CN116945166A - An optimization method for robot re-grasp based on tactile element slip feature feedback - Google Patents

Abstract

The invention belongs to the technical field of intelligent grabbing of robots, and relates to a robot re-grabbing optimization method based on haptic primitive sliding characteristic feedback, which comprises the following steps: step 1, acquiring touch mode information, namely sensing a touch primitive of a contact area by using an optical touch sensor arranged on a mechanical arm; step 2: based on the obtained haptic mode information, carrying out elliptical parameter fitting; step 3: constructing a test grabbing coordinate system of the mechanical arm; step 4: performing trial grabbing on the object to enable the object to slide, calculating a touch perception sliding vector of the object, and estimating the pose change of the object during the sliding; step 5: and correspondingly optimizing and adjusting sliding of different intensity degrees of the object in the process of trial grabbing, and reconstructing the re-grabbing pose of the mechanical arm. The invention solves the problem that the existing grabbing adjustment method depends on trial and error and cannot fully utilize the tactile mode information.

Description

A robot re-grasp optimization method based on tactile element slip feature feedback

Technical field

The invention belongs to the field of robot intelligent grasping technology, and relates to a robot re-grasp optimization method based on tactile element slip feature feedback.

Background technique

Vision-based grasping detection methods usually rely on the size, texture, color and other information of the object to obtain the grasping pose, but cannot perceive the contact surface characteristics, center of mass pose change information and contact force changes of the grasped object, as well as the grasping Information such as the relative motion status of the object and the gripper during the picking process makes it impossible to evaluate and provide feedback on the existing grabbing configuration. Therefore, when unstable grasping occurs, the grasping pose derived from visual reasoning needs to be further optimized to meet the requirements of stability, safety, and efficiency.

Contents of the invention

In order to solve the above technical problems existing in the prior art, the present invention proposes a robot re-grasp optimization method based on tactile element slip feature feedback, based on Canny edge detection and least squares ellipse fitting, to achieve optical Extract the tactile primitives sensed by the tactile sensor, and further determine the slip characteristics of the tactile primitives. According to the severity of the slip of the object during the trial grasp, it is divided into three types: initial slip and local slip. and comprehensive slipping. By combining the visual grasping detection map, tactile feedback features and visual-touch fusion quality analysis model method, a corresponding adaptive optimization adjustment strategy is developed to stably reconstruct the grasping pose. The specific technical solution is as follows:

A robot re-grasp optimization method based on tactile element slip feature feedback, including the following steps:

Step 1: Use the optical tactile sensor installed on the robotic arm to obtain tactile modal information, that is, sense and acquire the tactile primitives in the contact area;

Step 2: Perform ellipse parameter fitting based on the acquired tactile modal information;

Step 3: Construct the trial grasping coordinate system of the robotic arm;

Step 4: Trial grasp the object to cause the object to slip, calculate the tactile sensing sliding vector of the object, and infer the posture changes of the object during the sliding period;

Step 5: Make corresponding optimization adjustments to the different degrees of slippage that occur during the trial grasping process of the object, and reconstruct the re-grasping posture of the robotic arm.

Further, the step 1 is specifically: using a vision-based tactile sensor to determine whether the gripper of the robotic arm is in contact with the object through changes in the contact image. If so, obtain the sensor contact image; if not, obtain the non-contact tactile initial state. Image; obtain the area, position, and direction of object contact; obtain the movement of the contact part on the sensor surface through continuous video images;

Wherein, the tactile sensor estimates the contact area of the object through single-point and multi-point features, and realizes the perception of the contact area through the extraction and matching of geometric features, thereby obtaining a difference map from the initial image;

The characteristics of the elastomer material and the slightly convex surface are used to show the characteristics of an elliptical contact area or a circular contact area on the imaging image. The contact area is described through the tactile sensing modal elliptical primitive extraction, and the contact area is uniformly expressed as A The format of (p, a, b, θ); where, p (x, y) is the center coordinate of the contact ellipse in the tactile sensor image coordinate system, a is the semi-major axis length of the contact ellipse, and b is the semi-minor axis of the contact ellipse. Axis length, θ is the deflection angle of the contact ellipse, taking the long axis of the ellipse as a reference, its value is the angle of the long axis deflection in the horizontal direction, the value range is [-π/2, π/2]; when the shape of the contact area is In the case of a circle, the length of the major axis and the length of the minor axis are equal, that is, a=b, and θ is any value within [-π/2, π/2].

