A control method and system for a seven-degree-of-freedom robotic arm

Abstract

The application discloses a kind of seven degrees of freedom mechanical arm control method and system, related to mechanical arm control technical field, when receiving the target orientation of seven degrees of freedom mechanical arm, control seven degrees of freedom mechanical arm rotates to target orientation, and obtains the mechanical arm end position of seven degrees of freedom mechanical arm and the image of target to be grabbed, using pre-trained distance identification model to carry out distance detection to target to be grabbed image, generate target distance;Wherein, distance identification model includes first feature extraction group, second feature extraction group and distance detection network, based on preset impedance control function, mechanical arm end position and target distance, control seven degrees of freedom mechanical arm executes grabbing operation.The technical problem that it is solved by the way of artificial control mechanical arm, lack relevant measurement means, increase the difficulty of mechanical arm control, lead to the work efficiency of live cooperative work is lower.

Description

Technical Field

[0001] This invention relates to the field of robotic arm control technology, and in particular to a control method and system for a seven-degree-of-freedom robotic arm. Background Technology

[0002] In recent years, foreign objects tangling in power lines have become a major threat to the safe operation of power grids. Operational departments routinely use the ground potential method to handle this, but this method is greatly limited by the location of the foreign object and the surrounding terrain. Power outages for handling this issue affect transmission reliability. If workers are required to enter the electric field for equipotential work, it faces disadvantages such as aging tools, high labor intensity and safety risks, and low work efficiency. Therefore, robotic arms have become key equipment for power operations, undertaking live-line collaborative work.

[0003] Currently, live-line collaborative work on power transmission lines is mainly carried out by manually controlling robotic arms. However, with manual control, the distance between the robotic arm and the target point, as well as the force output, mainly rely on the experience of the operators, lacking relevant measurement methods. This increases the difficulty of controlling the robotic arm and results in low efficiency of live-line collaborative work. Summary of the Invention

[0004] This invention provides a control method and system for a seven-degree-of-freedom robotic arm, which solves the technical problem that when performing live-line collaborative work on power transmission lines by manually controlling the robotic arm, the distance between the robotic arm and the target point and the force output mainly rely on the operator's experience and lack relevant measurement methods, which increases the difficulty of controlling the robotic arm and leads to low work efficiency in live-line collaborative work.

[0005] The first aspect of this invention provides a control method for a seven-degree-of-freedom robotic arm, the method comprising:

[0006] When the target orientation of the seven-degree-of-freedom robotic arm is received, the seven-degree-of-freedom robotic arm is controlled to rotate to the target orientation, and the position of the end effector of the seven-degree-of-freedom robotic arm and the image of the target to be grasped are obtained.

[0007] A pre-trained distance recognition model is used to perform distance detection on the target image to be captured, generating the target distance; wherein, the distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network;

[0008] Based on a preset impedance control function, the end position of the robotic arm, and the target distance, the seven-degree-of-freedom robotic arm is controlled to perform a grasping operation.

[0009] Optionally, the step of using a pre-trained distance recognition model to perform distance detection on the target image to be captured and generating the target distance includes:

[0010] The first feature extraction group is used to extract features from the target image to be captured to generate a first distance feature map. The first feature extraction group includes a zero-padding layer, a convolutional layer, a normalization layer, a ReLU activation layer, a zero-padding layer, and a ReLU activation layer connected in sequence.

[0011] The second feature extraction group extracts features from the second distance feature map to generate a second distance feature map. The second feature extraction group includes four first feature extraction modules connected in sequence.

[0012] The distance detection network is used to perform distance detection on the second distance feature map to generate the target distance. The distance detection network includes a normalization layer, a ReLU activation layer, a convolutional layer, an upsampling layer, and four second feature extraction modules connected in sequence.

[0013] Optionally, the first feature extraction module includes dense blocks and transition blocks connected in sequence, and the specific processing procedure of the first feature extraction module is as follows:

[0014] The first feature map is processed by dense blocks to extract features from the input feature map and generate a second feature map. The dense blocks include a dense extraction layer, a feature fusion layer, a dense extraction layer, a feature fusion layer, and a dense extraction layer connected in sequence.

[0015] The second feature map is used to extract features by a transition block to generate a third feature map. The transition block includes a convolutional layer and an averaging layer connected in sequence.

[0016] Optionally, the step of extracting features from the input first feature map using dense blocks to generate a second feature map includes:

[0017] The first feature map is extracted by a dense extraction layer to generate a first transition feature map. The density extraction layer includes a convolutional layer, a normalization layer and a ReLU activation layer connected in sequence.

[0018] A feature fusion layer is used to fuse the first transition feature map and the first feature map to generate a first fused feature map.

[0019] The first fused feature map is used to extract features through a dense extraction layer to generate a second transition feature map;

[0020] A feature fusion layer is used to fuse the second transition feature map, the first transition feature map, and the first feature map to generate a second fused feature map.

[0021] A third transition feature map is generated by extracting features from the second fused feature map through a dense extraction layer.

