A method, device, equipment and medium for generating grasping force

Abstract

The present application provides a method, apparatus, device, and medium for generating a grasping force. The method includes: determining an initial image of an object to be grasped according to a preset image acquisition device, and determining the category of the object to be grasped according to the initial image and a preset recognition model; determining the grasping information of the object to be grasped according to a preset object information library and the category, and determining an initial grasping force corresponding to the object to be grasped according to the grasping information; grasping the object to be grasped according to the initial grasping force, and determining the displacement amount and deformation amount of the object to be grasped according to the image acquisition device and the recognition model; determining a grasping force increment of the object to be grasped according to the displacement amount and the deformation amount; and determining the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force. In this way, the accuracy and adaptability of the grasping force are improved, so as to better grasp the object.

Description

Technical Field

[0001] This application relates to the technical field of object grasping, and particularly to a method, apparatus, device, and medium for generating a grasping force. Background Art

[0002] Robot grasping refers to the process in which a robot uses its mechanical structure and control system to perform operations such as clamping, holding, or adsorbing on a target object through specific actions and operations, so as to achieve the picking up, transporting, or operating of the object. It has wide applications in many fields such as industrial manufacturing, logistics, medical treatment, and home services, and can improve production efficiency, reduce labor costs, and complete some dangerous or high-precision operation tasks.

[0003] Existing object grasping methods generally calculate the grasping force based on the force control closed-loop of physical sensors and mechanical models. The physical sensors real-time feedback the grasping force, and the grasping force is adjusted by combining PID control or adaptive algorithms.

[0004] However, the existing object grasping methods have the problems of low accuracy and adaptability of the generated grasping force. Summary of the Invention

[0005] This application provides a method, apparatus, device, and medium for generating a grasping force to solve the problems of low accuracy and adaptability of the generated grasping force in existing object grasping methods.

[0006] In a first aspect, this application provides a method for generating a grasping force, the method comprising:

[0007] Determine an initial image of the object to be grasped according to a preset image acquisition device, and determine the category of the object to be grasped according to the initial image and a preset recognition model;

[0008] Determine the grasping information of the object to be grasped according to a preset object information library and the category, and determine the initial grasping force corresponding to the object to be grasped according to the grasping information;

[0009] Grasp the object to be grasped according to the initial grasping force, and determine the displacement and deformation amount of the object to be grasped according to the image acquisition device and the recognition model;

[0010] Determine the grasping force increment of the object to be grasped according to the displacement and deformation amount;

[0011] Determine the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force.

[0012] In some embodiments of this application, determining the grasping information of the object to be grasped according to a preset object information library and the category, and determining the initial grasping force corresponding to the object to be grasped according to the grasping information includes:

[0013] Determine a preset object information library;

[0014] According to the object information library and the category, determine the density information of the object to be grasped and the corresponding static friction coefficient;

[0015] According to the image acquisition device, determine the volume information of the object to be grasped, and according to the density information, static friction coefficient and volume information, determine the grasping information;

[0016] According to the grasping information, determine the initial grasping force.

[0017] In some embodiments of the present application, determining the initial grasping force according to the grasping information includes:

[0018] Determine the density information, static friction coefficient and volume information in the grasping information, and according to the density information and volume information, determine the mass information corresponding to the object to be grasped;

[0019] Determine a preset grasping force threshold;

[0020] According to the mass information, static friction coefficient and grasping force threshold, determine the initial grasping force.

[0021] In some embodiments of the present application, determining the initial grasping force according to the mass information, static friction coefficient and grasping force threshold includes:

[0022] According to the mass information and static friction coefficient, determine the initial grasping force;

[0023] According to the object information library, determine a preset target object category;

[0024] Judge whether the category of the object to be grasped is the target object category;

[0025] If the category of the object to be grasped is the target object category, compare the size of the initial grasping force and the grasping force threshold to obtain a comparison result;

[0026] If the comparison result is that the initial grasping force is greater than the grasping force threshold, adjust the value corresponding to the initial grasping force to the value corresponding to the grasping force threshold.

[0027] In some embodiments of the present application, grasp the object to be grasped according to the initial grasping force, and according to the image acquisition device and the recognition model, determine the displacement amount and deformation amount of the object to be grasped, including:

[0028] Determine the marked points corresponding to the initial image;

[0029] According to the image acquisition device, obtain continuously acquired object images and determine the marked points in each object image;

[0030] Perform optical flow method calculation on each marked point to obtain the displacement amount;

[0031] Based on the recognition model, the object image is recognized in real time to obtain the deformation amount.