Further, step 2 specifically includes:

Step 2.1: After performing grayscale and morphological processing on the difference image, use the Canny edge detection operator to perform edge detection and extraction on the contact area;

Step 2.2: Use the geometric ellipse fitting method based on the least squares method to fit the detected edge image.

Further, step 3 specifically includes:

Step 3.1: Based on the current grab configuration map G (x, y, z, r _x , r _y , r _z , W) deduced by the grab detection algorithm, establish a grab with the grab center coordinates as the origin. Coordinate system o-xyz;

Step 3.2: Install two face-to-face tactile sensors on the working surface of the manipulator's gripper. Then redefine the coordinate system of the two contact images obtained through the tactile sensors. Take the right sensor imaging as the benchmark and 320×240 imaging. The center of the area is the center of the coordinate system to establish the right image coordinate system o-xz. The left sensor imaging is mirrored according to the zox plane of the o-xyz coordinate system. The left image coordinate system o-xz is established with the image center as the center. The left and right images are generated. The contact areas are marked A ₁ and A ₂ respectively, and the corresponding centers are p ₁ (x ₁ , z ₁ ) and p ₂ (x ₂ , z ₂ ) respectively; among them, the coordinate value of p (x, z) is the tactile Mapping value from image space to real contact surface dimensions.

Further, step 4 specifically includes:

Step 4.1: Unify the initial state into the form of deviation; the initial deviation of the object includes the initial position deviation (Δx, Δy, Δz) and the initial angle deviation (Δr _x , Δr _y , Δr _z ), where Δy is obtained from the camera field of view. , its size represents the zox plane distance of the object from the grasping coordinate system o-xyz; the pose deviation at the initial moment is as follows:

Step 4.2: Obtain the motion characteristics of the contact ellipse by reading the video stream of the tactile image;

Step 4.3: Filter the ellipse parameters, then describe the ellipse state vector, and finally determine the posture change of the object during the sliding period:

Further, step 5 specifically includes:

Step 5.1: Optimize and adjust the initial slip of the object during the trial grasping process;

Step 5.2: Optimize and adjust the local slip that occurs during the grasping process of the object;

Step 5.3: Optimize and adjust the overall slippage of the object during the trial grasping process.

Further, step 5.1 is specifically as follows: According to the characteristics of initial slip convergence and bounding, the following optimization adjustments are made:

(1) It is expected that the center position of the contact area is within the sensing range of the sensor. If the center of the fitting ellipse for contact detection is outside the sensing range of the sensor at the time of initial contact, full slip has occurred, and the contact area will be adjusted during adjustment. The center is along the direction close to the center of the sensor's sensing range;

(2) For a small initial sliding motion, that is, the center of the fitting ellipse is within the sensing range of the sensor, it is only necessary to adjust the posture of the end effector in the opposite direction of the object's posture change. In particular, due to gravity Influence _{of y} , r _z , W) are adjusted to:

G′(x-dx,y-dy,z-dz×0.6,r _x -dr _x ,r _y -dr _y ,r _z -dr _z ,W).

Further, the step 5.2 is specifically as follows: compare with the grasping position map of the current object, search at the non-zero position of the grasping position map along the x-axis of the current grasping coordinate system in the direction close to the center of gravity of the object, and based on the confidence near the area The grasping configuration is reconstructed from the point with the highest degree, and the new grasping configuration is obtained by changing the spatial mapping function f. The direction of the search is determined by the two tactile sensors. If the dr _y of the object is a positive value, then along the negative direction of the x-axis Search, if dr _y is a negative value, search along the positive direction of the x-axis.