[0022] A feature fusion layer is used to fuse the third transition feature map, the first transition feature map, and the first feature map to generate a second feature map.

[0023] Optionally, the second feature extraction module includes a UP block and an upsampling layer connected in sequence, and the specific processing procedure of the second feature extraction module is as follows:

[0024] The third feature map of the input is extracted by the UP block to generate the fourth transition feature map. The UP block includes a convolutional layer, a convolutional layer and a ReLU activation layer connected in sequence.

[0025] An upsampling layer is used to upsample the fourth transition feature map to generate a fifth transition feature map.

[0026] Optionally, the step of controlling the seven-degree-of-freedom robotic arm to perform a grasping operation based on a preset impedance control function, the end-effector position, and the target distance includes:

[0027] Determine whether the target distance is greater than or equal to a preset distance threshold;

[0028] If the target distance is less than or equal to the distance threshold, the position of the robotic arm end is input into a preset impedance control function to generate corresponding adjustment data;

[0029] The steady-state position of the seven-degree-of-freedom robotic arm is adjusted according to the adjustment data, and the seven-degree-of-freedom robotic arm is controlled to move according to the target distance and then perform a grasping operation;

[0030] If the target distance is greater than the distance threshold, then adjust the position of the seven-degree-of-freedom robotic arm;

[0031] Jump to execute the steps of controlling the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtaining the end-effector position of the seven-degree-of-freedom robotic arm and the image of the target to be grasped.

[0032] Optionally, the impedance control function is specifically:

[0033]

[0034]

[0035]

[0036] Among them, F d Let M be the desired contact force at the end effector of the robotic arm, B be the inertia coefficient matrix, K be the damping coefficient matrix, and x be the stiffness coefficient matrix. d Let x be the desired position of the robotic arm's end effector, and let x be the position of the robotic arm's end effector. Let be the second derivative of the desired position at the end effector of the robotic arm. Let be the second derivative at the end effector position of the robotic arm. Let be the first derivative of the desired position at the end of the robotic arm. Let x be the first derivative of the position of the robotic arm's end effector. e B represents the target position in the inertial coordinate system. e Let K be the environmental damping matrix. e Let x be the environmental stiffness matrix. SS F represents the steady-state position of the robotic arm's end effector. SS For steady-state contact force, E fss This represents the steady-state force error.

[0037] The second aspect of this invention provides a control system for a seven-degree-of-freedom robotic arm, comprising:

[0038] The response module is used to control the seven-degree-of-freedom robotic arm to rotate to the target orientation when the target orientation is received, and to obtain the end position of the robotic arm and the image of the target to be grasped.

[0039] The detection module is used to perform distance detection on the target image to be captured using a pre-trained distance recognition model to generate the target distance; wherein, the distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network;

[0040] The grasping module is used to control the seven-degree-of-freedom robotic arm to perform grasping operations based on a preset impedance control function, the position of the robotic arm's end effector, and the target distance.

[0041] Optionally, the detection module includes:

[0042] The first extraction submodule is used to extract features from the target image to be captured using the first feature extraction group to generate a first distance feature map. The first feature extraction group includes a zero-padding layer, a convolutional layer, a normalization layer, a ReLU activation layer, a zero-padding layer, and a ReLU activation layer connected in sequence.

[0043] The second extraction submodule is used to extract features from the second distance feature map through the second feature extraction group to generate the second distance feature map. The second feature extraction group includes four first feature extraction modules connected in sequence.

[0044] The detection submodule is used to perform distance detection on the second distance feature map using the distance detection network to generate the target distance. The distance detection network includes a normalization layer, a ReLU activation layer, a convolutional layer, an upsampling layer, and four second feature extraction modules connected in sequence.

[0045] Optionally, the grasping module includes:

[0046] The analysis submodule is used to determine whether the target distance is greater than or equal to a preset distance threshold;

[0047] If the target distance is greater than or equal to the distance threshold, the position of the robotic arm end is input into a preset impedance control function to generate corresponding adjustment data;

[0048] The first execution submodule is used to adjust the steady-state position of the seven-degree-of-freedom robotic arm according to the adjustment data, and control the seven-degree-of-freedom robotic arm to move according to the target distance and then perform a grasping operation;

[0049] The second execution submodule is used to adjust the position of the seven-degree-of-freedom robotic arm if the target distance is less than a preset distance threshold.

[0050] The steps include adjusting and executing the control of the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtaining the end-effector position of the seven-degree-of-freedom robotic arm and the image of the target to be grasped.

[0051] As can be seen from the above technical solutions, the present invention has the following advantages:

[0052] When the target orientation of the seven-degree-of-freedom (DOF) robotic arm is received, the arm is rotated to the target orientation. The end-effector position and the image of the target to be grasped are acquired. A pre-trained distance recognition model is used to detect the distance to the target image and generate the target distance. The distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network. Based on a preset impedance control function, the end-effector position, and the target distance, the seven-degree-of-freedom robotic arm is controlled to perform the grasping operation. This addresses the technical problem that manual control of the robotic arm lacks relevant measurement methods, increasing the difficulty of robotic arm control and resulting in low efficiency in live-line collaborative operations. This application improves the accuracy of the robotic arm's grasping of objects and increases the efficiency of live-line collaborative operations by acquiring the end-effector position and the target image, combined with a preset distance recognition model and impedance control function, and adjusting the state of the robotic arm. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1This is a flowchart illustrating the steps of a control method for a seven-degree-of-freedom robotic arm provided in Embodiment 1 of the present invention.

