A suction cup gripping device and control method based on tactile perception

Abstract

This invention discloses a suction cup grasping device and control method based on tactile perception. The device includes a connecting air pipe, a multi-layer vacuum suction cup, and a tactile sensor. The connecting air pipe has an air inlet and multiple vents, which are connected to the multi-layer vacuum suction cup. The multi-layer vacuum suction cup has a suction cup base connected to the connecting air pipe. A gas channel is provided within the suction cup base, and the vents connect to the inner cavity of the multi-layer vacuum suction cup through the gas channel. The tactile sensor includes a camera connecting pipe, a connecting pipe, an LED connecting pipe, and a soft contact head. The connecting air pipe and the camera connecting pipe are coaxially connected. The camera connecting pipe passes through the suction cup base and houses a camera system. The camera connecting pipe, the connecting pipe, and the LED connecting pipe are connected sequentially. An LED is located inside the LED connecting pipe, which is connected to the soft contact head. This invention can automatically assess the success of grasping based on tactile perception and actively adjust the grasping direction, improving the stability and efficiency of robot grasping.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically to a suction cup gripping device and control method based on tactile perception. Background Technology

[0002] In current production processes, the handling and installation of small, smooth materials, such as plastic, glass, and ceramic products, are common. Suction cup systems are typically used for this purpose. Existing suction cup gripping systems usually use the suction mechanism to directly hold the material for handling and installation. However, they cannot automatically detect and evaluate the gripping status. If the suction cup does not hold the object firmly and lifts it directly, the material is likely to fall, causing damage and even threatening the personal safety of human operators, which is not conducive to handling and installation.

[0003] For example, a suction cup gripping system with adjustable angle and lockability, disclosed in patent publication number CN209160915U, uses a ball joint to adjust the gripping direction of the suction cup gripping system, making it applicable to gripping materials of various shapes. It is equipped with a locking device, with its two ends connected to the suction cup locking cup and the connecting rod, respectively. This allows the entire suction cup gripping system to become a rigid whole under the fixing of the locking device after the material is gripped. However, it does not use an appropriate method to automatically detect and evaluate the gripping state. Even if the gripping direction can be adjusted, the direction may not be the optimal gripping direction.

[0004] The outdoor facility installation robotic arm, with patent publication number CN114012711A, is based on a tactile gripping tightness design. This allows the robotic arm to adjust the gripping distance and incorporates a structure to maintain equipment stability, ensuring the device remains stationary during gripping and guaranteeing the accuracy of cargo handling. Through its tactile automatic mechanical gripper, it can achieve intelligent and automatic cargo handling. However, it only uses a tactile sensor to detect whether an object has been grasped; it cannot actively sense gripping stability or adjust the gripping direction.

[0005] Therefore, how to automatically assess whether a grasping operation can be successful and make adjustments accordingly to improve the stability and safety of robot grasping is a technical challenge that urgently needs to be solved. Summary of the Invention

[0006] To overcome the defects and shortcomings of existing technologies, this invention provides a suction cup gripping device and control method based on tactile perception. This invention automatically detects the contact state between the suction cup and the material based on feedback from tactile sensors, eliminating the need for manual attachment of the suction cup to the material surface, thus achieving highly automated gripping. It can also automatically assess whether the gripping is successful, improving the stability and safety of robot gripping. Furthermore, it introduces a tactile perception-based adjustment mechanism, which can actively adjust the gripping direction based on tactile feedback, quickly and automatically generating a stable gripping configuration, thereby improving the success rate and efficiency of robot gripping.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The present invention provides a suction cup gripping device based on tactile perception, comprising: a connecting air tube, a multi-layer vacuum suction cup, and a tactile sensor, wherein the connecting air tube is connected to the multi-layer vacuum suction cup;

[0009] The connecting air pipe is provided with an air inlet and multiple air vents. The air vents are connected to a multi-layer vacuum suction cup. The air inlet is used to pass through multiple connecting hoses of an external vacuum pump. The connecting hoses are connected to the corresponding air vents.

[0010] The multi-layer vacuum suction cup is provided with a suction cup base, which is connected to a connecting air pipe. The suction cup base is provided with a gas channel, which is connected to a vent hole. The vent hole is connected to the inner cavity of the multi-layer vacuum suction cup through the gas channel.