Further, step 5.3 is specifically as follows: if the current crawling configuration completely slips, the current crawling area is discarded, the crawling point on the crawling configuration map is replaced, and the confidence level on the current crawling location map is discarded. Peak, find another local peak where dr _y exceeds the threshold, then generate a new coordinate from the crawl map, and use this to infer the new crawl configuration to obtain the updated crawl configuration; finally, through multiple crawls Quality analysis and local adjustment convert the overall slip problem into a local slip or initial slip problem, thereby adjusting the grasping posture near the new grasping configuration until a stable grasping state is achieved.

Beneficial effects:

Aiming at the sensing mechanism and imaging characteristics of optical sensors, this invention describes the characteristics of tactile contact elements through morphological methods, and completes contact ellipse detection based on edge detection and geometric features; the pairs of visual input images and tactile sensing images are captured according to continuous actions. While making judgments on the stability of the object, the continuous motion information is used to evaluate the unstable movement trend of the object in the robot clamping arm;

This invention aims at different degrees and types of sliding trends, and combines the global information of the vision-based grasping map and the local posture perception information based on touch, so that the unstable trend can be effectively suppressed, so that the grab can be more effective again. The grabbing configuration enables the robot to achieve more stable and safer grabbing tasks;

The present invention solves the problem that existing grasping adjustment methods rely on trial and error and cannot fully utilize tactile modal information.

Description of the drawings

Figure 1 is a flow chart of a robot re-grasp optimization method based on tactile primitive slip feature feedback according to the present invention;

Figure 2 is a fitting diagram of contact area primitives;

Figure 3 shows the grasping coordinate system and the tactile image coordinate system;

Figure 4 is a schematic diagram of the sliding characteristics of the contact ellipse between two frames of the video stream;

Figure 5 is a schematic diagram of the specific process of optimizing and adjusting the different degrees of slippage of an object during trial grasping according to the present invention;

Figure 6 is a schematic diagram of the force analysis of the robotic arm grabbing a rod-shaped object;

Figure 7 is a schematic diagram of the object model used in grasping experiments according to the embodiment of the present invention;

Figure 8 is a flow chart of the re-grasp simulation experiment according to the embodiment of the present invention;

Figure 9 is a grasping prediction and optimized grasping posture diagram implemented in the present invention;

Detailed ways

In order to make the purpose, technical solution and technical effect of the present invention more clear, the present invention will be further described in detail below with reference to the drawings and examples of the description.

As shown in Figure 1, the invention proposes a robot re-grasp optimization method based on the feedback of tactile primitives' slip characteristics, which realizes the perception of the grasping state through the extraction and analysis of tactile primitives, and then realizes the grasping strategy. The optimization can be applied to a conventional six-axis industrial robot arm. An optical tactile sensor is installed on the robot arm. The specific implementation steps of this method are as follows:

Step 1: When using a robotic arm to grab an object, use an optical tactile sensor to obtain tactile modal information, that is, to sense the tactile primitives in the contact area.

Specifically, a vision-based tactile sensor is used to obtain tactile modal information through changes in the contact image and make the following inferences:

(1) Whether there is contact with the object, if so, the sensor contact image is obtained, if not, the non-contact tactile initial image is obtained;

(2) The area, position and direction of object contact;

(3) Obtain the movement of the contact part on the sensor surface through continuous video images;

Among them, when using a tactile sensor, the contact area of the object is estimated through single-point and multi-point features, and the perception of the contact area is realized through the extraction and matching of geometric features, thereby obtaining a difference map from the initial image;

Most of the configurations using elastomer materials and surface micro-convexities appear as elliptical contact areas or circular contact areas on imaging images. The contact area is described through tactile sensing modal elliptical primitive extraction, and the contact area is uniformly expressed as The format of A(p,a,b,θ); where, p(x,y) is the center coordinate of the contact ellipse in the tactile sensor image coordinate system, a is the semi-major axis length of the contact ellipse, and b is the semi-major axis of the contact ellipse. The length of the minor axis, θ is the deflection angle of the contact ellipse, taking the major axis of the ellipse as a reference, its value is the angle of the major axis deflection in the horizontal direction, and the value range is [-π/2,π/2]. In particular, when the shape of the contact area is circular, the length of the major axis is equal to the length of the minor axis, that is, a=b, and θ is any value within [-π/2, π/2].