[0055] Figure 2 This is a flowchart illustrating the steps of a control method for a seven-degree-of-freedom robotic arm provided in Embodiment 2 of the present invention.

[0056] Figure 3 This is a schematic diagram of the distance recognition model provided in Embodiment 2 of the present invention;

[0057] Figure 4 This is a structural block diagram of a control system for a seven-degree-of-freedom robotic arm provided in Embodiment 3 of the present invention. Detailed Implementation

[0058] This invention provides a control method and system for a seven-degree-of-freedom robotic arm, which addresses the technical problem of low efficiency in live-line collaborative work on power transmission lines when the robotic arm is manually controlled. This is because the distance between the robotic arm and the target point, as well as the force output, rely mainly on the operator's experience and lack relevant measurement methods, which increases the difficulty of controlling the robotic arm and leads to low efficiency in live-line collaborative work.

[0059] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0060] Please see Figure 1 , Figure 1 The flowchart illustrates the steps of a control method for a seven-degree-of-freedom robotic arm provided in Embodiment 1 of the present invention.

[0061] This invention provides a control method for a seven-degree-of-freedom robotic arm, the method comprising:

[0062] Step 101: When the target orientation of the seven-degree-of-freedom robotic arm is received, control the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtain the position of the end effector of the seven-degree-of-freedom robotic arm and the image of the target to be grasped.

[0063] The target location refers to the location of the target object that needs to be captured on the power transmission line.

[0064] A seven-degree-of-freedom robotic arm refers to a robotic arm mechanism with ≥7 joints capable of independent movement.

[0065] In this embodiment of the invention, when the orientation of the target object to be grasped by the seven-degree-of-freedom robotic arm is received, the seven-degree-of-freedom robotic arm is controlled to rotate so that the camera on the robotic arm faces the target object, and the position of the end of the seven-degree-of-freedom robotic arm and the image of the target object to be grasped captured by the camera are obtained at the current moment.

[0066] Step 102: Use a pre-trained distance recognition model to perform distance detection on the target image to be captured and generate the target distance; wherein, the distance recognition model includes a first feature extraction group, a second feature extraction group and a distance detection network.

[0067] In this embodiment of the invention, the target image to be captured is input into a pre-trained distance recognition model for distance detection, a depth map is generated, and the corresponding target distance is extracted from the depth map. The distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network.

[0068] It should be noted that before inputting the target image to be captured into the pre-trained distance recognition model for distance detection, training data (including images of foreign objects on the line, bird nests, grounding wires, and voltage detection targets) needs to be acquired. Simultaneously, Photoshop is used to adjust the brightness and contrast of the training and test sets within the training data, and labels are applied. The training data is then used to train the preset distance recognition model, resulting in a trained model. The training data consists of a training set and a test set. The training process is as follows: the training set is input into the preset distance recognition model to generate training feature maps; a standard feature map corresponding to the training feature map is selected from the test set; the overlap between the training feature map and the standard feature map is calculated; and the percentage overlap between the training feature map and the standard feature map is determined to be less than or equal to 99%. If the overlap is greater than 99%, the trained distance recognition model is output; if the overlap is less than or equal to 99%, the model parameters of the insulator defect detection model are adjusted using gradient descent, and the process jumps to the step of inputting the training set into the preset distance recognition model to generate training feature maps.

[0069] Notably, in one embodiment, the distance recognition model uses a skip-connection encoder-decoder network, DenseNet-169, as the encoder for feature extraction from power monocular images. The last layer of each convolutional block in the encoder consists of two bilinear sampling blocks and a ReLU activation function with downsampling parameters, which reduces computational complexity while obtaining more spatial features. The decoder comprises convolution operations and bilinear upsampling operations. The convolutional blocks of the corresponding encoder are skip-connected to the upsampling blocks of the corresponding decoder. This expands the feature map while obtaining a refined edge structure feature map, reducing feature loss. The feature map is the depth map directly output after the convolution operation. The resolution of the output depth map is half that of the input image.

[0070] Step 103: Based on the preset impedance control function, the position of the robotic arm end effector and the target distance, control the seven-degree-of-freedom robotic arm to perform a grasping operation.

[0071] In this embodiment of the invention, if the target distance exceeds a preset distance threshold, and if the target distance does not exceed the distance threshold, the end position of the robotic arm is input into a preset impedance control function to generate a corresponding steady-state position, steady-state contact force, and steady-state force error. The steady-state position, steady-state contact force, and steady-state force error are then used to generate a joint adjustment signal to adjust the state of the seven-degree-of-freedom robotic arm. After adjusting the state of the robotic arm, the robotic arm is controlled to perform a grasping operation. If the target distance exceeds the distance threshold, the position of the seven-degree-of-freedom robotic arm is adjusted, and the process jumps to step 101.