[0032] In some embodiments of the present application, according to the displacement amount and the deformation amount, determining the grasping force increment of the object to be grasped includes:

[0033] According to the category of the object to be grasped, determine the corresponding displacement threshold and deformation threshold;

[0034] According to the displacement amount and the displacement threshold, as well as the deformation amount and the deformation threshold, determine the reward value corresponding to the object to be grasped;

[0035] Compare the reward value with the preset reward threshold to obtain a comparison result;

[0036] If the comparison result is that the reward value is less than the reward threshold, determine the preset reward determination model, and according to the displacement amount, the deformation amount and the reward determination model, determine the grasping force increment;

[0037] If the comparison result is that the reward value is not less than the reward threshold, do not determine the grasping force increment.

[0038] In some embodiments of the present application, if the comparison result is that the reward value is less than the reward threshold, determine the preset reward determination model, and according to the displacement amount, the deformation amount and the reward determination model, determine the grasping force increment, including:

[0039] According to the reward determination model, determine the predicted reward value corresponding to the displacement amount and the deformation amount;

[0040] According to the predicted reward value, determine the grasping force increment.

[0041] In a second aspect, the present application provides a grasping force generation device, the device includes:

[0042] An image determination module, configured to determine the initial image of the object to be grasped according to a preset image acquisition device, and determine the category of the object to be grasped according to the initial image and a preset recognition model;

[0043] An information determination module, configured to determine the grasping information of the object to be grasped according to a preset object information library and the category, and determine the initial grasping force corresponding to the object to be grasped according to the grasping information;

[0044] An object grasping module, configured to grasp the object to be grasped according to the initial grasping force, and determine the displacement amount and the deformation amount of the object to be grasped according to the image acquisition device and the recognition model;

[0045] An increment determination module, configured to determine the grasping force increment of the object to be grasped according to the displacement amount and the deformation amount;

[0046] A grasping force determination module, configured to determine the real-time grasping force of an object to be grasped according to the grasping force increment and the initial grasping force.

[0047] In a third aspect, the present application provides a computer device, including: a processor, and a memory communicatively connected to the processor;

[0048] The memory stores computer-executable instructions;

[0049] The processor executes the computer-executable instructions stored in the memory to implement the method of the present application.

[0050] In a fourth aspect, the present application provides a computer-readable storage medium, in which program code is stored, and when the program code is executed by a processor, it is used to implement the method of the present application.

[0051] A method, device, equipment and medium for generating a grasping force provided by the present application, by determining an initial image of an object to be grasped according to a preset image acquisition device, and determining the category of the object to be grasped according to the initial image and a preset recognition model; determining the grasping information of the object to be grasped according to a preset object information library and the category, and determining the initial grasping force corresponding to the object to be grasped according to the grasping information; grasping the object to be grasped according to the initial grasping force, and determining the displacement and deformation amount of the object to be grasped according to the image acquisition device and the recognition model; determining the grasping force increment of the object to be grasped according to the displacement and deformation amount; and determining the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force.

[0052] In this way, the category and grasping information of the object to be grasped can be determined through the image acquisition device and the recognition model, so as to determine the initial grasping force, and the grasping force increment can be determined through the displacement and deformation amount during the grasping process, so as to realize the real-time adjustment of the grasping force of the object during the object grasping process, improve the grasping success rate, enhance the generality and flexibility of the grasping, and improve the overall grasping efficiency and stability. Description of the Drawings

[0053] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0054] Figure 1 It is a schematic flowchart of a method for generating a grasping force provided by an embodiment of the present application;

[0055] Figure 2 It is a schematic flowchart of another method for generating a grasping force provided by an embodiment of the present application;

[0056] Figure 3Schematic structural diagram of a gripping force generation device provided by an embodiment of the present application;

[0057] Figure 4 Block diagram of the structure of a device for executing a gripping force generation method according to an embodiment of the present application. Detailed implementation manners

[0058] Here, exemplary embodiments will be described in detail, and examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0059] The technical solution of the present application and how the technical solution of the present application solves the above technical problems will be described in detail below with specific embodiments. The following several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below with reference to the drawings.

[0060] Figure 1 Flow schematic diagram of a gripping force generation method provided by an embodiment of the present application. As Figure 1 shown, the gripping force generation method may include the following steps:

[0061] S110. Determine an initial image of an object to be gripped according to a preset image acquisition device, and determine the category of the object to be gripped according to the initial image and a preset recognition model.

[0062] Among them, gripping is a key action for operating an object, which can be realized by an intelligent robot. The robot can learn a large amount of gripping data, autonomously optimize the gripping strategy, and improve the success rate and efficiency of gripping.