[0011] The tactile sensor is located inside the connecting air tube and the multi-layer vacuum suction cup. The tactile sensor includes a camera connecting tube, a connecting tube, an LED connecting tube, and a soft contact head. The connecting air tube and the camera connecting tube are coaxially connected. The camera connecting tube passes through the suction cup base and contains a camera system. The camera system is used to acquire tactile images of the deformation caused by the soft contact head after contact with an object. The camera connecting tube is connected to the connecting tube, and the connecting tube is connected to the LED connecting tube. The inner edge of the LED connecting tube has a ring of uniformly distributed LEDs for supplementing light to the camera system. The LED connecting tube is connected to the soft contact head, and the soft contact head contains multiple arrayed marked synapses.

[0012] As a preferred technical solution, the soft contact head is further provided with a mounting buckle, and the soft contact head is connected to the LED connecting tube through the mounting buckle.

[0013] As a preferred technical solution, the camera connecting tube is provided with an external thread, the connecting air tube is provided with an internal thread, and the camera connecting tube and the connecting air tube are connected by threads.

[0014] As a preferred technical solution, the interior of the soft contact head is filled with transparent gel.

[0015] As a preferred technical solution, the top of the soft contact head is sealed with a transparent acrylic plate.

[0016] This invention also provides a control method for a suction cup gripping device based on tactile perception, comprising the following steps:

[0017] The tactile sensor in the multi-layer vacuum suction cup is controlled to make perpendicular contact with the surface of the material to be grasped. The interval and range of the tilt angle change and the interval and range of the contact depth change are set. The tactile sensor is controlled to contact and rotate on the object surface. The contact position between the tactile sensor and the material surface is changed. Tactile images and corresponding tilt angles and contact depths are obtained. The datasets are divided into training datasets and test datasets. Tactile image preprocessing is performed.

[0018] Using the preprocessed tactile image as input data and the corresponding tilt angle and contact depth as output data labels, a convolutional neural network is trained. By minimizing the mean square error between the predicted output and the data labels, the convolutional neural network learns the nonlinear mapping between the input and output, thus obtaining the trained convolutional neural network.

[0019] Input the test dataset, estimate the tilt angle and contact depth between the tactile sensor and the object surface corresponding to different tactile images based on the trained convolutional neural network, set the tilt angle threshold and contact depth threshold, and use the tilt angle not greater than the set tilt angle threshold and the contact depth not less than the set contact depth threshold as the evaluation criteria for grasping stability.

[0020] After estimating the position of the material to be grasped based on visual positioning technology, the multi-layer vacuum suction cup is controlled to contact the material surface vertically downward. The camera in the camera connecting tube continuously captures tactile images. When the tactile image changes, that is, when the difference between the current tactile image and the initial tactile image before contact with the material is not 0, it indicates that contact has occurred. Then, the multi-layer vacuum suction cup is controlled to increase the contact depth vertically downward to the first depth value d1.

[0021] The gripping stability assessment standard evaluates whether the gripping can be stable under the current relative pose of the tactile sensor and the material surface. If it can, the air in the multi-layer vacuum suction cup is extracted through the vent, creating a negative pressure inside the multi-layer vacuum suction cup and generating suction force to start gripping. If it cannot, the multi-layer vacuum suction cup is controlled vertically upward to reduce the contact depth to the second depth value d2, keeping the contact position between the tactile sensor and the surface of the material to be gripped unchanged. It is then rotated in the direction that reduces the tilt angle, so that the central axis of the multi-layer vacuum suction cup tends to be parallel to the normal of the material surface. The multi-layer vacuum suction cup is then controlled vertically downward to increase the contact depth to the first depth value d1. The relative pose of the tactile sensor and the material surface is repeatedly adjusted until the gripping stability assessment standard is met, and then gripping is performed.

[0022] As a preferred technical solution, the tactile sensor in the multi-layer vacuum suction cup is controlled to make perpendicular contact with the surface of the material to be grasped. At this time, the tilt angle and contact depth are both 0. The tilt angle is varied at intervals and ranges of 5°, -60° to 60°, and the contact depth is varied at intervals and ranges of 1mm and 0 to 15mm.

[0023] As a preferred technical solution, the tilt angle refers to the angle between the central axis of the tactile sensor and the normal to the material surface. The tilt angle is a vector, including the pitch angle and the roll angle. The pitch angle refers to the angle of left and right rotation of the tactile sensor, and the roll angle refers to the angle of front and back rotation of the tactile sensor. The contact depth refers to the deformation distance of the soft contact head along the central axis of the tactile sensor after contacting the object. The positional changes of the end effector of the robotic arm and the tactile sensor in the coordinate system of the robotic arm body are calculated using the dimensions of the multi-layer vacuum suction cup and the positive kinematics of the robotic arm.