[0072] It should be noted that the impedance control function is as follows:

[0073]

[0074]

[0075]

[0076] Among them, F d Let M be the desired contact force at the end effector of the robotic arm, B be the inertia coefficient matrix, K be the damping coefficient matrix, and x be the stiffness coefficient matrix. d Let x be the desired position of the robotic arm's end effector, and let x be the position of the robotic arm's end effector. Let be the second derivative of the desired position at the end effector of the robotic arm. Let be the second derivative at the end effector position of the robotic arm. Let be the first derivative of the desired position at the end of the robotic arm. Let x be the first derivative of the position of the robotic arm's end effector. e B represents the target position in the inertial coordinate system. e Let K be the environmental damping matrix. e Let x be the environmental stiffness matrix.SS F represents the steady-state position of the robotic arm's end effector. SS For steady-state contact force, E fss This represents the steady-state force error.

[0077] In this embodiment of the invention, when the target orientation of the seven-degree-of-freedom (DOF) robotic arm is received, the robotic arm is controlled to rotate to the target orientation. The end-effector position of the robotic arm and the image of the target to be grasped are acquired. A pre-trained distance recognition model is used to detect the distance to the target image and generate the target distance. The distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network. Based on a preset impedance control function, the end-effector position, and the target distance, the seven-degree-of-freedom robotic arm is controlled to perform the grasping operation. This addresses the technical problem that manual control of the robotic arm lacks relevant measurement methods, increasing the difficulty of robotic arm control and resulting in low efficiency in live-line collaborative operations. This application improves the accuracy of the robotic arm's grasping of objects and increases the efficiency of live-line collaborative operations by acquiring the end-effector position and the target image, combined with a preset distance recognition model and impedance control function, and adjusting the state of the robotic arm.

[0078] Please see Figure 2 , Figure 2 The flowchart illustrates the steps of a control method for a seven-degree-of-freedom robotic arm provided in Embodiment 2 of the present invention.

[0079] This invention provides a control method for a seven-degree-of-freedom robotic arm, the method comprising:

[0080] Step 201: When the target orientation of the seven-degree-of-freedom robotic arm is received, control the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtain the position of the end effector of the seven-degree-of-freedom robotic arm and the image of the target to be grasped.

[0081] The target image refers to the color monocular image of the target point captured by the camera of the seven-DOF robotic arm.

[0082] In this embodiment of the invention, when the seven-degree-of-freedom robotic arm reaches the expected position, the camera on the robotic arm is turned on and probes the target point or target object to see if it is within the 100° field of view of the camera. When the target point is within the detection range of the camera, the camera is switched to a normal camera, and the image of the target to be grasped at the current moment is collected through the camera and the position of the end of the seven-degree-of-freedom robotic arm is obtained.

[0083] It should be noted that the seven-DOF robotic arm employs a shell design combining carbon fiber and aluminum alloy, achieving lightweight construction while ensuring shell strength. The arm uses brushless DC servo motors to drive the joint movements and incorporates a harmonic reducer for transmission. The robotic arm is designed with a repeatability error of no more than 1%, therefore, an absolute encoder has been added to the existing incremental encoder, accurately reflecting joint angles and improving positional accuracy. The robotic arm can handle a load of up to 5 kg and has a spherical working radius exceeding 800 mm. The base can be mounted on any fixed support, and the end effector can be equipped with a robotic gripper for grasping and other tasks. This robotic arm comprises seven degrees of freedom: base, upper arm, forearm, and wrist joint. Odd-numbered joints are horizontal rotary joints, and even-numbered joints are vertical rotary joints, allowing free rotation within the range of [-π, π].

[0084] Step 202: Use the first feature extraction group to extract features from the target image to be captured and generate a first distance feature map. The first feature extraction group includes a zero-padding layer, a convolutional layer, a normalization layer, a ReLU activation layer, a zero-padding layer, and a ReLU activation layer connected in sequence.

[0085] In the embodiments of the present invention, see Figure 3 As shown, the first feature extraction group extracts features from the target image to be captured to obtain a first distance feature map. The first feature extraction group includes a zero-padding layer, a convolutional layer, a normalization layer, a ReLU activation layer, a zero-padding layer, and a ReLU activation layer connected in sequence.

[0086] It should be noted that if the number of convolutional layers is equal to the output feature vector of the current layer, then it is represented as follows:

[0087]

[0088] Wherein, the ReLU function is, and the output of the current layer is The convolution operation is *, and the single convolution kernel of the current convolutional layer is The offset of the convolutional layer is...

[0089] If the number of convolutional layers is n, then the following formula is the output feature vector of the pooling layer:

[0090]

[0091] The softmax activation function is f(.), and the connection weights are... The input to the current pooling layer is The input matrix summation operation is represented as (.), and the current offset is...

