining the main axis direction and gripping area of ​​the slender probe. The corresponding candidate grasping posture is determined to be a three-finger gripping posture, where two fingers provide lateral gripping and the third finger provides posture stability. The center of gravity position is determined based on the probe length, and a gripping position close to the center of gravity is selected. Safety verification and preliminary grasping tests of the three-finger gripping posture both pass, confirming it as the target grasping posture. During the insertion of the slender probe, if the robot's wrist camera detects that the probe axis deviates from the insertion port axis, the robot's dexterous hand adjusts the parameters of the three-finger gripping posture, such as reducing gripping stiffness and adjusting the gripping posture.

[0133] In the embodiments of this application, such as Figure 9 As shown, taking a sheet-like pressure plate as an example, by acquiring measurement data, the edge and rotation axis of the pressure plate inside the terminal box are identified, the object is classified as a sheet-like pressure plate, and the candidate gripping postures are determined to be either a lateral flicking posture or a single-finger pressing posture. The parameters of the candidate gripping postures are determined according to the gripping posture library. The lateral flicking posture passes both safety verification and preliminary gripping detection, thus confirming it as the target gripping posture. If there is a wire beside the pressure plate, the wire side is designated as a no-touch zone, and the pressure plate edge is contacted from the other side. During operation based on the target gripping posture, the tactile sensor confirms that the target contact point is within the edge range of the pressure plate, and then force is applied along the legal flicking direction of the pressure plate to the gripping force range, thus realizing the operation of the object.

[0134] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0135] Based on the same inventive concept, this application also provides a robot dexterous hand gripping control device for implementing the robot dexterous hand gripping control method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more robot dexterous hand gripping control device embodiments provided below can be found in the limitations of the robot dexterous hand gripping control method above, and will not be repeated here.

[0136] In one exemplary embodiment, such as Figure 10 As shown, a robot dexterous hand grasping control device is provided, comprising: an acquisition module, a first determining module, a second determining module, and a control module, wherein:

[0137] The acquisition module is used to acquire measurement data of the object to be operated on, and determine the object type of the object to be operated on based on the measurement data;

[0138] The first determination module is used to determine candidate gripping postures based on the object type and the task type of the object to be operated on.

[0139] The second determination module is used to verify the candidate gripping posture based on the measurement data and determine the target gripping posture.

[0140] The control module is used to control the grasping of the object to be operated based on the target grasping posture.

[0141] In one embodiment, the measurement data includes image data and depth data, and the candidate gripping postures include a primary candidate gripping posture and alternative candidate gripping postures. The second determining module is specifically used to determine the location information and no-touch area of ​​the object to be operated based on the image data and depth data; to perform a security verification on the primary candidate gripping posture based on the location information and no-touch area; if the security verification passes, a preliminary gripping detection is performed based on the primary candidate gripping posture; if the preliminary gripping detection is qualified, the primary candidate gripping posture is used as the target gripping posture; if the preliminary gripping detection fails or the security verification fails, the alternative candidate gripping posture is used as the primary candidate gripping posture, and the step of performing a security verification on the primary candidate gripping posture based on the location information and no-touch area is repeated.

[0142] In one embodiment, the second determining module is specifically used to transform the primary candidate gripping posture to the robot coordinate system and determine the target contact point; if the target contact point does not fall into the no-touch area, or if the distance between the target contact point and the adjacent object of the object to be operated is greater than a preset distance threshold, then the safety verification of the primary candidate gripping posture is determined to be passed.

[0143] In one embodiment, the second determining module is specifically used to determine the preset contact area of ​​the object to be operated; to make low-force contact with the object to be operated based on the pre-forming angle corresponding to the main candidate grasping posture; to obtain the actual contact point detected by the tactile sensor; and if the actual contact point is located within the preset contact area, then the grasping preliminary detection is determined to be qualified.

[0144] In one embodiment, the control module is specifically used to determine the target gripping force range based on the target gripping posture; if the actual contact point is located within a preset contact area, increase the gripping force of the object to be operated to the target gripping force range; acquire the pressure distribution of the object to be operated detected by the pressure sensor; and adjust the gripping force of the object to be operated based on the pressure distribution.

[0145] In one embodiment, the acquisition module is specifically used to extract features based on measurement data to obtain attribute feature vectors; and to perform classification detection based on the attribute feature vectors to determine the object type.

[0146] The modules in the aforementioned robot dexterous hand grasping control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0147] In one exemplary embodiment, a computer device is provided, which may be a robot, and its internal structure diagram may be as follows: Figure 11As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a robotic dexterous hand grasping control method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0148] Those skilled in the art will understand that Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0149] In one exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: acquiring measurement data of an object to be operated on, and determining the object type of the object to be operated on based on the measurement data; determining candidate gripping postures based on the object type and the task type of the object to be operated on; verifying the candidate gripping postures based on the measurement data to determine the target gripping posture; and performing gripping control on the object to be operated on based on the target gripping posture.