Step 2: Based on the acquired tactile modal information, perform ellipse parameter fitting, specifically including:

Step 2.1: After performing grayscale and morphological processing on the difference image, use the Canny edge detection operator to perform edge detection and extraction on the contact area;

Step 2.2: Use the geometric ellipse fitting method based on the least squares method to fit the edge image extracted by the detection, as shown in Figure 2, which improves the detection speed while ensuring the accuracy of the ellipse parameter fitting.

Step 3: Construct the trial grasping coordinate system of the robotic arm, including:

Step 3.1: Based on the current grab configuration map G (x, y, z, r _x , r _y , r _z , W) deduced by the grab detection algorithm, establish a grab with the grab center coordinates as the origin. The coordinate system o-xyz is used to conveniently describe the motion characteristics of objects in the visual and tactile sensing link after the gripper of the robotic arm is stably closed;

Step 3.2: As shown in Figure 3, two face-to-face tactile sensors are installed on the working surface of the manipulator gripper in this embodiment. In order to facilitate the description of the contact information of the same object, the coordinate system of the two contact images is redefined to The right sensor imaging is used as a benchmark, and the right image coordinate system o-xz is established with the center of the 320×240 imaging area as the coordinate system center. The left sensor imaging is mirrored according to the zox plane of the o-xyz coordinate system, and the left image coordinate system is established with the image center as the center. In the image coordinate system o-xz, the contact areas generated by the left and right images are marked A ₁ and A ₂ respectively, and the corresponding centers are p ₁ (x ₁ ,z ₁ ) and p ₂ (x ₂ ,z ₂ ) respectively; where , the coordinate value of p(x,z) is the mapping value from the tactile image space to the real contact surface size.

Step 4: Trial grasp the object to cause the object to slip, calculate the tactile sensing sliding vector of the object, and infer the posture changes of the object during the sliding period, including:

Step 4.1: The present invention uniformly describes the initial state in the form of deviation; the initial deviation of the object includes the initial position deviation (Δx, Δy, Δz) and the initial angle deviation (Δr _x , Δr _y , Δr _z ), where Δy is determined by the camera Obtained from the field of view, its size represents the zox plane distance of the object from the grasping coordinate system o-xyz; the pose deviation at the initial moment is as follows:

Step 4.2: By reading the video stream of the tactile image, obtain the motion characteristics of the contact ellipse and the sliding characteristics between the two frame images, as shown in Figure 4;

Step 4.3: Due to errors in the ellipse fitting method, determining the state vector of the ellipse directly from the first and last frames of the video will lead to large errors. In order to reduce the error, filter the ellipse parameters first, and then describe the ellipse state vector. , and finally determine the pose change of the object during the sliding period:

Step 5: As shown in Figure 5, make corresponding optimization adjustments to the different degrees of slippage that occur during the trial grasping process of the object, and reconstruct the re-grasping posture of the robotic arm, specifically including:

Step 5.1: Optimize and adjust the initial slip of the object during the trial grasping process, specifically as follows:

According to the convergent and bounded characteristics of initial slip, the following two adjustment strategies are proposed:

(1) It is expected that the center position of the contact area is within the sensing range of the sensor. If the center of the fitting ellipse for contact detection is outside the sensing range of the sensor at the time of initial contact, global slip is very likely to occur, and adjustments should be made so that The center of the contact area is along a direction close to the center of the sensor's sensing range.

(2) For small initial sliding movements, the adjustment strategy only needs to adjust the pose of the end effector in the opposite direction of the object pose change. In particular, due to the influence of gravity, objects naturally have a tendency to move in the negative direction of the z-axis. For changes in the z-axis direction, scaling is used. The scaling factor used in the present invention is 0.6, which means that the configuration map G(x, y) will be captured. ,z,r _x ,r _y ,r _z ,W) adjusted to:

G′(x-dx,y-dy,z-dz×0.6,r _x -dr _x ,r _y -dr _y ,r _z -dr _z ,W).

