Intelligent grabbing method and system of mechanical arm based on internet of things

Abstract

The application discloses an intelligent grabbing method and system of a mechanical arm based on an Internet of Things, relates to the technical field of mechanical arm grabbing control, and constructs a digital twin model integrating a mechanical arm, a to-be-grabbed object and an environment through the Internet of Things.The system and method quantize an arm health risk index (integrating joint load rate, vibration entropy and other mechanical arm state parameters) and a target grabbing risk index (integrating grabbing offset, stacking stability and other task-related parameters) through multi-dimensional indexes, comprehensively capture and simulate potential risks in a grabbing process of each grabbing area, dynamically determine an optimal grabbing area based on a comprehensive risk index, guide the mechanical arm to grab in a sequence for different grabbing areas, prolong the service life of the mechanical arm while improving grabbing efficiency (reducing invalid operations), and realize the collaborative optimization of safety and efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical arm grabbing control, more particularly, it relates to an intelligent grabbing method and system of a mechanical arm based on Internet of Things. BACKGROUND

[0002] In the current industrial automation field, the mechanical arm scattered object grabbing faces the common challenge that safety and efficiency are difficult to be optimized simultaneously. The traditional grabbing strategy focuses on a single dimension index, or only pays attention to the running state of the mechanical arm body (such as joint load, temperature and other basic parameters), or simply considers the physical characteristics of the grabbing target (such as position offset, stacking form and the like), lacking the systematic fusion analysis of the device health loss and task execution risk. Especially in the complex stacking scene, the mechanical arm frequently enters the high load posture area for work, not only accelerates the wear of the key components, but also may cause material collapse due to improper grabbing sequence, resulting in work interruption and even device damage. The existing technology has not established a dynamic risk assessment model covering the whole element of "mechanical arm-grabbing object-environment", and it is difficult to realize the multi-objective coordinated optimization of the grabbing process.

[0003] The existing grabbing control system generally has two limitations: in the aspect of mechanical arm health management, a passive protection mechanism such as threshold alarm is adopted, which cannot predict the cumulative damage risk combined with the characteristics of the working area; in the aspect of grabbing decision, it depends on static priority rules or local optimization algorithm, and cannot quantitatively evaluate the chain effect of different grabbing sequences on the overall stacking stability. Such a fragmented control logic leads to two key defects: one is that the service life of the mechanical arm is significantly reduced due to repeated high load actions in the region, and the other is that the grabbing efficiency is limited by frequent obstacle avoidance adjustment and secondary sorting operation.

[0004] Based on the above, the present application provides an intelligent grabbing method and system of a mechanical arm based on Internet of Things. SUMMARY

[0005] In view of the deficiencies of the prior art, the purpose of the present application is to provide an intelligent grabbing method and system of a mechanical arm based on Internet of Things.

[0006] To achieve the above purpose, the present application provides the following technical scheme:

[0007] An intelligent grabbing method of a mechanical arm based on Internet of Things, the steps are as follows:

[0008] Step 1: Before the mechanical arm works, a grabbing scene twin model is constructed by using the physical scene data collected by the Internet of Things sensor;

[0009] Step 2: The working space of the mechanical arm is divided into working areas, and the grabbing area where the object to be grabbed exists is marked;

[0010] Step three: for each grabbing area, calculate the arm health risk index and the target grabbing risk index in real time;

[0011] Step four: calculate the optimal grabbing area, and control the robot arm to grab the to-be-grabbed object in the optimal grabbing area.

[0012] Further, an intelligent grabbing system of a robot arm based on the Internet of Things includes a grabbing scene twin unit, which is used to construct a grabbing scene twin model before the robot arm starts grabbing work;

[0013] A grabbing area risk analysis unit is used to divide the workspace of the robot arm in the grabbing scene twin model into multiple work areas, mark the work area where the to-be-grabbed object exists as a grabbing area, and define the grabbing comprehensive risk index of each grabbing area in real time during the work process of the robot arm.

