CN-122008215-A - Anti-slip control method for mechanical arm grabbing

CN122008215ACN 122008215 ACN122008215 ACN 122008215ACN-122008215-A

Abstract

The invention discloses a control method for preventing a mechanical arm from slipping off during grabbing, which relates to the technical field of intelligent sorting and recycling and comprises the following steps of collecting visual information of an object in real time and determining grabbing information of a clamping jaw on the mechanical arm based on the visual information; the visual information comprises displacement data, three-dimensional shape data and anti-skid data, and the grabbing information comprises grabbing positions, grabbing forces, clamping jaw opening and closing angles and tail end attitude angles of the mechanical arm. According to the method, the characteristics of the object are accurately depicted through the displacement data, the three-dimensional shape data and the anti-skid data, the adaptive grabbing information is output by combining the machine learning model, the sliding risk is reduced from the source, a secondary grabbing mechanism for matching the preset sliding path with the real-time sliding path is designed according to the sliding risk, a plurality of preset sliding paths are simulated through dynamics, the optimal preset sliding path is dynamically matched by combining the real-time track, the object can be grabbed in a rescuing mode, and the grabbing success rate can be improved.

Inventors

Quan Zhengwen
YANG XIAOLIN
LUO HONGTAI
LIANG LONGYI

Assignees

广西升禾资源再生利用技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260224

Claims (9)

1. The mechanical arm grabbing anti-slip control method is characterized by comprising the following steps of: the method comprises the steps of collecting visual information of an object in real time, and determining grabbing information of a clamping jaw on a mechanical arm based on the visual information, wherein the visual information comprises displacement data, three-dimensional shape data and anti-skid data, and the grabbing information comprises grabbing positions, grabbing forces, opening and closing angles of the clamping jaw and the attitude angle of the tail end of the mechanical arm; The clamping jaw is used for grabbing the object through grabbing information, judging whether the object has a sliding risk or not after grabbing the object, generating a plurality of preset sliding paths if the object has the sliding risk, grabbing the object for the second time according to the optimal preset sliding path in the preset sliding paths, and moving the object to a corresponding storage position if the object does not have the sliding risk.
2. The method for controlling the gripping and sliding prevention of the mechanical arm according to claim 1, wherein the step of collecting the visual information of the object in real time comprises the following steps: the method comprises the steps that an industrial camera and a laser radar are adopted for collaborative collection, the industrial camera obtains two-dimensional images and depth information of an object, and the laser radar scans the three-dimensional outline of the object; Performing Gaussian filtering on the two-dimensional image to remove image noise and calibrate depth information; calculating the two-dimensional coordinates and the vertical height of the object relative to the conveyor belt in real time, and updating the displacement variation amount every frame to obtain displacement data; generating object volume, length, width and height, barycentric coordinates and surface contours through point cloud splicing, and obtaining three-dimensional shape data; and calculating the surface roughness and the liquid coverage of the object based on the image gray level co-occurrence matrix to obtain anti-skid data.
3. The method for controlling the gripping and anti-slip of the mechanical arm according to claim 1, wherein the step of determining the gripping information of the gripping jaw on the mechanical arm based on the visual information comprises the following steps: Building and training a machine learning model; combining the visual information into feature vectors, and inputting a machine learning model which is trained in advance; And outputting a grabbing information number according to the input feature vector by the machine learning model, wherein the grabbing information number corresponds to the grabbing information.
4. The method for controlling the gripping and anti-slip control of the mechanical arm according to claim 3, wherein the step of constructing and training a machine learning model comprises the steps of: collecting historical visual information and corresponding historical grabbing information as samples, and constructing a sample set by using a plurality of samples; Numbering the history grabbing information, combining the history visual information into a history feature vector, defining an input layer and an output layer, and building a machine learning framework; Dividing the sample set into a training set and a testing set, constructing a machine learning model, training the machine learning model, testing the machine learning model by using the testing set, and outputting the machine learning model meeting the preset accuracy to obtain the machine learning model after training.
5. The method for controlling the gripping and anti-slip of the mechanical arm according to claim 1, wherein the step of gripping the object by the gripping jaw through gripping information comprises the steps of: The mechanical arm motion planner receives the grabbing information and plans a collision-free motion track based on the grabbing information; the mechanical arm is driven to drive the clamping jaw to grab the object, wherein in the grabbing process, the clamping jaw acquires force feedback data of the object.
6. The method for controlling the gripping and sliding prevention of the mechanical arm according to claim 5, wherein the step of determining whether the object has a sliding risk after the object is gripped comprises the steps of: Setting a displacement threshold based on specification data of the object; After the object is grabbed off the conveyor belt, continuously monitoring sliding data of the object relative to the clamping jaw through a miniature short-distance laser sensor or a visual mark point arranged on the clamping jaw; And comparing the sliding data with a displacement threshold value to judge whether the object has a sliding risk or not, wherein if the sliding data is larger than or equal to the displacement threshold value, the object is judged to have the sliding risk, and if the sliding data is smaller than the displacement threshold value, the object is judged to not have the sliding risk.
7. The method for controlling the gripping and sliding prevention of the mechanical arm according to claim 6, wherein the step of generating a plurality of preset sliding paths comprises: Once judging that the object has a sliding risk, carrying out simulation deduction based on visual information of the object so as to obtain a plurality of preset sliding paths of the object; the preset sliding path consists of a plurality of path points and corresponding point information on a time sequence, wherein the point information comprises sliding speed and position information.
8. The method for controlling the grabbing and sliding preventing of the mechanical arm according to claim 7, wherein the step of secondarily grabbing the object according to the optimal preset sliding path among the preset sliding paths comprises the following steps: tracking a real-time sliding path of an object; Real-time matching is carried out on the real-time sliding paths, the path points of the preset sliding paths and the corresponding point information, so that the preset sliding paths with the highest matching degree are obtained, and the preset sliding paths with the highest matching degree are used as optimal preset sliding paths; the clamping jaw performs secondary grabbing on the object based on the non-passing path points and corresponding point information of the object in the optimal preset sliding path; The secondary grabbing process comprises the steps of taking non-passing path points and corresponding point information of an object in an optimal preset sliding path as new displacement data, inputting the new displacement data, three-dimensional shape data and new anti-slip data into a machine learning model which is completed through training, outputting grabbing information numbers by the machine learning model which is completed through training, grabbing information numbers correspond to the new grabbing information, and grabbing the object secondarily by clamping jaws based on the new grabbing information, wherein a mapping table of historical force feedback data and anti-slip data is established in advance, and the anti-slip data in visual information is corrected based on the force feedback data and the mapping table to obtain the new anti-slip data.
9. The method of claim 1, wherein the step of moving the object to the corresponding storage position comprises: establishing a corresponding relation between materials and storage positions; determining the material quality of the object, and moving the object to the corresponding storage position based on the material quality and the corresponding relation of the object.