[0150] In one embodiment, the measurement data includes image data and depth data, and the candidate gripping postures include a primary candidate gripping posture and alternative candidate gripping postures. When the processor executes the computer program, it further implements the following steps: determining the location information and no-touch area of ​​the object to be operated based on the image data and depth data; performing a security check on the primary candidate gripping posture based on the location information and no-touch area; if the security check passes, performing a preliminary gripping detection based on the primary candidate gripping posture; if the preliminary gripping detection passes, using the primary candidate gripping posture as the target gripping posture; if the preliminary gripping detection fails or the security check fails, using the alternative candidate gripping posture as the primary candidate gripping posture, and repeating the step of performing a security check on the primary candidate gripping posture based on the location information and no-touch area.

[0151] In one embodiment, when the processor executes the computer program, it further performs the following steps: transforming the primary candidate gripping posture to the robot coordinate system and determining the target contact point; if the target contact point does not fall into the no-collision zone, or if the distance between the target contact point and the adjacent object of the object to be operated is greater than a preset distance threshold, then the safety verification of the primary candidate gripping posture is determined to be passed.

[0152] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining a preset contact area of ​​the object to be operated; making low-force contact with the object to be operated based on the pre-forming angle corresponding to the main candidate gripping posture; obtaining the actual contact point detected by the tactile sensor; if the actual contact point is located within the preset contact area, then determining that the preliminary gripping detection is qualified.

[0153] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the target gripping force range based on the target gripping posture; if the actual contact point is located within a preset contact area, increasing the gripping force of the object to be operated to the target gripping force range; acquiring the pressure distribution of the object to be operated detected by the pressure sensor; and adjusting the gripping force of the object to be operated based on the pressure distribution.

[0154] In one embodiment, when the processor executes the computer program, it further performs the following steps: extracting features from the measurement data to obtain an attribute feature vector; and performing classification detection based on the attribute feature vector to determine the object type.

[0155] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon. When the computer program is executed by a processor, it performs the following steps: acquiring measurement data of an object to be operated on, and determining the object type of the object to be operated on based on the measurement data; determining candidate gripping postures based on the object type and the task type of the object to be operated on; verifying the candidate gripping postures based on the measurement data to determine the target gripping posture; and performing gripping control on the object to be operated on based on the target gripping posture.

[0156] In one embodiment, the measurement data includes image data and depth data, and the candidate gripping postures include a primary candidate gripping posture and alternative candidate gripping postures. When the computer program is executed by the processor, it further implements the following steps: determining the location information and no-touch area of ​​the object to be operated based on the image data and depth data; performing a security check on the primary candidate gripping posture based on the location information and no-touch area; if the security check passes, performing a preliminary gripping detection based on the primary candidate gripping posture; if the preliminary gripping detection passes, using the primary candidate gripping posture as the target gripping posture; if the preliminary gripping detection fails or the security check fails, using the alternative candidate gripping posture as the primary candidate gripping posture, and repeating the step of performing a security check on the primary candidate gripping posture based on the location information and no-touch area.

[0157] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: transforming the primary candidate gripping posture to the robot coordinate system and determining the target contact point; if the target contact point does not fall into the no-touch area, or if the distance between the target contact point and the adjacent object of the object to be operated is greater than a preset distance threshold, then the safety verification of the primary candidate gripping posture is determined to be passed.

[0158] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining a preset contact area of ​​the object to be operated; making low-force contact with the object to be operated based on the pre-forming angle corresponding to the primary candidate gripping posture; obtaining the actual contact point detected by the tactile sensor; if the actual contact point is located within the preset contact area, then determining that the preliminary gripping detection is qualified.

[0159] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the target gripping force range based on the target gripping posture; if the actual contact point is located within a preset contact area, increasing the gripping force of the object to be operated to the target gripping force range; acquiring the pressure distribution of the object to be operated detected by the pressure sensor; and adjusting the gripping force of the object to be operated based on the pressure distribution.

[0160] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: extracting features from the measurement data to obtain an attribute feature vector; and performing classification detection based on the attribute feature vector to determine the object type.

[0161] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps: acquiring measurement data of an object to be operated on, and determining the object type of the object to be operated on based on the measurement data; determining candidate gripping postures based on the object type and the task type of the object to be operated on; verifying the candidate gripping postures based on the measurement data to determine the target gripping posture; and performing gripping control on the object to be operated on based on the target gripping posture.