[0063] The preset image acquisition device is a device preset for image acquisition, which may be a depth camera. A depth camera is an imaging device capable of obtaining the depth information of objects in a scene, and can collect the RGB-D images of objects. The RGB-D image is an image type that fuses the information of a color image (RGB) and a depth image (D). The depth information is very helpful for target recognition and positioning. The RGB-D image allows the robot to more accurately calculate the three-dimensional position and pose of an object, thereby improving the accuracy of gripping.

[0064] The object to be grasped is the object that the robot needs to grasp. The initial image is the image of the object to be grasped determined by the image acquisition device before the grasping action starts, and is used to identify the grasping information of the object. Thus, before grasping, the corresponding grasping force is determined according to the grasping information, and then the object is grasped by the grasping force.

[0065] The preset recognition model is a model that is preset and used to recognize the initial image. It can be a large vision model, such as the CLIP model. Through learning and analyzing a large amount of visual data, the large vision model can automatically extract features and patterns in visual information such as images and videos, so as to realize various tasks such as understanding, classifying, recognizing, and generating visual content.

[0066] The category of the object to be grasped is the corresponding type of the object. Different types of objects require different grasping forces when being grasped. For example, the grasping forces for fragile objects such as eggs and hard objects such as stones are different. Otherwise, it is easy to fail in grasping and even damage the object.

[0067] Based on this, the initial image of the object to be grasped is determined by the image acquisition device, and the initial image is recognized according to the recognition model, so as to determine the category of the object to be grasped, so that the corresponding initial grasping force of the object can be determined subsequently, and the object to be grasped can be grasped according to the initial grasping force.

[0068] S120. According to the preset object information library and the category, determine the grasping information of the object to be grasped, and according to the grasping information, determine the corresponding initial grasping force of the object to be grasped.

[0069] Among them, the preset object information library is an information library that is preset and stores the object information of multiple objects. The object information can include information such as the category and density of the object.

[0070] The grasping information is the information used to determine the corresponding initial grasping force of the object. For example, it can be information such as the category and density of the object. The initial grasping force is the force when initially grasping the object.

[0071] Based on this, by determining the category of the object to be grasped, the corresponding object information in the object information library is determined, and then the grasping information corresponding to the object to be grasped is determined, so as to determine the corresponding initial grasping force according to the grasping information, and the object to be grasped can be grasped according to the initial grasping force subsequently.

[0072] S130. Grasp the object to be grasped according to the initial grasping force, and determine the displacement and deformation amount of the object to be grasped according to the image acquisition device and the recognition model.

[0073] Among them, the displacement is the displacement change of the object relative to the robot during the grasping process. For example, the object may slip during the grasping process, resulting in a displacement relative to the robot; the deformation is the dimensional change of the object in each direction during the grasping process, such as the change in length, width or height. The small deformation of a rigid object can be ignored.

[0074] Based on this, after grasping the object to be grasped with the initial grasping force, according to the image acquisition device and the recognition model, the displacement change and the dimensional deformation of the object to be grasped during the grasping process are determined in real time; in the actual grasping process, the object may undergo dynamic changes such as slipping and deformation, resulting in the initial grasping force no longer meeting the current grasping situation, and it is easy to result in grasping failures such as the object slipping. Therefore, it is necessary to adjust the initial grasping force in real time according to the dynamic changes of the object during the actual grasping process to ensure successful grasping.

[0075] S140. Determine the grasping force increment of the object to be grasped according to the displacement and the deformation.

[0076] Among them, the grasping force increment is the adjustment increment of the grasping force according to the dynamic changes of the object during the actual grasping process.

[0077] Based on this, by determining the dynamic changes such as the displacement and deformation of the object relative to the robot during the actual grasping process, the corresponding grasping force increment is determined, so as to adjust the grasping force in real time according to the grasping force increment subsequently.

[0078] S150. Determine the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force.

[0079] Based on this, before the grasping starts, the corresponding initial grasping force is determined according to the category of the object, and during the grasping process, the corresponding grasping force increment is determined according to the dynamic changes such as the displacement and deformation of the object, so as to determine the real-time grasping force according to the initial grasping force and the grasping force increment, thereby realizing precise grasping of the object and improving the success rate of grasping.

[0080] On the basis of the feasible implementation manner of the above S120, the present application further provides steps for determining the grasping information of the object to be grasped according to the preset object information library and category, and determining the corresponding initial grasping force of the object to be grasped according to the grasping information, including:

[0081] Determine the preset object information library;

[0082] Determine the density information and the corresponding static friction coefficient of the object to be grasped according to the object information library and the category;

[0083] Determine the volume information of the object to be grasped according to the image acquisition device, and determine the grasping information according to the density information, static friction coefficient and volume information;

[0084] Determine the initial grasping force according to the grasping information.