[0014] The definition process of the grabbing comprehensive risk index of the grabbing area: select a grabbing area, obtain the arm health risk index and the target grabbing risk index of the grabbing area, and calculate the grabbing comprehensive risk index of the grabbing area based on the arm health risk index and the target grabbing risk index.

[0015] A robot arm grabbing control unit marks the grabbing area with the smallest grabbing comprehensive risk index value as the optimal grabbing area, and controls the robot arm to preferentially grab the to-be-grabbed object in the optimal grabbing area.

[0016] Further, the grabbing scene twin model includes a robot arm digital twin model, a to-be-grabbed object digital twin model, and a grabbing environment digital twin model, and the grabbing scene twin model is constructed by fusing the robot arm digital twin model, the to-be-grabbed object digital twin model, and the grabbing environment digital twin model.

[0017] Further, the grabbing area arm health risk index acquisition process: select a grabbing area, determine all possible poses of the end of the robot arm in the grabbing area in the grabbing scene twin model, obtain the joint load rate and vibration entropy of each possible pose, set a joint load rate threshold, when the joint load rate of a possible pose is higher than the joint load rate threshold, increase the high-load possibility frequency by one, compare all possible poses in pairs, calculate the absolute difference of the joint load rates of the two compared possible poses, calculate the joint load fluctuation rate, calculate the average joint load fluctuation rate by summing and averaging all joint load fluctuation rates, calculate the average vibration entropy by summing and averaging all possible poses, and calculate the arm health risk index of the grabbing area based on the high-load possibility frequency, the average joint load fluctuation rate, and the average vibration entropy.

[0018] Further, the acquisition process of the joint load rate of the possible posture: select a possible posture, input the pose parameters of the possible posture by using KDL, and reversely calculate the joint angle combination satisfying the pose by using the inverse kinematics algorithm. Each joint angle included in the joint angle combination is input into the multi-body dynamics model of the robot arm, the torque output by each joint is calculated, the load rate of each joint is further calculated, and the load rates of each joint are weighted and summed to calculate the joint load rate of the possible posture.

[0019] Further, the acquisition process of the target grasping risk index R2 of the grasping area: select a grasping area, determine all first-graspable objects in the grasping area, further acquire the grasping offset of each first-graspable object, sum and average the grasping offsets of all first-graspable objects to calculate the average grasping offset, acquire the grasping stack stability index of each first-graspable object, set a grasping stack stability threshold index, mark the first-graspable object as a reasonable grasping object when the grasping stack stability index of the first-graspable object is higher than the grasping stack stability threshold index, and calculate the target grasping risk index of the grasping area based on the average grasping offset and the total number of reasonable grasping objects.

[0020] Further, the acquisition process of the grasping offset of the first-graspable object: select a first-graspable object, simulate grasping of the first-graspable object in the grasping area by the gripper of the robot arm based on the multi-body dynamics model, acquire the center of gravity coordinates of the first-graspable object and the center coordinates of the gripper after the simulation grasping, and calculate the grasping offset of the first-graspable object based on the center of gravity coordinates of the first-graspable object and the center coordinates of the gripper.

[0021] Further, the acquisition process of the grasping stack stability index of the first-graspable object: select a first-graspable object, simulate grasping of the first-graspable object in the grasping area by the gripper of the robot arm based on the multi-body dynamics model, identify the stacking relationship of all remaining graspable objects after the simulation grasping, calculate the stability coefficient of each remaining graspable object, sum and average the stability coefficients of all remaining graspable objects to calculate the grasping stack stability index.