[0092] Step 203: Extract features from the second distance feature map using the second feature extraction group to generate the second distance feature map. The second feature extraction group includes four first feature extraction modules connected in sequence.

[0093] In the embodiments of the present invention, see Figure 3 As shown, the second feature extraction group is used to extract features from the second distance feature map to obtain the second distance feature map. The second feature extraction group includes four first feature extraction modules connected in sequence.

[0094] It should be noted that the first feature extraction module includes dense blocks and transition blocks connected in sequence. The specific processing procedure of the first feature extraction module is as follows:

[0095] S11. Extract features from the input first feature map using dense blocks to generate a second feature map. The dense block includes a dense extraction layer, a feature fusion layer, a dense extraction layer, a feature fusion layer, and a dense extraction layer connected in sequence.

[0096] Furthermore, S11 includes the following sub-steps:

[0097] S111. The first feature map of the input is extracted through a dense extraction layer to generate a first transition feature map. The density extraction layer includes a convolutional layer, a normalization layer and a ReLU activation layer connected in sequence.

[0098] In the embodiments of the present invention, see Figure 3 As shown, the first feature map is obtained by extracting features from the input first feature map through a dense extraction layer (a convolutional layer, a normalization layer, and a ReLU activation layer connected in sequence).

[0099] S112. A feature fusion layer is used to fuse the first transition feature map and the first feature map to generate a first fused feature map.

[0100] In the embodiments of the present invention, see Figure 3 As shown, a feature fusion layer is used to fuse the first transition feature map and the first feature map to generate the first fused feature map.

[0101] S113. The first fused feature map is extracted through a dense extraction layer to generate a second transition feature map.

[0102] In the embodiments of the present invention, see Figure 3 As shown, a dense extraction layer is used to extract features from the first fused feature map to generate a second transition feature map.

[0103] S114. A feature fusion layer is used to fuse the second transition feature map, the first transition feature map, and the first feature map to generate a second fused feature map.

[0104] In the embodiments of the present invention, see Figure 3 As shown, a feature fusion layer is used to fuse the second transition feature map, the first transition feature map, and the first feature map to generate the second fused feature map.

[0105] S115. The second fused feature map is extracted through a dense extraction layer to generate a third transition feature map.

[0106] In the embodiments of the present invention, see Figure 3 As shown, a dense extraction layer is used to extract features from the second fused feature map to generate a third transition feature map.

[0107] S116. A feature fusion layer is used to fuse the third transition feature map, the first transition feature map, and the first feature map to generate a second feature map.

[0108] In the embodiments of the present invention, see Figure 3 As shown, a feature fusion layer is used to fuse the third transition feature map, the first transition feature map, and the first feature map to generate the second feature map.

[0109] S12. Use transition blocks to extract features from the second feature map and generate a third feature map. The transition block includes a convolutional layer and an averaging layer connected in sequence.

[0110] In the embodiments of the present invention, see Figure 3 As shown, a transition block (a convolutional layer and a mean layer connected in sequence) is used to extract features from the second feature map to generate the third feature map.

[0111] Step 204: Use a distance detection network to perform distance detection on the second distance feature map to generate the target distance. The distance detection network includes a normalization layer, a ReLU activation layer, a convolutional layer, an upsampling layer, and four second feature extraction modules connected in sequence.

[0112] In the embodiments of the present invention, see Figure 3 As shown, a distance detection network (a normalization layer, a ReLU activation layer, a convolutional layer, an upsampling layer, and four second feature extraction modules connected in sequence) is used to perform distance detection on the second distance feature map, generate a depth map, and extract the corresponding target distance from the depth map.

[0113] It should be noted that the second feature extraction module includes a UP block and an upsampling layer connected in sequence. The specific processing procedure of the second feature extraction module is as follows:

[0114] S21. The third feature map of the input is extracted through the UP block to generate the fourth transition feature map. The UP block includes a convolutional layer, a convolutional layer and a ReLU activation layer connected in sequence.

[0115] In this embodiment of the invention, a fourth transition feature map is generated by extracting features from the input third feature map through an UP block (a series of convolutional layers and a ReLU activation layer connected in sequence).

[0116] S22. Use an upsampling layer to perform an upsampling operation on the fourth transition feature map to generate the fifth transition feature map.

[0117] In this embodiment of the invention, an upsampling layer is used to upsample the fourth transition feature map to generate the fifth transition feature map.

[0118] Step 205: Based on the preset impedance control function, the position of the robotic arm end effector and the target distance, control the seven-degree-of-freedom robotic arm to perform a grasping operation.

[0119] Furthermore, step 205 includes the following sub-steps:

[0120] S31. Determine whether the target distance is greater than or equal to the preset distance threshold.

[0121] Distance threshold refers to the working distance of a seven-degree-of-freedom robotic arm.

[0122] In this embodiment of the invention, it is determined whether the target distance reaches the working distance of the seven-degree-of-freedom robotic arm.

[0123] S32. If the target distance is less than or equal to the distance threshold, the position of the robotic arm end is input into the preset impedance control function to generate the corresponding adjustment data.