[0024] As a preferred technical solution, the neural network structure adopts the PoseNet CNN network.

[0025] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0026] (1) The present invention automatically detects the contact state between the suction cup and the material based on the feedback of the tactile sensor, which solves the problem that traditional suction cup devices cannot automatically determine whether they are in contact with the material, realizes highly automated grasping, can automatically evaluate the stability of suction cup grasping, and improves the stability and safety of robot grasping.

[0027] (2) The present invention adopts a tactile sensing suction cup gripping direction adjustment method, which can quickly optimize contact and force to achieve the purpose of gently gripping materials, thereby improving the success rate and efficiency of robot gripping. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the overall structure of the suction cup gripping device based on tactile perception according to the present invention.

[0029] Figure 2 This is a side cross-sectional view of the suction cup gripping device based on tactile perception of the present invention;

[0030] Figure 3 This is a schematic diagram of the structure of the soft contact head of the present invention;

[0031] Figure 4 This is a schematic diagram of the control method for the suction cup gripping device based on tactile perception according to the present invention.

[0032] Figure 5 This is a schematic diagram of the grasping direction adjustment based on tactile perception according to the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0034] Example

[0035] like Figure 1 , Figure 2 As shown, this embodiment provides a suction cup gripping device based on tactile perception, including: a connecting air tube 100, a multi-layer vacuum suction cup 200 and a tactile sensor 300, wherein the connecting air tube 100 is connected to the multi-layer vacuum suction cup 200.

[0036] In this embodiment, the connecting air pipe 100 is provided with an air inlet 110 and multiple air vents 120. Preferably, there are eight air vents 120. Multiple connecting hoses of the external vacuum pump pass through the air inlet 110. Branches of the connecting hoses are connected to the multiple evenly distributed air vents 120. The air vents 120 are connected to the inner cavity 200 of the multi-layer vacuum suction cup. The air vents 120 are used for gas circulation, which facilitates the external vacuum pump to evenly extract the air inside the multi-layer vacuum suction cup.

[0037] In this embodiment, the connection points of the external vacuum pump's connecting hoses are sealed with rubber sealing sleeves to prevent air leakage and improve adsorption stability.

[0038] In this embodiment, the multi-layer vacuum suction cup 200 is provided with a suction cup base 210, which is connected to the connecting air pipe 100. The suction cup base 210 is provided with a gas channel, which is connected to a vent hole 120. The vent hole is connected to the inner cavity of the multi-layer vacuum suction cup through the gas channel.

[0039] In this embodiment, the tactile sensor can adopt the structure of the open-source tactip tactile sensor from the Bristol Robotics Laboratory. The structure of the camera connecting tube 310 can be modified according to the size and structure of the connecting air tube 100 and the multi-layer vacuum suction cup 200. For example, threads can be added to the outside of the camera connecting tube 310 to connect with the connecting air tube 100.

[0040] In this embodiment, the tactile sensor 300 is disposed inside the connecting air tube 100 and the multilayer vacuum suction cup 200. The tactile sensor 300 includes a camera connecting tube 310, a connecting tube 320, an LED connecting tube 330, and a soft contact head 340.

[0041] The connecting tube 100 is coaxially connected to the camera connecting tube 310, which passes through the suction cup base 210. The camera connecting tube 310 has a small camera system inside for acquiring tactile images. The camera connecting tube 310 is connected to the connecting tube 320, which is connected to the LED connecting tube 330. The inner edge of the LED connecting tube 330 has small LEDs evenly distributed in a ring to provide supplementary lighting for the small camera system in order to obtain high-quality tactile images. The LED connecting tube 330 is connected to the soft contact head 340.

[0042] like Figure 3 As shown, the bottom of the LED connector tube 330 and the top of the flexible contact head 340 have mutually offset mounting buckles 341 structures, and the flexible contact head 340 is connected to the LED connector tube 330 through the mounting buckles 341.

[0043] like Figure 3 As shown, the soft contact head 340 has 127 arrayed marked synapses 342 distributed inside, and the soft contact head is filled with transparent gel. The top of the soft contact head 340 is sealed with a transparent acrylic plate. The soft contact head 340 is preferably a 3D printed soft gel.