[0162] In one embodiment, the measurement data includes image data and depth data, and the candidate gripping postures include a primary candidate gripping posture and alternative candidate gripping postures. When the computer program is executed by the processor, it further implements the following steps: determining the location information and no-touch area of ​​the object to be operated based on the image data and depth data; performing a security check on the primary candidate gripping posture based on the location information and no-touch area; if the security check passes, performing a preliminary gripping detection based on the primary candidate gripping posture; if the preliminary gripping detection passes, using the primary candidate gripping posture as the target gripping posture; if the preliminary gripping detection fails or the security check fails, using the alternative candidate gripping posture as the primary candidate gripping posture, and repeating the step of performing a security check on the primary candidate gripping posture based on the location information and no-touch area.

[0163] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: transforming the primary candidate gripping posture to the robot coordinate system and determining the target contact point; if the target contact point does not fall into the no-touch area, or if the distance between the target contact point and the adjacent object of the object to be operated is greater than a preset distance threshold, then the safety verification of the primary candidate gripping posture is determined to be passed.

[0164] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining a preset contact area of ​​the object to be operated; making low-force contact with the object to be operated based on the pre-forming angle corresponding to the primary candidate gripping posture; obtaining the actual contact point detected by the tactile sensor; if the actual contact point is located within the preset contact area, then determining that the preliminary gripping detection is qualified.

[0165] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the target gripping force range based on the target gripping posture; if the actual contact point is located within a preset contact area, increasing the gripping force of the object to be operated to the target gripping force range; acquiring the pressure distribution of the object to be operated detected by the pressure sensor; and adjusting the gripping force of the object to be operated based on the pressure distribution.

[0166] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: extracting features from the measurement data to obtain an attribute feature vector; and performing classification detection based on the attribute feature vector to determine the object type.

[0167] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0168] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0169] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0170] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for controlling the grasping action of a robot's dexterous hand, characterized in that, The method includes: Obtain measurement data of the object to be operated on, and determine the object type of the object to be operated on based on the measurement data; Based on the object type and the task type of the object to be operated on, candidate gripping postures are determined; The candidate gripping postures are verified based on the measurement data to determine the target gripping posture; Based on the target gripping posture, the object to be operated is gripped and controlled.

2. The method according to claim 1, characterized in that, The measurement data includes image data and depth data. The candidate gripping poses include a primary candidate gripping pose and alternative candidate gripping poses. The candidate gripping poses are validated based on the measurement data to determine the target gripping pose, including: Based on the image data and the depth data, the location information and no-touch zone of the object to be operated are determined; Based on the location information and the no-touch zone, the primary candidate gripping posture is subjected to security verification; If the security check passes, a preliminary gripping test is performed based on the primary candidate gripping posture. If the preliminary gripping test is qualified, the primary candidate gripping posture is taken as the target gripping posture. If the initial gripping test fails or the security check fails, the alternative candidate gripping posture is used as the primary candidate gripping posture, and the step of performing security check on the primary candidate gripping posture based on the location information and the no-touch zone is repeated.

3. The method according to claim 2, characterized in that, The security verification of the primary candidate gripping posture based on the location information and the no-touch zone includes: Transform the primary candidate grasping posture to the robot coordinate system and determine the target contact point; If the target contact point does not fall into the no-touch zone, or if the distance between the target contact point and the adjacent object of the object to be operated is greater than a preset distance threshold, then the safety verification of the primary candidate gripping posture is determined to be passed.

4. The method according to claim 3, characterized in that, The preliminary grip detection based on the primary candidate grip posture includes: Determine the preset contact area of ​​the object to be operated on; Based on the pre-forming angle corresponding to the main candidate gripping posture, low-force contact is made with the object to be operated; The actual contact point detected by the tactile sensor is obtained. If the actual contact point is located within the preset contact area, the preliminary gripping test is determined to be qualified.

5. The method according to claim 4, characterized in that, The grasping control of the object to be operated based on the target grasping posture includes: Determine the target gripping force range based on the target gripping posture; If the actual contact point is located within the preset contact area, increase the gripping force on the object to be operated to the target gripping force range; Acquire the pressure distribution of the object to be operated as detected by the pressure sensor; Based on the pressure distribution, adjust the gripping force of the object to be operated.

6. The method according to any one of claims 1 to 5, characterized in that, Determining the object type of the object to be operated on based on the measurement data includes: Based on the measurement data, feature extraction is performed to obtain attribute feature vectors; The object type is determined by performing classification and detection based on the attribute feature vector.

7. A robot dexterous hand grasping control device, characterized in that, The device includes: The acquisition module is used to acquire measurement data of the object to be operated on, and determine the object type of the object to be operated on based on the measurement data; The first determining module is used to determine candidate gripping postures based on the object type and the task type of the object to be operated on. The second determining module is used to verify the candidate gripping posture based on the measurement data and determine the target gripping posture. The control module is used to control the gripping of the object to be operated based on the target gripping posture.

8. A robot comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.