[0124] The adjustment data refers to the steady-state position, steady-state contact force, and steady-state force error of the seven-degree-of-freedom robotic arm.

[0125] In this embodiment of the invention, when the target distance reaches the working distance, the position of the robotic arm end is input into a preset impedance control function to generate the corresponding steady-state position, steady-state contact force, and steady-state force error.

[0126] S33. Adjust the steady-state position of the seven-degree-of-freedom robotic arm according to the adjustment data, and control the seven-degree-of-freedom robotic arm to move according to the target distance and then perform the grasping operation.

[0127] In this embodiment of the invention, a corresponding joint control signal is generated based on the steady-state position, steady-state contact force, and steady-state force error to adjust the joints of the robotic arm to the desired steady-state position, and then the seven-degree-of-freedom robotic arm is controlled to move according to the target distance and perform a grasping operation on the target point or target object.

[0128] S34. If the target distance is greater than the distance threshold, adjust the position of the seven-degree-of-freedom robotic arm.

[0129] In this embodiment of the invention, when the target distance is less than the working distance, the position of the seven-degree-of-freedom robotic arm is adjusted.

[0130] S35. Jump to execute the steps of controlling the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtaining the end position of the seven-degree-of-freedom robotic arm and the image of the target to be grasped.

[0131] In this embodiment of the invention, the execution jumps to step 201.

[0132] In this embodiment of the invention, when the target orientation of the seven-degree-of-freedom (DOF) robotic arm is received, the robotic arm is controlled to rotate to the target orientation. The end-effector position of the robotic arm and the image of the target to be grasped are acquired. A pre-trained distance recognition model is used to detect the distance to the target image and generate the target distance. The distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network. Based on a preset impedance control function, the end-effector position, and the target distance, the seven-degree-of-freedom robotic arm is controlled to perform the grasping operation. This addresses the technical problem that manual control of the robotic arm lacks relevant measurement methods, increasing the difficulty of robotic arm control and resulting in low efficiency in live-line collaborative operations. This application improves the accuracy of the robotic arm's grasping of objects and increases the efficiency of live-line collaborative operations by acquiring the end-effector position and the target image, combined with a preset distance recognition model and impedance control function, and adjusting the state of the robotic arm.

[0133] Please see Figure 4 , Figure 4 This is a structural block diagram of a control system for a seven-degree-of-freedom robotic arm provided in Embodiment 3 of the present invention.

[0134] The present invention provides a control system for a seven-degree-of-freedom robotic arm, comprising:

[0135] The response module 301 is used to control the seven-degree-of-freedom robotic arm to rotate to the target orientation when the target orientation of the seven-degree-of-freedom robotic arm is received, and to obtain the position of the end of the seven-degree-of-freedom robotic arm and the image of the target to be grasped.

[0136] The detection module 302 is used to perform distance detection on the target image to be captured using a pre-trained distance recognition model to generate the target distance; wherein, the distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network;

[0137] The grasping module 303 is used to control the seven-degree-of-freedom robotic arm to perform grasping operations based on a preset impedance control function, the position of the robotic arm end effector, and the target distance.

[0138] Furthermore, the detection module 302 includes:

[0139] The first extraction submodule is used to extract features from the target image to be captured using the first feature extraction group to generate a first distance feature map. The first feature extraction group includes a zero-padding layer, a convolutional layer, a normalization layer, a ReLU activation layer, a zero-padding layer, and a ReLU activation layer connected in sequence.

[0140] The second extraction submodule is used to extract features from the second distance feature map through the second feature extraction group to generate the second distance feature map. The second feature extraction group includes four first feature extraction modules connected in sequence.

[0141] The detection submodule is used to perform distance detection on the second distance feature map using a distance detection network to generate the target distance. The distance detection network includes a normalization layer, a ReLU activation layer, a convolutional layer, an upsampling layer, and four second feature extraction modules connected in sequence.

[0142] Furthermore, the capture module 302 includes:

[0143] The analysis submodule is used to determine whether the target distance is greater than or equal to a preset distance threshold;

[0144] If the target distance is less than or equal to the distance threshold, the position of the robotic arm end effector is input into a preset impedance control function to generate corresponding adjustment data;

[0145] The first execution submodule is used to adjust the steady-state position of the seven-degree-of-freedom robotic arm according to the adjustment data, and control the seven-degree-of-freedom robotic arm to move according to the target distance and then perform the grasping operation.

[0146] The second execution submodule is used to adjust the position of the seven-degree-of-freedom robotic arm if the target distance is greater than the distance threshold.

[0147] Jump to execute the steps of controlling the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtaining the position of the end effector of the seven-degree-of-freedom robotic arm and the image of the target to be grasped.

[0148] Furthermore, the first feature extraction module includes dense blocks and transition blocks connected in sequence, and the specific processing procedure of the first feature extraction module is as follows:

[0149] The first feature map is processed by dense blocks to extract features from the input feature map and generate a second feature map. The dense blocks include a dense extraction layer, a feature fusion layer, a dense extraction layer, a feature fusion layer, and a dense extraction layer connected in sequence.