[0044] In this embodiment, a small camera system inside the camera connector tube acquires tactile images, specifically images of the deformation inside the soft contact head. When the soft contact head comes into contact with an object, it deforms. This deformation is reflected in the displacement changes of the marker synapses inside the soft contact head (the displacement changes of the marker synapses can amplify the deformation of the soft contact head). The deformation of the soft contact head will vary depending on the contact depth and direction with the object, resulting in different tactile images captured by the small camera system. Tactile information is obtained by analyzing the displacement changes of the marker synapses. The relative pose of the tactile sensor and the material surface is estimated based on the deformation images of the marker synapses, thereby evaluating the gripping stability of the multilayer vacuum suction cup on the object.

[0045] like Figure 4 As shown, this embodiment also provides a control method for a suction cup gripping device based on tactile perception, including the following steps:

[0046] S1: Obtain the training dataset and test dataset, train the convolutional neural network based on the training dataset, and estimate the relative pose of the tactile sensor and the material surface;

[0047] In this embodiment, a multi-layer vacuum suction cup is mounted on the end of a robotic arm via a flange. The robotic arm first controls the tactile sensor in the multi-layer vacuum suction cup to make perpendicular contact with the material surface with manual assistance. At this time, the tilt angle θ and the contact depth d are both 0. Then, the tilt angle θ is varied at intervals and ranges of 5°, -60° to 60°, and the contact depth d is varied at intervals and ranges of 1mm and 0 to 15mm, respectively. The robotic arm automatically controls the tactile sensor to contact and rotate on the object surface, while recording the tactile image, tilt angle θ, and contact depth d as training and testing data. In order to obtain sufficient data, the contact position between the tactile sensor and the material surface can be changed, and the above steps can be repeated.

[0048] In this embodiment, an image preprocessing step is also included, which involves scaling, Gaussian filtering, and binarization of the tactile image.

[0049] The convolutional neural network is trained based on a training dataset. Specifically, it uses preprocessed tactile images as input data and corresponding tilt angles θ and contact depths d as data labels (output). By minimizing the mean square error between the predicted output and the data labels, the convolutional neural network learns a nonlinear mapping between the input and output. Based on the trained convolutional neural network, tactile images are identified, and the tilt angles θ and contact depths d of the tactile sensor and the object surface corresponding to different tactile images are estimated. Tilt angle thresholds and contact depth thresholds are set. The tilt angle θ not exceeding the set tilt angle threshold and the contact depth d not less than the set contact depth threshold are used as the grasping stability evaluation criteria. Based on the grasping stability evaluation criteria, the stability of grasping under the current relative pose of the tactile sensor and the material surface is evaluated (i.e., when the multi-layer vacuum suction cup is in perpendicular contact with the material surface and pressed down a certain distance before air is pumped out, the adsorption of the object will be the most stable). When the tilt angle is greater than the set tilt angle threshold, the contact depth is set to 0, indicating that the grasping under this configuration is very unstable.

[0050] In this embodiment, the tilt angle θ refers to the angle between the central axis of the tactile sensor and the normal to the material surface. θ is a vector that includes the pitch angle and the roll angle (excluding the yaw angle of rotation around the central axis of the tactile sensor). The pitch angle refers to the angle of left-right rotation of the tactile sensor, and the roll angle refers to the angle of front-back rotation of the tactile sensor. The contact depth d refers to the deformation distance of the soft contact head along the central axis of the tactile sensor after contacting the object. It can be obtained by calculating the pose changes of the end effector of the robotic arm and the tactile sensor in the coordinate system of the robotic arm body by utilizing the size of the multi-layer vacuum suction cup and the positive kinematics of the robotic arm.

[0051] In this embodiment, the preferred neural network architecture is the PoseNet CNN network;

[0052] S2: The robot controls the multi-layer vacuum suction cups to contact the material, determining whether the tactile sensor is in contact with the material surface. If it is determined that the tactile sensor is not in contact with the material surface, the robot continues to control the multi-layer vacuum suction cups to contact the material. Figure 5 As shown, if it is determined that the tactile sensor is in contact with the material surface, the contact depth is increased to the first depth value d1, and a tactile image is further acquired;

[0053] In this embodiment, after estimating the approximate position of the material based on visual positioning technology (such as YOLO), the robotic arm first controls the multi-layer vacuum suction cup to make vertical downward contact with the material surface. The camera in the camera connecting tube continuously captures tactile images. When the tactile image changes, that is, when the difference between the current tactile image and the initial tactile image before contact with the material is not 0, it indicates that contact has occurred. The robotic arm then further increases the contact depth to the first depth value d1, the multi-layer vacuum suction cup moves downward as a whole, the deformation of the soft contact head increases, and the sealed space formed by the multi-layer vacuum suction cup and the material is compressed and reduced.