[0022] Compared with the prior art, the present application has the following advantages:

[0023] The system and method of the application build a digital twin model integrating a mechanical arm, a to-be-grabbed object and an environment through the Internet of Things, quantize an arm health risk index (integrating joint load rate, vibration entropy and other mechanical arm state parameters) and a target grabbing risk index (integrating grabbing offset, stacking stability and other task-related parameters) through multi-dimensional indexes, comprehensively capture and simulate potential risks in the grabbing process of each grabbing area, dynamically determine an optimal grabbing area based on a comprehensive risk index, guide the grabbing order of the mechanical arm for different grabbing areas, prolong the service life of the mechanical arm while improving the grabbing efficiency (reducing invalid operations), and realize the cooperative optimization of safety and efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0024] Fig. 1 is a principle block diagram of the application;

[0025] Fig. 2 is a flowchart of acquisition of the arm health risk index of the grabbing area;

[0026] Fig. 3 is a flowchart of acquisition of the target grabbing risk index of the grabbing area. DETAILED DESCRIPTION

[0027] Embodiment one: refer to Figs. 1-3 An intelligent grabbing system of a mechanical arm based on the Internet of Things, comprising a grabbing scene twin unit, a grabbing area risk analysis unit and a mechanical arm grabbing control unit.

[0028] The grabbing scene twin unit, before the mechanical arm grabbing work starts, builds a grabbing scene twin model (the grabbing scene twin model comprises a mechanical arm digital twin model, a to-be-grabbed object digital twin model and a grabbing environment digital twin model, and the mechanical arm digital twin model, the to-be-grabbed object digital twin model and the grabbing environment digital twin model are fused to build the grabbing scene twin model).

[0029] The construction process of the robot digital twin model: Step 1: robot data acquisition (completed by multi-modal sensors, including joint level, global level, thermal management, joint level: torque sensor (accuracy ±0.01 N•m), magnetic encoder (17-bit accuracy), MEMS vibration sensor (5 kHz sampling); global level: laser radar (Speedten Creative RS-LiDAR-M1) scans the overall shape of the robot, UWB positioning base station (Decawave DWM1000) realizes sub-meter positioning; thermal management: infrared thermal imager (FLIRA65) monitors the motor temperature field in real time (sampling rate 10 Hz)); Step 2: generate high-precision grid models of each joint of the robot (STL format, surface number ≥100,000) through reverse engineering (3D scanning + CAD reconstruction); use LOD (Level of Detail) technology to generate lightweight models (surface number ≤50,000), which are suitable for edge real-time simulation. Based on the improved DH parameter method, the joint coordinate system is established, and the end position is calculated in real time through the KDL library (error <0.1°); introduce the "flexible joint compensation model" (considering the elastic deformation of the joint), dynamically correct the kinematic error (accuracy improved by 20%). Based on Newton-Euler equation, derive the multi-body dynamics equation, and fuse the measured friction (Coulomb friction + viscous friction), inertia tensor data; use model predictive control (MPC) algorithm, real-time update joint torque prediction value (error <5% with actual value); Step 3: synchronize the joint angle and speed of the physical robot to the virtual model every 5 ms, drive the virtual robot to move; input the measured joint load rate and motor temperature into the virtual model, correct the dynamics simulation parameters (such as the friction coefficient decreases with the increase of temperature).

[0030] The construction process of the digital twin model of the object to be grabbed: Step 1: data collection of the object to be grabbed (visual perception: binocular RGB-D camera reconstructs the object's three-dimensional point cloud at 30 fps, and generates a mesh model (accuracy ±0.1 mm) through Poisson surface reconstruction; physical property library construction: object ID is read through a code scanning gun or RFID, and pre-stored properties (such as glass fragility coefficient 0.9, rubber 0.2) are retrieved); Step 2: geometric model: point cloud data is segmented into basic shapes such as planes and cylinders through RANSAC algorithm, generating parameterized models (such as cubes and spheres); stacked objects are extracted through the convex hull algorithm to obtain the overall bounding box, and the "easy-to-slide area" (contact area <50%) is marked. Physical model: based on the discrete element method (DEM), a particle model is constructed to simulate the contact mechanics behavior of the object during stacking (such as the collision and splashing of glass fragments); introduce "brittleness-force feedback mapping" (brittleness coefficient x contact force = maximum allowed deformation), predict the risk of breakage during grabbing. Step 3: position and attitude synchronization: real-time update the object's pose in the virtual space through visual SLAM (such as ORB-SLAM3); physical property synchronization: during the grabbing process, the contact force data of the force sensor is fed back to the virtual model, and the elastic modulus of the object is dynamically adjusted (such as the modulus of rubber temporarily increases by 30% after being compressed).