Description

Anti-slip control method for mechanical arm grabbing Technical Field The invention relates to the technical field of intelligent sorting and recycling, in particular to a control method for grabbing and anti-slipping of a mechanical arm. Background In the field of intelligent sorting and recycling, automatic sorting of mechanical arms is a core technology for realizing efficient and accurate treatment of mass materials. The anti-slip control in the grabbing process is a key link for guaranteeing continuous and stable operation of the sorting assembly line, and the sorting efficiency, accuracy and reliability of the system are directly determined. Currently, the common control of robot anti-slip in industry is mostly dependent on a rule-based feedback control method. The typical implementation path is that a force sensor and the like are used for detecting whether the object to be grabbed slides relatively, and once a sliding signal is detected, a control system increases the normal grabbing force of the clamping jaw according to a preset and fixed proportional gain coefficient so as to inhibit sliding by increasing static friction force. Although this method is intuitive in principle and effective under certain working conditions, it reveals significant limitations in the face of complex and varied actual industrial scenarios, especially applications such as renewable resource sorting where objects are highly non-standardized and the working environment is much disturbed, mainly appearing in the following aspects: For example, regular feedback control is difficult to achieve good trade-off between real-time response and system robustness, control performance is highly dependent on a gain coefficient determined by pre-debugging, for smooth or heavier objects, too small gain can cause too slow or too small reinforcement to stop slipping in time, for light or rough objects, too large gain can cause over-excitation response, instantaneous rising of grabbing force can cause mechanical oscillation, processing capacity for non-rigid or easily deformable objects is weak, and when the objects such as soft plastic films and fluffy paper masses are faced, a simple slipping-increasing grabbing force strategy is quite suitable, overall friction force is easy to decline, slipping risks are aggravated, when anti-slipping control is invalid, model parameters are not updated immediately by failure information and a more reliable grabbing attempt is initiated quickly when the objects enter an irreversible slipping process. Disclosure of Invention The invention aims to provide a control method for preventing a mechanical arm from slipping off during grabbing, which aims to solve the problems in the background art. In order to achieve the purpose, the invention provides the following technical scheme that the mechanical arm grabbing anti-slip control method comprises the following steps: the method comprises the steps of collecting visual information of an object in real time, and determining grabbing information of a clamping jaw on a mechanical arm based on the visual information, wherein the visual information comprises displacement data, three-dimensional shape data and anti-skid data, and the grabbing information comprises grabbing positions, grabbing forces, opening and closing angles of the clamping jaw and the attitude angle of the tail end of the mechanical arm; The clamping jaw is used for grabbing the object through grabbing information, judging whether the object has a sliding risk or not after grabbing the object, generating a plurality of preset sliding paths if the object has the sliding risk, grabbing the object for the second time according to the optimal preset sliding path in the preset sliding paths, and moving the object to a corresponding storage position if the object does not have the sliding risk. In a preferred embodiment, the step of collecting visual information of the object in real time includes: the method comprises the steps that an industrial camera and a laser radar are adopted for collaborative collection, the industrial camera obtains two-dimensional images and depth information of an object, and the laser radar scans the three-dimensional outline of the object; Performing Gaussian filtering on the two-dimensional image to remove image noise and calibrate depth information; calculating the two-dimensional coordinates and the vertical height of the object relative to the conveyor belt in real time, and updating the displacement variation amount every frame to obtain displacement data; generating object volume, length, width and height, barycentric coordinates and surface contours through point cloud splicing, and obtaining three-dimensional shape data; and calculating the surface roughness and the liquid coverage of the object based on the image gray level co-occurrence matrix to obtain anti-skid data. In a preferred embodiment, the step of determining the grasping information of the gripper