[0054] S3: Based on the tactile image, determine whether the grip is stable. If it is determined that the grip is not stable, reduce the contact depth to the second depth value d2 and adjust the gripping direction of the multi-layer vacuum suction cup according to the tactile perception. If it is determined that the grip is stable, the external vacuum pump works, the multi-layer vacuum suction cup generates negative pressure, and the robot performs the gripping action.

[0055] In this embodiment, the tilt angle θ and contact depth d are estimated from the acquired tactile image using the convolutional neural network. Based on the grasping stability evaluation criteria that the tilt angle θ is not greater than a set tilt angle threshold and the contact depth d is not less than a set contact depth threshold, it is determined whether the multi-layer vacuum suction cup can firmly grasp the material. If it can, the external vacuum pump starts working, the air inside the multi-layer vacuum suction cup is extracted through the vent, a negative pressure is formed inside the multi-layer vacuum suction cup, thereby generating suction force, and then the robot begins to grasp. If it cannot, this embodiment introduces a tactile sensing-based adjustment mechanism. When the material has curvature, a suction cup without tactile sensing function is difficult to grasp successfully. For example, if the suction cup is adsorbing the material vertically, further increasing the suction force (pressure) may not produce a stable grasping configuration. Therefore, after stopping the external vacuum pump, if... Figure 5 As shown, to avoid damaging the tactile sensor, the robotic arm first reduces the contact depth to a second depth value d2 (d2 is less than d1), the multi-layer vacuum suction cup moves upward as a whole, the deformation of the soft contact head decreases, and then keeps the contact position between the tactile sensor and the material surface unchanged, rotates in the direction that reduces the above-mentioned tilt angle θ, so that the central axis of the multi-layer vacuum suction cup tends to be parallel to the normal of the material surface, and then further increases the contact depth to the first depth value d1, and repeats the following steps until it is determined that the multi-layer vacuum suction cup can hold the material firmly, and then the external vacuum pump starts working again, automatically and quickly generating a stable gripping configuration.

[0056] In summary, by introducing tactile sensing information to automatically detect and evaluate the adsorption state of multi-layer vacuum suction cups and actively change the adsorption direction of the multi-layer vacuum suction cups, a stable gripping configuration can be automatically and quickly generated, optimizing contact and force to achieve the purpose of gently gripping materials, thereby improving the success rate and efficiency of robot gripping.