[0150] A transition block is used to extract features from the second feature map to generate a third feature map. The transition block consists of a convolutional layer and an averaging layer connected in sequence.

[0151] Furthermore, the second feature extraction module includes a UP block and an upsampling layer connected in sequence. The specific processing procedure of the second feature extraction module is as follows:

[0152] The third feature map of the input is extracted by the UP block to generate the fourth transition feature map. The UP block includes a convolutional layer, a convolutional layer and a ReLU activation layer connected in sequence.

[0153] An upsampling layer is used to upsample the fourth transition feature map to generate the fifth transition feature map.

[0154] Furthermore, the step of extracting features from the input first feature map using dense blocks to generate the second feature map includes:

[0155] The first feature map is extracted by a dense extraction layer to generate a first transition feature map. The density extraction layer includes a convolutional layer, a normalization layer and a ReLU activation layer connected in sequence.

[0156] A feature fusion layer is used to fuse the first transition feature map and the first feature map to generate a first fused feature map.

[0157] The first fused feature map is used to extract features through a dense extraction layer to generate a second transition feature map;

[0158] A feature fusion layer is used to fuse the second transition feature map, the first transition feature map, and the first feature map to generate the second fused feature map.

[0159] The second fused feature map is used to extract features through a dense extraction layer to generate a third transition feature map;

[0160] A feature fusion layer is used to fuse the third transition feature map, the first transition feature map, and the first feature map to generate the second feature map.

[0161] Furthermore, the impedance control function is specifically as follows:

[0162]

[0163]

[0164]

[0165] Among them, F d Let M be the desired contact force at the end effector of the robotic arm, B be the inertia coefficient matrix, K be the damping coefficient matrix, and x be the stiffness coefficient matrix. d Let x be the desired position of the robotic arm's end effector, and let x be the position of the robotic arm's end effector. Let be the second derivative of the desired position at the end effector of the robotic arm. Let be the second derivative at the end effector position of the robotic arm. Let be the first derivative of the desired position at the end of the robotic arm. Let x be the first derivative of the position of the robotic arm's end effector. e B represents the target position in the inertial coordinate system. e Let K be the environmental damping matrix. e Let x be the environmental stiffness matrix. SS F represents the steady-state position of the robotic arm's end effector. SS For steady-state contact force, E fss This represents the steady-state force error.

[0166] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0167] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and 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 an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0168] 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.

[0169] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A control method for a seven-degree-of-freedom robotic arm, characterized in that, The method includes: When the target orientation of the seven-degree-of-freedom robotic arm is received, the seven-degree-of-freedom robotic arm is controlled to rotate to the target orientation, and the position of the end effector of the seven-degree-of-freedom robotic arm and the image of the target to be grasped are obtained. A pre-trained distance recognition model is used to perform distance detection on the target image to be captured, generating the target distance; wherein, the distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network; Based on a preset impedance control function, the end position of the robotic arm, and the target distance, the seven-degree-of-freedom robotic arm is controlled to perform a grasping operation. The step of using a pre-trained distance recognition model to perform distance detection on the target image to be captured and generating the target distance includes: The first feature extraction group is used to extract features from the target image to be captured to generate a first distance feature map. The first feature extraction group includes a zero-padding layer, a convolutional layer, a normalization layer, a ReLU activation layer, a zero-padding layer, and a ReLU activation layer connected in sequence. The second feature extraction group extracts features from the first distance feature map to generate a second distance feature map. The second feature extraction group includes four first feature extraction modules connected in sequence. The distance detection network is used to perform distance detection on the second distance feature map to generate the target distance. The distance detection network includes a normalization layer, a ReLU activation layer, a convolutional layer, an upsampling layer and four second feature extraction modules connected in sequence. The step of controlling the seven-degree-of-freedom robotic arm to perform a grasping operation based on a preset impedance control function, the end-effector position, and the target distance includes: Determine whether the target distance is less than or equal to a preset distance threshold; If the target distance is less than or equal to the distance threshold, the position of the robotic arm end is input into a preset impedance control function to generate corresponding adjustment data; The steady-state position of the seven-degree-of-freedom robotic arm is adjusted according to the adjustment data, and the seven-degree-of-freedom robotic arm is controlled to move according to the target distance and then perform a grasping operation; If the target distance is greater than the distance threshold, then adjust the position of the seven-degree-of-freedom robotic arm; Jump to execute the steps of controlling the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtaining the end-effector position of the seven-degree-of-freedom robotic arm and the image of the target to be grasped; The impedance control function is specifically as follows: ； ； ； in, The desired contact force at the end effector of the robotic arm. The inertia coefficient matrix, Here is the damping coefficient matrix. This is the stiffness coefficient matrix. The desired position of the robotic arm's end effector. This is the position of the robotic arm's end effector. Let be the second derivative of the desired position at the end effector of the robotic arm. Let be the second derivative at the end effector position of the robotic arm. Let be the first derivative of the desired position at the end of the robotic arm. The first derivative of the position of the robotic arm's end effector. The target position in the inertial coordinate system. Here is the environmental damping matrix. Here is the environmental stiffness matrix. This represents the steady-state position of the robotic arm's end effector. For steady-state contact force, This represents the steady-state force error.