[0031] The construction process of the digital twin model of the grabbing environment: Step 1: environment data collection (including static environment data and dynamic environment data, static environment data: laser scanning fixed obstacles (shelves, walls) to generate three-dimensional point cloud; infrared ranging sensor is deployed along the edge of the workspace to mark the boundary of the restricted area; dynamic environment data: UWB positioning base station tracks dynamic targets such as AGV and human bodies in real time; MEMS vibration sensor collects environmental vibration amplitude, and light sensor monitors light intensity); Step 2: reconstruct the laser radar point cloud into a high-precision mesh model (STL format, face number ≥300,000) through reverse engineering, and generate a lightweight model (face number ≤50,000) using LOD technology to adapt to the edge; obstacles are semantically labeled (such as "metal shelf" and "plastic box") and associated with threat level parameters. Based on UWB data, construct a dynamic target trajectory prediction model (LSTM algorithm, predict path 1-3 seconds in advance); fuse vibration and light data to establish an "environmental parameter-interference intensity" mapping (such as vibration > 0.5g, mark as "high interference area"); Step 3: synchronize dynamic target position, vibration intensity, and light data to the virtual model every 100 ms to drive the environment model update; measured environmental parameters (such as the actual trajectory deviation of AGV from the predicted trajectory) are used to correct the virtual model parameters (such as trajectory prediction coefficients), ensuring that the coordinate error of static obstacles is less than 5 cm and the trajectory deviation of dynamic targets is less than 20 cm.

[0032] ​The fusion process of the digital twin models of the robotic arm, the object to be grasped, and the grasping environment is as follows: Spatial reference unification: Using the robotic arm base as the origin of the global coordinate system, the virtual models of the three are mapped to the same coordinate system through the coordinate transformation matrix of UWB positioning (robotic arm), visual SLAM (object to be grasped), and LiDAR (environment), ensuring that the spatial position error of static obstacles, objects, and the robotic arm is <5cm. Time synchronization: The PTP protocol is used to unify the clocks of sensors and edge devices (error <1μs). The robotic arm joint data (updated every 5ms), object pose (updated every 30ms), and environmental dynamic data (updated every 100ms) are aligned by timestamp to avoid timing misalignment. Static data association: The robotic arm workspace parameters (such as maximum extension radius) are matched with the coordinates of static obstacles in the environment, and "robotic arm inaccessible areas" (such as areas obscured by shelves) are marked; the attributes of the object to be grasped (such as brittleness coefficient) are associated with environmental parameters (such as vibration intensity). Dynamic data interaction: The dynamic data of the three are integrated in real time—environmental vibration intensity corrects the positioning accuracy of the robotic arm end effector (vibration > When the positioning error is +0.3mm, a collision risk warning is triggered when the robotic arm's movement path intersects with the trajectory of a dynamic environmental target (such as an AGV). The object's pose offset is combined with the ambient light intensity to correct the visual recognition confidence (confidence ×0.8 when the light intensity is <500 lux). Geometric collision detection: The collision risk between the robotic arm and environmental obstacles, between the robotic arm and objects, and between objects and the environment is monitored in real time using the GJK algorithm (e.g., the robotic arm decelerates when it is <0.3m from the shelf, and marks "easy to slip" when the contact area of ​​stacked objects is <50%). Behavioral collaborative simulation: The entire process of "robotic arm movement - grasping objects - dealing with environmental interference" is simulated in virtual space - the robotic arm adjusts its grasping speed according to environmental vibration (speed ×0.5 when there is high vibration); when grasping objects, the visually guided gripper angle is corrected in combination with the ambient light (fine-tuning is done by force feedback when the light intensity is insufficient).