[0085] Among them, the density information is the density corresponding to the object to be grasped. Density is a measure of the mass within a specific volume and is a characteristic of matter; the volume information is the volume corresponding to the object to be grasped. Volume refers to the size of the space occupied by the object; the static friction coefficient is a physical quantity that describes the friction characteristics between two objects that are in contact and relatively stationary.

[0086] Based on this, by determining the category of the object to be grasped, the corresponding density and static friction coefficient of the object of this category are determined according to the object information library, and the volume of the object is determined through the depth camera, so as to determine the corresponding grasping information according to the density, volume and static friction coefficient of the object, so as to determine the corresponding initial grasping force according to the grasping information subsequently.

[0087] On the basis of the feasible implementation manner of the above S120, the present application further provides steps for determining the initial grasping force according to the grasping information, including:

[0088] Determine the density information, static friction coefficient and volume information in the grasping information, and determine the mass information corresponding to the object to be grasped according to the density information and volume information;

[0089] Determine the preset grasping force threshold;

[0090] Determine the initial grasping force according to the mass information, static friction coefficient and grasping force threshold.

[0091] Among them, the mass information is the mass of the object to be grasped. Mass is a physical quantity that measures the inertia of an object, is a physical property possessed by the object, is a measure of the amount of matter, and can be determined according to the density and volume of the object.

[0092] The preset grasping force threshold is a critical value of the grasping force set in advance, which is used for subsequent comparison with the grasping force, so as to adjust the initial grasping force according to the comparison result.

[0093] Based on this, the corresponding mass is determined through the density and volume of the object to be grasped, and the grasping force threshold is determined, so as to determine the initial grasping force according to the mass, static friction coefficient and grasping force threshold of the object to be grasped subsequently.

[0094] On the basis of the feasible implementation manner of the above S130, the present application further provides steps for determining the initial grasping force according to the mass information, static friction coefficient and grasping force threshold, including:

[0095] Determine the initial grasping force according to the quality information and the static friction coefficient;

[0096] Determine the preset target object category according to the object information library;

[0097] Judge whether the category of the object to be grasped is the target object category;

[0098] If the category of the object to be grasped is the target object category, compare the size of the initial grasping force and the grasping force threshold to obtain a comparison result;

[0099] If the comparison result is that the initial grasping force is greater than the grasping force threshold, adjust the value corresponding to the initial grasping force to the value corresponding to the grasping force threshold.

[0100] Among them, the preset target object category is the object category that is preset and requires separate adjustment of the grasping force. For example, it can be easily damaged objects such as eggs and digital products. At this time, the grasping force threshold is 5N, that is, the maximum initial grasping force for such easily damaged objects cannot exceed 5N.

[0101] Based on this, in the actual grasping process, after determining the initial grasping force according to the mass and static friction coefficient of the object to be grasped, it is also necessary to determine whether the object is an easily damaged item or other objects that require corresponding adjustment of the grasping force. Therefore, by comparing the category of the object to be grasped and the target object category stored in the object information library, when the object to be grasped is the target object category, compare the size of the initial grasping force and the grasping force threshold, so that when the initial grasping force is greater than the grasping force threshold, reduce the numerical size of the initial grasping force to the value corresponding to the grasping force threshold, so as to ensure that the initial grasping force meets the actual grasping requirements of this category of objects; when the object is not the target object category or the grasping force is not greater than the grasping force threshold, the initial grasping force is not adjusted.

[0102] In some embodiments of the present application, through an image acquisition device such as a depth camera, obtain the initial image of the object to be grasped, so as to determine the category of the object according to the preset recognition model, so as to determine the category of the object according to the object information library, and then determine the corresponding density and static friction coefficient, and determine the volume of the object according to the depth camera, so as to determine the initial grasping force of the object, and according to the preset grasping force threshold, compare the category of the object to be grasped and the target object category stored in the object information library to determine that when the object to be grasped is the target object category, adjust the initial grasping force according to the preset grasping force threshold, so as to determine that the adjusted initial grasping force meets the actual grasping situation of the target object, and determine the grasping force increment of the object by determining the displacement and deformation amount of the object during the actual grasping process, and determine the real-time grasping force according to the initial grasping force and the grasping force increment.

[0103] In this way, during the process of object grasping, the grasping force of the object is adjusted in real time, the grasping success rate is improved, and the generality and flexibility of grasping are enhanced, enabling the robot to adapt to the actual grasping requirements of the object, and effectively improving the overall grasping efficiency and stability.