2. The control method for a seven-degree-of-freedom robotic arm according to claim 1, characterized in that, The first feature extraction module includes dense blocks and transition blocks connected in sequence. The specific processing procedure of the first feature extraction module is as follows: The first feature map is processed by dense blocks to extract features from the input feature map and generate a second feature map. The dense blocks include a dense extraction layer, a feature fusion layer, a dense extraction layer, a feature fusion layer, and a dense extraction layer connected in sequence. The second feature map is used to extract features by a transition block to generate a third feature map. The transition block includes a convolutional layer and an averaging layer connected in sequence.

3. The control method for a seven-degree-of-freedom robotic arm according to claim 2, characterized in that, The step of extracting features from the input first feature map using dense blocks to generate a second feature map includes: The first feature map is extracted from the input feature map by a dense extraction layer to generate a first transition feature map. The dense extraction layer includes a convolutional layer, a normalization layer and a ReLU activation layer connected in sequence. A feature fusion layer is used to fuse the first transition feature map and the first feature map to generate a first fused feature map. The first fused feature map is used to extract features through a dense extraction layer to generate a second transition feature map; A feature fusion layer is used to fuse the second transition feature map, the first transition feature map, and the first feature map to generate a second fused feature map. A third transition feature map is generated by extracting features from the second fused feature map through a dense extraction layer. A feature fusion layer is used to fuse the third transition feature map, the first transition feature map, and the first feature map to generate a second feature map.

4. The control method for a seven-degree-of-freedom robotic arm according to claim 1, characterized in that, The second feature extraction module includes a UP block and an upsampling layer connected in sequence. The specific processing procedure of the second feature extraction module is as follows: The third feature map of the input is extracted by the UP block to generate the fourth transition feature map. The UP block includes a convolutional layer, a convolutional layer and a ReLU activation layer connected in sequence. An upsampling layer is used to upsample the fourth transition feature map to generate a fifth transition feature map.

5. A control system for a seven-degree-of-freedom robotic arm, characterized in that, include: The response module is used to control the seven-degree-of-freedom robotic arm to rotate to the target orientation when the target orientation is received, and to obtain the end position of the robotic arm and the image of the target to be grasped. The detection module is used to perform distance detection on the target image to be captured using a pre-trained distance recognition model to generate the target distance; wherein, the distance recognition model includes a first feature extraction group, a second feature extraction group, and a distance detection network; The grasping module is used to control the seven-degree-of-freedom robotic arm to perform grasping operations based on a preset impedance control function, the end position of the robotic arm, and the target distance; The detection module includes: The first extraction submodule is used to extract features from the target image to be captured using the first feature extraction group to generate a first distance feature map, wherein the first feature extraction group includes a zero-padding layer, a convolutional layer, a normalization layer, a ReLU activation layer, a zero-padding layer, and a ReLU activation layer connected in sequence. The second extraction submodule is used to extract features from the first distance feature map through the second feature extraction group to generate a second distance feature map, wherein the second feature extraction group includes four first feature extraction modules connected in sequence. The detection submodule is used to perform distance detection on the second distance feature map using the distance detection network to generate the target distance, wherein the distance detection network includes a normalization layer, a ReLU activation layer, a convolutional layer, an upsampling layer and four second feature extraction modules connected in sequence; The crawling module includes: The analysis submodule is used to determine whether the target distance is less than or equal to a preset distance threshold; If the target distance is less than or equal to the distance threshold, the position of the robotic arm end is input into a preset impedance control function to generate corresponding adjustment data; The first execution submodule is used to adjust the steady-state position of the seven-degree-of-freedom robotic arm according to the adjustment data, and control the seven-degree-of-freedom robotic arm to move according to the target distance and then perform a grasping operation; The second execution submodule is used to adjust the position of the seven-degree-of-freedom robotic arm if the target distance is greater than a preset distance threshold. The steps of adjusting and executing the control of the seven-degree-of-freedom robotic arm to rotate to the target orientation, and obtaining the end-effector position of the seven-degree-of-freedom robotic arm and the image of the target to be grasped; The impedance control function is specifically as follows: ； ； ； in, The desired contact force at the end effector of the robotic arm. The inertia coefficient matrix, Here is the damping coefficient matrix. This is the stiffness coefficient matrix. The desired position of the robotic arm's end effector. This is the position of the robotic arm's end effector. Let be the second derivative of the desired position at the end effector of the robotic arm. Let be the second derivative at the end effector position of the robotic arm. Let be the first derivative of the desired position at the end of the robotic arm. The first derivative of the position of the robotic arm's end effector. The target position in the inertial coordinate system. Here is the environmental damping matrix. Here is the environmental stiffness matrix. This represents the steady-state position of the robotic arm's end effector. For steady-state contact force, This represents the steady-state force error.

Images