[0057] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A control method for a suction cup gripping device based on tactile perception, characterized in that, A suction cup gripping device based on tactile perception is provided, including: a connecting air tube, a multi-layer vacuum suction cup and a tactile sensor, wherein the connecting air tube is connected to the multi-layer vacuum suction cup; The connecting air pipe is provided with an air inlet and multiple air vents. The air vents are connected to a multi-layer vacuum suction cup. The air inlet is used to pass through multiple connecting hoses of an external vacuum pump. The connecting hoses are connected to the corresponding air vents. The multi-layer vacuum suction cup is provided with a suction cup base, which is connected to a connecting air pipe. The suction cup base is provided with a gas channel, which is connected to a vent hole. The vent hole is connected to the inner cavity of the multi-layer vacuum suction cup through the gas channel. The tactile sensor is located inside the connecting air tube and the multi-layer vacuum suction cup. The tactile sensor includes a camera connecting tube, a connecting tube, an LED connecting tube, and a soft contact head. The connecting air tube and the camera connecting tube are coaxially connected. The camera connecting tube passes through the suction cup base and contains a camera system. The camera system is used to acquire tactile images of the deformation caused by the soft contact head after contact with an object. The camera connecting tube is connected to the connecting tube, and the connecting tube is connected to the LED connecting tube. The inner edge of the LED connecting tube has a ring of uniformly distributed LEDs for supplementing light to the camera system. The LED connecting tube is connected to the soft contact head, and the soft contact head contains multiple arrayed marked synapses. Includes the following steps: The tactile sensor in the multi-layer vacuum suction cup is controlled to make perpendicular contact with the surface of the material to be grasped. The interval and range of the tilt angle change and the interval and range of the contact depth change are set. The tactile sensor is controlled to contact and rotate on the object surface. The contact position between the tactile sensor and the material surface is changed. Tactile images and corresponding tilt angles and contact depths are obtained. The datasets are divided into training datasets and test datasets. Tactile image preprocessing is performed. Using the preprocessed tactile image as input data and the corresponding tilt angle and contact depth as output data labels, a convolutional neural network is trained. By minimizing the mean square error between the predicted output and the data labels, the convolutional neural network learns the nonlinear mapping between the input and output, thus obtaining the trained convolutional neural network. Input the test dataset, estimate the tilt angle and contact depth between the tactile sensor and the object surface corresponding to different tactile images based on the trained convolutional neural network, set the tilt angle threshold and contact depth threshold, and use the tilt angle not greater than the set tilt angle threshold and the contact depth not less than the set contact depth threshold as the evaluation criteria for grasping stability. After estimating the position of the material to be grasped based on visual positioning technology, the multi-layer vacuum suction cup is controlled to contact the material surface vertically downward. The camera in the camera connecting tube continuously captures tactile images. When the tactile image changes, that is, when the difference between the current tactile image and the initial tactile image before contact with the material is not 0, it indicates that contact has occurred. Then, the multi-layer vacuum suction cup is controlled to increase the contact depth vertically downward to the first depth value d1. The gripping stability assessment standard evaluates whether the gripping can be stable under the current relative pose of the tactile sensor and the material surface. If it can, the air in the multi-layer vacuum suction cup is extracted through the vent, creating a negative pressure inside the multi-layer vacuum suction cup and generating suction force to start gripping. If it cannot, the multi-layer vacuum suction cup is controlled vertically upward to reduce the contact depth to the second depth value d2, keeping the contact position between the tactile sensor and the surface of the material to be gripped unchanged. It is then rotated in the direction that reduces the tilt angle, so that the central axis of the multi-layer vacuum suction cup tends to be parallel to the normal of the material surface. The multi-layer vacuum suction cup is then controlled vertically downward to increase the contact depth to the first depth value d1. The relative pose of the tactile sensor and the material surface is repeatedly adjusted until the gripping stability assessment standard is met, and then gripping is performed.

2. The control method for the suction cup gripping device based on tactile perception according to claim 1, characterized in that, The tactile sensor in the multi-layer vacuum suction cup is controlled to make perpendicular contact with the surface of the material to be grasped. At this time, the tilt angle and contact depth are both 0. The tilt angle is varied at intervals and ranges of 5°, -60° to 60°, and the contact depth is varied at intervals and ranges of 1mm and 0 to 15mm.

3. The control method for the suction cup gripping device based on tactile perception according to claim 1, characterized in that, The tilt angle refers to the angle between the central axis of the tactile sensor and the normal to the material surface. The tilt angle is a vector, including the pitch angle and the roll angle. The pitch angle is the angle at which the tactile sensor rotates left and right, and the roll angle is the angle at which the tactile sensor rotates forward and backward. The contact depth refers to the deformation distance of the soft contact head along the central axis of the tactile sensor after contacting the object. It is obtained by calculating the pose changes of the end effector of the robotic arm and the tactile sensor in the coordinate system of the robotic arm body using the dimensions of the multi-layer vacuum suction cup and the positive kinematics of the robotic arm.

4. The control method for the suction cup gripping device based on tactile perception according to claim 1, characterized in that, The neural network structure uses the PoseNet CNN network.

5. The control method for the suction cup gripping device based on tactile perception according to claim 1, characterized in that, The flexible contact head is also equipped with a mounting clip, which connects the flexible contact head to the LED connecting tube.

6. The control method for the suction cup gripping device based on tactile perception according to claim 1, characterized in that, The camera connecting tube has an external thread, and the connecting air tube has an internal thread. The camera connecting tube and the connecting air tube are connected by threads.

7. The control method for the suction cup gripping device based on tactile perception according to claim 1, characterized in that, The soft contact head is filled with transparent gel.

8. The control method for the suction cup gripping device based on tactile perception according to claim 1, characterized in that, The top of the soft contact head is sealed with a transparent acrylic sheet.

Images