[0104] Figure 2 FIG. is a schematic flow chart of another method for generating a grasping force provided by an embodiment of the present application. As Figure 2 shown, the method for generating a grasping force may include the following steps:

[0105] S210. Determine an initial image of the object to be grasped according to a preset image acquisition device, and determine the category of the object to be grasped according to the initial image and a preset recognition model.

[0106] S220. Determine the grasping information of the object to be grasped according to a preset object information library and the category, and determine an initial grasping force corresponding to the object to be grasped according to the grasping information.

[0107] In some embodiments of the present application, the specific implementation manners of steps S210 to S220 may refer to the content in the foregoing embodiments, and will not be elaborated herein.

[0108] S230. Determine the marked points corresponding to the initial image.

[0109] Among them, the marked points are the marked points on the object to be grasped in the initial image, and the marked points are special points set on the object to facilitate operations such as object recognition, positioning, tracking, and measurement.

[0110] Based on this, by determining the marked points of the object to be grasped in the initial image, it is possible to subsequently determine whether the object to be grasped has a displacement and the corresponding displacement amount according to the positions of the marked points during the grasping process.

[0111] S240. Obtain continuously acquired object images according to the image acquisition device, and determine the marked points in each object image.

[0112] Among them, the object image is the object image of the object to be grasped acquired by an image acquisition device such as a depth camera during the grasping process.

[0113] Based on this, by determining the positions of the marked points on the object to be grasped in each object image, it is possible to subsequently determine the displacement amount.

[0114] S250. Perform optical flow method calculation on each marked point to obtain a displacement amount.

[0115] Among them, the optical flow method is a technology used in the fields of computer vision and image processing for estimating the motion of objects in images. Optical flow refers to the apparent motion of objects in an image, that is, due to the relative motion between the object and the camera, the positions of pixel points in the image change over time, forming a visual flow. By extracting the feature points in the image, that is, the marker points of the object to be grasped, and then calculating the optical flow by tracking the position changes of these marker points between different frames, it is possible to determine whether the object to be grasped has been displaced during the grasping process and the corresponding displacement amount.

[0116] Based on this, by determining the positions of the marker points of the object to be grasped in each object image during the actual grasping process, calculations are performed according to the optical flow method, so as to determine the corresponding displacement amount when the object is displaced.

[0117] S260. Perform real-time recognition on the object image based on the recognition model to obtain the deformation amount.

[0118] Based on this, the size change amount of the object to be grasped during the grasping process can be compared through a recognition model such as a large visual model, so as to determine the deformation amount.

[0119] S270. Determine the grasping force increment of the object to be grasped according to the displacement amount and the deformation amount.

[0120] S280. Determine the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force.

[0121] In some embodiments of the present application, the specific implementation manners of steps S270 to S280 can refer to the content in the foregoing embodiments and will not be elaborated herein.

[0122] Based on the feasible implementation manner of the above S270, the present application further provides steps for determining the grasping force increment of the object to be grasped according to the displacement amount and the deformation amount, including:

[0123] Determine the corresponding displacement threshold and deformation threshold according to the category of the object to be grasped;

[0124] Determine the corresponding reward value of the object to be grasped according to the displacement amount and the displacement threshold, as well as the deformation amount and the deformation threshold;

[0125] Compare the reward value with a preset reward threshold to obtain a comparison result;

[0126] If the comparison result is that the reward value is less than the reward threshold, determine a preset reward determination model, and determine the grasping force increment according to the displacement amount, the deformation amount, and the reward determination model;

[0127] If the comparison result is that the reward value is not less than the reward threshold, do not determine the grasping force increment.

[0128] Among them, the displacement threshold and the deformation threshold are the maximum critical values of displacement and deformation respectively.

[0129] The reward value is the value of the reward function determined according to the displacement and deformation amounts. The reward function is a key factor for evaluating the quality of the agent's behavior. According to the actions taken by the agent in the environment and the state changes feedback by the environment, a numerical reward signal is output for the agent. For example, if the reward function value is represented by R, the slip distance is Δx, and the deformation amount is ΔL, then R = 1 - Δx / threshold - ΔL / threshold; the preset reward threshold is used to compare with the reward value, which can be 0, that is, when R < 0, it is determined that the object has a dynamic change, and the increment of the grasping force needs to be determined, so as to adjust the initial grasping force in time.

[0130] The preset reward determination model is a pre-set model that can determine the increment of the grasping force based on predicting the future cumulative reward based on the state. It can be a Critic network model. The Critic network model is an important model in reinforcement learning. The Critic network is a neural network. Its input is the state of the environment (sometimes including the actions taken by the agent), and the output is an estimate of the value of the current state (or state-action pair). Simply put, it is responsible for evaluating the quality of the agent's behavior in the current state; the Critic network can make better decisions by learning a value function, and can continuously adjust its parameters according to the reward signal feedback by the environment, so that the estimate of the state value becomes more and more accurate.