Step 5.2: Optimize and adjust the local slip that occurs during the trial grasping process of the object, specifically as follows:

Compare the grab position map of the current object, search at the non-zero position of the grab map in the direction close to the center of gravity of the object along the x-axis of the current grab coordinate system, and reconstruct the grab configuration based on the point with the highest confidence near the area. A new crawling configuration is obtained by changing f through the spatial mapping function. The direction of the search is determined by two tactile sensors. If the dr _y of the object is a positive value, the search is along the negative direction of the x-axis. If the dr _y is a negative value, the search is along the positive direction of the x-axis.

Taking a rod-shaped object in the three-dimensional space as an example, as shown in Figure 6, through force analysis, it can be seen that the critical condition for stable grasping is that the friction torque of the fingers at the end of the mechanical arm is equal to gravity times the force arm, where the distance of the force arm It can be obtained from the formula:

Among them, L is the gravity arm. Moving this distance along the x-axis direction under the current grasping coordinate can correct the local slip into a stable grasping action; T is the torque of the friction pair; θ is the angle with the horizontal direction.

Since the type, shape, quality, distribution characteristics, surface contact characteristics and other attributes of the grasped object are unknown, the above equation needs to be evaluated. In this embodiment, the mass of the grabbed object does not exceed 1kg, so G=9.8 can be set, and the friction torque T can be obtained by the following formula.

T＝μF∫ _A rdr

Among them, μ is the friction coefficient between the finger and the object, and the approximate value in this embodiment is 0.7; F is the clamping force, the size of which can be read from the state of the clamp; A is the contact area between the object and the tactile sensor, because the surface curvature of the object The distribution is inconsistent. To facilitate subsequent processing, this embodiment assumes that the force distribution in the deformation area of the elastic body meets the requirements of uniform distribution. In summary, the friction torque T can be estimated. Finally, take the estimated L as the explored distance and map it to the image space to obtain the explored distance L _i , vector That is the iteration direction and step size of the search in the image space. The grab point is replaced in each iteration, and the new grab coordinates are expressed as follows:

Step 5.3: Optimize and adjust the overall slippage of the object during the trial grasping process, specifically as follows:

If the current grabbing configuration completely slips, it means that the current grabbing posture is not suitable for the current grabbing task; the adjustment strategy should discard the current grabbing area, replace the grabbing points on the grabbing configuration map, and discard the current grabbing position. The peak confidence level on the map is used to find another local peak value that exceeds the threshold, and then a new coordinate is generated from the capture map, and the new capture configuration is inferred from this to obtain the updated capture configuration; finally, through multiple Secondary grasping quality analysis and local adjustment can convert the overall slip problem into a local slip or initial slip problem, so that the grasping posture can be adjusted according to the above strategy near the new grasping configuration until a stable grasping state is achieved.

The purpose and effects of the present invention will become more apparent when the present invention is described below based on the embodiments.

Performance measurement of the present invention: The object used in this example comes from EGAD. Before conducting the grasping experiment, first construct its collision model and scale it to a size suitable for clamping, and simulate the properties of the grasped object by randomly changing the position of its center of gravity. Uncertainty: 10 random center-of-gravity changes were performed for each grasped object, and 10 grasping experiments were conducted for each object with a changed center of gravity. The object model used in the experiment is shown in Figure 7. In this embodiment, a simulation experiment is carried out based on Pybullet. The selected robot model is Rethink sawyer, the parallel gripper model is WSG-50, and the tactile sensor is DIGIT. The experimental process is shown in Figure 8. The grabbing configuration results before and after adjustment are shown in Figure 8. 9 shown.

Example 1:

In order to simulate the uncertainty of the grasping object's pose, this embodiment randomly samples the grasping object's pose uncertainty space shown in the following formula; in order to simulate the robot execution error and the influence of other uncertain environments, when executing the grasp When adding random perturbations to the predicted grasping posture.