[0033] The grasping area risk analysis unit divides the workspace of the robotic arm in the grasping scenario twin model into multiple work areas (the workspace of the robotic arm refers to the set of all spatial locations that the end effector (such as gripper) can reach). The work area containing the object to be grasped is marked as the grasping area (the number of objects to be grasped in the grasping area is not limited to one, and multiple objects to be grasped may be stacked or interact with each other). During the operation of the robotic arm, the comprehensive grasping risk index of each grasping area is defined in real time.

[0034] The robotic arm grasping control unit marks the grasping area with the lowest comprehensive grasping risk index value as the superior grasping area, and controls the robotic arm to prioritize grasping the object to be grasped in the superior grasping area.

[0035] The definition process of the comprehensive risk index of the grabbing area: selecting a grabbing area, obtaining the arm health risk index R1 and the target grabbing risk index R2 of the grabbing area, and calculating the comprehensive risk index of the working area by u1 is the arm health risk coefficient, u2 is the target grabbing risk coefficient, The values of u1 and u2 are different in different scenarios. For example, the coefficients of the industrial scenario are as follows: u1 is 0.5, and u2 is 0.5 (because in the industrial scenario, the arm health and grabbing stacking are important (such as high collision risk of AGV in the workshop)).

[0036] The acquisition process of the arm health risk index R1 of the grabbing area: selecting a grabbing area, determining all possible poses of the end of the robotic arm in the grabbing area in the grabbing scene twin model (the possible pose is all reachable pose combinations of the end of the robotic arm in the grabbing area), obtaining the joint load rate and vibration entropy of each possible pose, setting a joint load rate threshold (the joint load rate threshold is set based on the rated load of the hardware and the fatigue life of the joint material), and increasing the high-load possibility number by one when the joint load rate of the possible pose is higher than the joint load rate threshold (no need to increase the high-load possibility number if it is not higher than the threshold). Finally, the high-load possibility number is marked as The arm health risk index R1 of the grabbing area is calculated by

[0037] The acquisition process of the joint load rate of the possible pose: selecting a possible pose, inputting the pose parameters (position + orientation) of the possible pose by KDL, and calculating the joint angle combination that satisfies the pose by inverse kinematics algorithm (if it is a 6-axis, the joint angle combination contains ​​​​​), each joint angle included in the joint angle combination is input into the multi-body dynamics model of the robotic arm (constructed based on Newton-Euler equation), the torque required to be output by each joint is calculated, and the load rate of each joint is further calculated (e.g., the large arm joint requires 12 N•m, and the rated torque of the large arm joint is 15 N•m → the load rate is 80%), the load rates of each joint are weighted and summed to calculate the joint load rate of the possible pose.

[0038] The acquisition process of the vibration entropy of the possible pose is as follows: a possible pose is selected, the KDL is adopted, the pose parameters of the possible pose are input, the joint angle combination satisfying the pose is backstepped through the inverse kinematics algorithm, each joint angle included in the joint angle combination is input into the multi-body dynamics model of the robotic arm, the robotic arm is controlled to move according to the input joint angle, the vibration acceleration signals of the key positions of the transmission chain (such as the input shaft of the reducer and the large arm joint) are output, the vibration acceleration signals are converted into frequency domain spectrum through fast Fourier transform (FFT) to obtain the "frequency-energy" distribution (e.g., 120 Hz corresponds to energy , 200 Hz corresponds to ), the frequency range is divided into n characteristic frequency bands (e.g., 50-200 Hz, 200-500 Hz, and 500-1000 Hz), the energy of each characteristic frequency band is calculated (the integral energy of each characteristic frequency band is obtained), the energy proportion of each characteristic frequency band is calculated , , i is the serial number of each characteristic frequency band, n is the total number of characteristic frequency bands, and the vibration entropy of the possible pose is calculated through .