[0131] Based on this, through reinforcement learning, dynamic adjustment is performed on the basis of the initial grasping force to obtain the increment of the grasping force.

[0132] On the basis of the feasible implementation manner of S270 above, the present application further provides that if the comparison result shows that the reward value is less than the reward threshold, then the preset reward determination model is determined, and the increment of the grasping force is determined according to the displacement amount, the deformation amount and the reward determination model, including the steps of:

[0133] According to the reward determination model, determine the predicted reward value corresponding to the displacement amount and the deformation amount;

[0134] According to the predicted reward value, determine the increment of the grasping force.

[0135] Among them, the predicted reward value is the predicted value of the displacement amount and the deformation amount determined by the reward determination model according to the prediction strategy.

[0136] Furthermore, in the robot grasping force training iteration, the following modules work together to determine the value of the grasping force:

[0137] 1) State input: The vision real-time collects the slip amount (Δx), the deformation amount (ΔL), and the end force state information, and converts them into vectors through an encoder.

[0138] 2) Reward function calculation:

[0139] The basic reward is calculated based on the deviation between the slip / deformation and the threshold (e.g., R = 1 - Δx / 3 - ΔL / 2);

[0140] Penalty term trigger: When Δx or ΔL exceeds the threshold, a large negative reward of -10 points is applied to force policy adjustment;

[0141] Additional reward: +10 points for successful grasping.

[0142] 3) Advantage function guidance: The Critic network predicts the future cumulative reward (V(s)) based on the state and calculates the advantage value A_t = Q(s,a) - V(s).

[0143] When A_t > 0 (the action is better than the baseline), the policy network increases the action probability corresponding to ΔF;

[0144] When A_t < 0 (the action is worse than the baseline), the probability is reduced and new actions are explored.

[0145] 4) Policy update amplitude control:

[0146] The PPO clipping mechanism is used to limit the update ratio to prevent a single ΔF adjustment from exceeding ±0.5N;

[0147] The entropy increase term (such as a coefficient of 0.001) encourages action diversity and avoids local optima;

[0148] Through the above collaborative mechanism, PPO balances the immediate reward and long-term stability in each iteration, and finally realizes the refined dynamic adjustment of the grasping force.

[0149] Based on this, through the reward determination model, the predicted values corresponding to the displacement and deformation amounts are determined, so as to determine the grasping force increment.

[0150] In some embodiments of the present application, through the visual large model combined with language instructions, the object attributes are dynamically parsed to generate the initial grasping force in real time; and the slip and deformation of the object during the grasping process are detected through high-speed images and recognition models, so as to dynamically fine-tune the object grasping according to reinforcement learning and optimize the grasping force in combination with the actual situation; in addition, in actual applications, after the object to be grasped is grasped, the category, shape, initial grasping force and grasping force increment of the object can be stored in the object information library. Thus, when grasping a new object to be grasped next time, if the grasping information of the new object to be grasped matches the information already existing in the information library, the corresponding initial grasping force and grasping force increment can be called, which improves the grasping speed.

[0151] Figure 3 This is a schematic structural diagram of a grasping force generation device 300 provided by an embodiment of the present application. As Figure 3As shown in the figure, the generating device 300 for the grasping force includes: an image determination module 310, an information determination module 320, an object grasping module 330, an increment determination module 340, and a grasping force determination module 350; where:

[0152] The image determination module 310 is configured to determine an initial image of the object to be grasped according to a preset image acquisition device, and determine the category of the object to be grasped according to the initial image and a preset recognition model.

[0153] The information determination module 320 is configured to determine the grasping information of the object to be grasped according to a preset object information library and the category, and determine the initial grasping force corresponding to the object to be grasped according to the grasping information.

[0154] The object grasping module 330 is configured to grasp the object to be grasped according to the initial grasping force, and determine the displacement and deformation amount of the object to be grasped according to the image acquisition device and the recognition model.

[0155] The increment determination module 340 is configured to determine the grasping force increment of the object to be grasped according to the displacement and deformation amount.

[0156] The grasping force determination module 350 is configured to determine the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force.

[0157] In an embodiment of the present application, the information determination module 320 may also specifically be configured to:

[0158] Determine a preset object information library;

[0159] According to the object information library and the category, determine the density information and the corresponding static friction coefficient of the object to be grasped;

[0160] According to the image acquisition device, determine the volume information of the object to be grasped, and determine the grasping information according to the density information, the static friction coefficient, and the volume information;

[0161] According to the grasping information, determine the initial grasping force.