S＝{＜x,y,θ＞∣x∈[-20,20],y∈[-20,20],θ∈[0,2π]}

In the embodiment, the mass of the grabbed object is 1kg, the clamping force is 20N, and the upper limit of adjustment times is five. The change in posture of the object before and after grabbing is obtained through Pybullet's getBasePositionAndOrientation method. If the change reaches the ideal value If the change reaches 80% to 100%, it is considered a successful capture, otherwise it is considered a failed capture. Among them, if the change reaches 90% to 100% of the ideal value, it is considered a stable capture.

The experimental results are shown in the following table:

As can be seen from the table, with the application of re-grasp posture adjustment, the average grasp success rate increased from 87.6% to 92.2%, of which the stable grasp rate was 88.4%. This result shows that the reactive re-grasp optimization strategy improves the performance of Stability of grasping process under uncertainty of object and its pose.

The above are only preferred implementation examples of the present invention, and do not impose any formal restrictions on the present invention. Although the implementation process of the present invention has been described in detail above, those familiar with the art can still modify the technical solutions recorded in the foregoing examples, or make equivalent substitutions for some of the technical features. All modifications, equivalent substitutions, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (9)

1. A robot re-grasp optimization method based on tactile primitive slip feature feedback, which is characterized by including the following steps: Step 1: Use the optical tactile sensor installed on the robotic arm to obtain tactile modal information, that is, sense and acquire the tactile primitives in the contact area; Step 2: Perform ellipse parameter fitting based on the acquired tactile modal information; Step 3: Construct the trial grasping coordinate system of the robotic arm; Step 4: Trial grasp the object to cause the object to slip, calculate the tactile sensing sliding vector of the object, and infer the posture changes of the object during the sliding period; Step 5: Make corresponding optimization adjustments to the different degrees of slippage that occur during the trial grasping process of the object, and reconstruct the re-grasping posture of the robotic arm. 2. A robot re-grasp optimization method based on tactile element slip feature feedback as claimed in claim 1, characterized in that step 1 is specifically: using a vision-based tactile sensor to pass changes in the contact image , determine whether the gripper of the manipulator is in contact with the object, if so, obtain the sensor contact image, if not, obtain the non-contact tactile initial image; obtain the area, position, and direction of the object contact; obtain the contact part on the sensor through continuous video images Surface motion conditions; Wherein, the tactile sensor estimates the contact area of the object through single-point and multi-point features, and realizes the perception of the contact area through the extraction and matching of geometric features, thereby obtaining a difference map from the initial image; The characteristics of the elastomer material and the slightly convex surface are used to show the characteristics of an elliptical contact area or a circular contact area on the imaging image. The contact area is described through the tactile sensing modal elliptical primitive extraction, and the contact area is uniformly expressed as A The format of (p, a, b, θ); where, p (x, y) is the center coordinate of the contact ellipse in the tactile sensor image coordinate system, a is the semi-major axis length of the contact ellipse, and b is the semi-minor axis of the contact ellipse. Axis length, θ is the deflection angle of the contact ellipse, taking the long axis of the ellipse as a reference, its value is the angle of the long axis deflection in the horizontal direction, the value range is [-π/2, π/2]; when the shape of the contact area is In the case of a circle, the length of the major axis and the length of the minor axis are equal, that is, a=b, and θ is any value within [-π/2, π/2]. 3. A robot re-grasp optimization method based on tactile element slip feature feedback as claimed in claim 2, characterized in that step 2 specifically includes: Step 2.1: After performing grayscale and morphological processing on the difference image, use the Canny edge detection operator to perform edge detection and extraction on the contact area; Step 2.2: Use the geometric ellipse fitting method based on the least squares method to fit the detected edge image. 