[0162] In an embodiment of the present application, the information determination module 320 may also specifically be configured to:

[0163] Determine the density information, the static friction coefficient, and the volume information in the grasping information, and determine the mass information corresponding to the object to be grasped according to the density information and the volume information;

[0164] Determine a preset grasping force threshold;

[0165] According to the mass information, the static friction coefficient, and the grasping force threshold, determine the initial grasping force.

[0166] In the embodiment of the present application, the information determination module 320 may further specifically be configured to:

[0167] Determine an initial grasping force according to the quality information and the static friction coefficient;

[0168] Determine a preset target object category according to the object information library;

[0169] Judge whether the category of the object to be grasped is the target object category;

[0170] If the category of the object to be grasped is the target object category, compare the initial grasping force with the grasping force threshold to obtain a comparison result;

[0171] If the comparison result is that the initial grasping force is greater than the grasping force threshold, adjust the value corresponding to the initial grasping force to the value corresponding to the grasping force threshold.

[0172] In the embodiment of the present application, the object grasping module 330 may further specifically be configured to:

[0173] Determine the marked points corresponding to the initial image;

[0174] Obtain the continuously acquired object images according to the image acquisition device, and determine the marked points in each object image;

[0175] Perform optical flow method calculation on each marked point to obtain a displacement amount;

[0176] Perform real-time recognition on the object image based on the recognition model to obtain a deformation amount.

[0177] In the embodiment of the present application, the increment determination module 340 may further specifically be configured to:

[0178] Determine the corresponding displacement threshold and deformation threshold according to the category of the object to be grasped;

[0179] Determine the reward value corresponding to the object to be grasped according to the displacement amount, the displacement threshold, the deformation amount, and the deformation threshold;

[0180] Compare the reward value with the preset reward threshold to obtain a comparison result;

[0181] If the comparison result is that the reward value is less than the reward threshold, determine the preset reward determination model, and determine the grasping force increment according to the displacement amount, the deformation amount, and the reward determination model;

[0182] If the comparison result is that the reward value is not less than the reward threshold, do not determine the grasping force increment.

[0183] In the embodiment of the present application, the increment determination module 340 may further specifically be configured to:

[0184] Determine the predicted reward values corresponding to the displacement and deformation amounts according to the reward determination model;

[0185] Determine the grasping force increment according to the predicted reward values.

[0186] Figure 4 It is a schematic structural diagram of the device provided in the embodiment of the present application. As Figure 4 shown, the device 400 includes:

[0187] The device 400 may include a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a communication component 403, and other components. Among them, the processor 401, the memory 402, and the communication component 403 are connected through a bus 404.

[0188] In a specific implementation process, at least one processor 401 executes the computer execution instructions stored in the memory 402, so that at least one processor 401 executes the above-mentioned method for generating the grasping force.

[0189] For the specific implementation process of the processor 401, reference can be made to the above method embodiment, and its implementation principle and technical effects are similar, which will not be elaborated here in this embodiment.

[0190] Furthermore, the processor may be a central processing unit (English: Central Processing Unit, abbreviated as: CPU), or may also be other general-purpose processors, digital signal processors (English: Digital Signal Processor, abbreviated as: DSP), application specific integrated circuits (English: Application Specific Integrated Circuit, abbreviated as: ASIC), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the present application can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0191] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.

[0192] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience in representation, the buses in the drawings of this application are not limited to only one bus or one type of bus.

[0193] In some embodiments, a computer program product is also proposed, including a computer program or instruction, which, when executed by a processor, implements the steps in any of the above-mentioned methods for generating a grasping force.

[0194] For the specific implementation of each of the above operations, reference can be made to the previous embodiments and will not be elaborated here.

[0195] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions or by controlling relevant hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0196] Therefore, an embodiment of this application provides a computer-readable storage medium, in which multiple program codes are stored. The program codes can be loaded by a processor to execute the steps in any of the methods for generating a grasping force provided by the embodiments of this application.

[0197] Among them, the storage medium can include: a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disc, or the like.

[0198] According to one aspect of this application, a computer program product or a computer program is provided. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium.

[0199] Since the instructions stored in the storage medium can execute the steps in any of the methods for generating a grasping force provided by the embodiments of this application, the beneficial effects that can be achieved by any of the methods for generating a grasping force provided by the embodiments of this application can be realized. For details, refer to the previous embodiments and will not be elaborated here.

[0200] Other embodiments of the present application will be readily apparent to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and examples are only illustrative, and the true scope and spirit of the present application are pointed out by the claims above.