4. A robot re-grasp optimization method based on tactile element slip feature feedback as claimed in claim 3, characterized in that step 3 specifically includes: Step 3.1: Based on the current grab configuration map G (x, y, z, r _x , r _y , r _z , W) deduced by the grab detection algorithm, establish a grab with the grab center coordinates as the origin. Coordinate system o-xyz; Step 3.2: Install two face-to-face tactile sensors on the working surface of the manipulator's gripper. Then redefine the coordinate system of the two contact images obtained through the tactile sensors. Take the right sensor imaging as the benchmark and 320×240 imaging. The center of the area is the center of the coordinate system to establish the right image coordinate system o-xz. The left sensor imaging is mirrored according to the zox plane of the o-xyz coordinate system. The left image coordinate system o-xz is established with the image center as the center. The left and right images are generated. The contact areas are marked A ₁ and A ₂ respectively, and the corresponding centers are p ₁ (x ₁ , z ₁ ) and p ₂ (x ₂ , z ₂ ) respectively; among them, the coordinate value of p (x, z) is the tactile Mapping value from image space to real contact surface dimensions. 5. A robot re-grasp optimization method based on tactile element slip feature feedback as claimed in claim 4, characterized in that step 4 specifically includes: Step 4.1: Unify the initial state into the form of deviation; the initial deviation of the object includes the initial position deviation (Δx, Δy, Δz) and the initial angle deviation (Δr _x , Δr _y , Δr _z ), where Δy is obtained from the camera field of view. , its size represents the zox plane distance of the object from the grasping coordinate system o-xyz; the pose deviation at the initial moment is as follows: Step 4.2: Obtain the motion characteristics of the contact ellipse by reading the video stream of the tactile image; Step 4.3: Filter the ellipse parameters, then describe the ellipse state vector, and finally determine the posture change of the object during the sliding period: 6. A robot re-grasp optimization method based on tactile element slip feature feedback as claimed in claim 5, characterized in that step 5 specifically includes: Step 5.1: Optimize and adjust the initial slip of the object during the trial grasping process; Step 5.2: Optimize and adjust the local slip that occurs during the grasping process of the object; Step 5.3: Optimize and adjust the overall slippage of the object during the trial grasping process. 7. A robot re-grasp optimization method based on feedback of tactile primitive slip characteristics as claimed in claim 6, characterized in that step 5.1 is specifically: according to the characteristics of initial slip convergence and bounding, Make the following optimization adjustments: (1) It is expected that the center position of the contact area is within the sensing range of the sensor. If the center of the fitting ellipse for contact detection is outside the sensing range of the sensor at the time of initial contact, full slip has occurred, and the contact area will be adjusted during adjustment. The center is along the direction close to the center of the sensor's sensing range; (2) For a small initial sliding motion, that is, the center of the fitting ellipse is within the sensing range of the sensor, it is only necessary to adjust the posture of the end effector in the opposite direction of the object's posture change. In particular, due to gravity Influence _{of y} , r _z , W) are adjusted to: G′(x-dx,y-dy,z-dz×0.6,r _x -dr _x ,r _y -dr _y ,r _z -dr _z ,W). 8. A robot re-grasp optimization method based on tactile element slip feature feedback as claimed in claim 6, characterized in that step 5.2 is specifically: comparing the grasp position map of the current object, Search the non-zero position of the position map along the x-axis of the current grasping coordinate system in the direction close to the center of gravity of the object, reconstruct the grasping configuration based on the point with the highest confidence near the area, and obtain a new grasp by changing the spatial mapping function f Taking the configuration, the direction of the search is determined by the two tactile sensors. If the dr _y of the object is a positive value, the search is along the negative direction of the x-axis. If the dr _y is a negative value, the search is along the positive direction of the x-axis. 9. A robot re-grasp optimization method based on tactile primitive slip feature feedback as claimed in claim 6, characterized in that step 5.3 is specifically: if the current gripping configuration fully slips, Then abandon the current crawling area, replace the crawling point on the crawling configuration map, discard the peak confidence level on the current crawling location map, find another local peak value where dr _y exceeds the threshold, and then generate a new one from the crawling map. coordinates, and use this to infer the new grasping configuration, and obtain the updated grasping configuration; finally, through multiple grasping quality analyzes and local adjustments, the overall slip problem is transformed into a local slip or initial slip problem. Thereby adjusting the grasping posture near the new grasping configuration until a stable grasping state is achieved.