[0201] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A method for generating a grasping force, characterized in that, The method includes: Determine an initial image of the object to be grasped according to a preset image acquisition device, and determine the category of the object to be grasped according to the initial image and a preset recognition model; Determine the grasping information of the object to be grasped according to a preset object information library and the category, and determine the corresponding initial grasping force of the object to be grasped according to the grasping information; Grasp the object to be grasped according to the initial grasping force, and determine the displacement and deformation amount of the object to be grasped according to the image acquisition device and the recognition model; Determine the corresponding displacement threshold and deformation threshold according to the category of the object to be grasped; Determine the corresponding reward value of the object to be grasped according to the displacement amount and the displacement threshold, and the deformation amount and the deformation threshold; Compare the reward value with a preset reward threshold to obtain a comparison result; If the comparison result is that the reward value is less than the reward threshold, determine a preset reward determination model, and determine the predicted reward value corresponding to the displacement amount and the deformation amount according to the reward determination model; Determine the grasping force increment according to the predicted reward value; If the comparison result is that the reward value is not less than the reward threshold, do not determine the grasping force increment; Determine the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force.

2. The method according to claim 1, characterized in that, The step of determining the grasping information of the object to be grasped according to a preset object information library and the category, and determining the corresponding initial grasping force of the object to be grasped according to the grasping information includes: Determine the preset object information library; Determine the density information and the corresponding static friction coefficient of the object to be grasped according to the object information library and the category; Determine the volume information of the object to be grasped according to the image acquisition device, and determine the grasping information according to the density information, the static friction coefficient and the volume information; Determine the initial grasping force according to the grasping information.

3. The method according to claim 2, wherein The step of determining the initial grasping force according to the grasping information includes: Determine the density information, the static friction coefficient and the volume information in the grasping information, and determine the corresponding mass information of the object to be grasped according to the density information and the volume information; Determine a preset grasping force threshold; Determine the initial grasping force according to the mass information, the static friction coefficient and the grasping force threshold.

4. The method according to claim 3, characterized in that, The step of determining the initial grasping force according to the mass information, the static friction coefficient and the grasping force threshold includes: Determine the initial grasping force according to the mass information and the static friction coefficient; Determine a preset target object category according to the object information library; Judge whether the category of the object to be grasped is the target object category; If the category of the object to be grasped is the target object category, compare the size of the initial grasping force and the grasping force threshold to obtain a comparison result; If the comparison result is that the initial grasping force is greater than the grasping force threshold, adjust the value corresponding to the initial grasping force to the value corresponding to the grasping force threshold.

5. The method according to claim 1, wherein Grasping the object to be grasped according to the initial grasping force, and determining the displacement and deformation of the object to be grasped according to the image acquisition device and the recognition model, including: Determining the marked points corresponding to the initial image; Acquiring the continuously acquired object images according to the image acquisition device, and determining the marked points in each of the object images; Performing optical flow method calculation on each of the marked points to obtain the displacement; Performing real-time recognition on the object images based on the recognition model to obtain the deformation.

6. A gripping force generating device, characterized in that, The device includes: An image determination module, configured to determine the initial image of the object to be grasped according to a preset image acquisition device, and determine the category of the object to be grasped according to the initial image and a preset recognition model; An information determination module, configured to determine the grasping information of the object to be grasped according to a preset object information library and the category, and determine the initial grasping force corresponding to the object to be grasped according to the grasping information; An object grasping module, configured to grasp the object to be grasped according to the initial grasping force, and determine the displacement and deformation of the object to be grasped according to the image acquisition device and the recognition model; A threshold determination module, configured to determine the corresponding displacement threshold and deformation threshold according to the category of the object to be grasped; A reward value determination module, configured to determine the reward value corresponding to the object to be grasped according to the displacement and the displacement threshold, and the deformation and the deformation threshold; A comparison module, configured to compare the reward value with a preset reward threshold to obtain a comparison result; A predicted reward value determination module, configured to, if the comparison result is that the reward value is less than the reward threshold, determine a preset reward determination model, and determine the predicted reward value corresponding to the displacement and the deformation according to the reward determination model; An increment determination module, configured to determine the grasping force increment according to the predicted reward value; A comparison result determination module, configured to, if the comparison result is that the reward value is not less than the reward threshold, not determine the grasping force increment; A grasping force determination module, configured to determine the real-time grasping force of the object to be grasped according to the grasping force increment and the initial grasping force.

7. A device, characterized in that, Including: One or more processors; A memory; One or more programs, where one or more programs are stored in the memory and configured to be executed by one or more processors, and one or more programs are configured to execute the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, Program code is stored in a computer-readable storage medium, and the program code can be called by a processor to execute the method according to any one of claims 1 to 5.

Images