[0039] The acquisition process of the target grasping risk index R2 of the grasping region is as follows: a grasping region is selected, all first-graspable objects in the grasping region are determined (the first-graspable object is a to-be-grasped object that can be currently grasped in the grasping region, for example, the bottom object A supports the middle object B, and the middle object B supports the upper objects C and D, so the upper objects C and D are to-be-grasped objects that can be currently grasped, and the upper objects C and D are first-graspable objects), the grasping offset of each first-graspable object is further acquired, the grasping offsets of all first-graspable objects are summed and averaged to calculate the average grasping offset , the grasping stack stability index of each first-graspable object is acquired, the grasping stack stability threshold index is set (the grasping stack stability threshold index is set based on the scene risk tolerance), when the grasping stack stability index of the first-graspable object is higher than the grasping stack stability threshold index, the first-graspable object is marked as a reasonable grasping object (if it is not higher than the threshold, it does not need to be marked), and the total number of reasonable grasping objects is marked as​ , by calculating a target grasping risk index R2 of the grasping area, wherein h1 is a grasping offset coefficient, and h2 is a reasonable grasping quantity coefficient, and in an industrial assembly scene, h1 can be 0.6, and h2 can be 0.4 (accuracy dominant, quantity compensation weak).

[0040] The acquisition process of the grasping offset of the first-graspable object can be as follows: selecting a first-graspable object, controlling the gripper of the robot arm to simulate grasping the first-graspable object in the grasping area based on a multi-body dynamics model (such as ADAMS), after the simulation grasping, obtaining the coordinates (x g , y g , z g ) of the center of gravity of the first-graspable object and the coordinates (x c , y c , z c ) of the center of the gripper, and calculating the grasping offset of the first-graspable object by the formula: grasping offset .

[0041] The acquisition process of the grasping stack stability index of the first-graspable object can be as follows: selecting a first-graspable object, controlling the gripper of the robot arm to simulate grasping the first-graspable object in the grasping area based on a multi-body dynamics model, after the simulation grasping, identifying the stacking relationship of all remaining graspable objects (remaining graspable objects are graspable objects that have not been grasped), calculating the stability coefficient of each remaining graspable object (the stability coefficient is the ratio of the stability moment to the overturning moment of the remaining graspable object), and calculating the grasping stack stability index by summing and averaging the stability coefficients of all remaining graspable objects.

[0042] The acquisition process of the stability moment of the remaining graspable object (moment resisting overturning): the stability moment is generated by the support contact force between the remaining objects, reflecting the ability of the stacking structure to resist collapse. The acquisition process is as follows: 1. Stacking topology identification (virtual layer): in ADAMS, the support relationship of the remaining objects is identified (such as object B is supported by object A, object C is supported by object B) by a contact detection algorithm (such as the GJK algorithm), and a "support tree" (A→B→C) is generated; marking the support contact points of each object (such as the contact point between object B and A is , and the contact point between object C and B is ; 2. Contact force calculation (virtual layer + physical layer): virtual simulation: in ADAMS, set the contact model (such as "face-face contact"), define the contact parameters (friction coefficient 0.6, elastic modulus 2GPa), and simulate the size , and direction (perpendicular to the contact surface) of the contact force; 3. Stability moment synthesis (algorithm layer): the stability moment is the sum of the moments of all support contact forces on the overturning axis: ; Support force of the a-th contact point (such as ); Vertical distance from the contact force line of action to the overturning axis (such as ).

[0043] The acquisition process of the overturning moment of the remaining objects to be grabbed (the moment causing the collapse): the overturning moment is generated by the gravity of the remaining objects, reflecting the driving force for the collapse of the stacked structure. 1. Calculation of the center of gravity of the remaining objects (virtual layer + physical layer): virtual simulation: through the mass attribute module of ADAMS, the center of gravity coordinates of each remaining object (such as the center of gravity of object B (x=0.1m, y=0.2m, z=0.3m)) are calculated; 2. Synthesis of overturning moment (algorithm layer): the overturning moment is the sum of the moments of the gravity of all remaining objects on the overturning axis: ; Wherein: The gravity of the z-th object (such as the weight of object B ); The horizontal distance from the center of gravity of the object to the overturning axis (such as ).

[0044] The above system builds a digital twin model integrating the mechanical arm, the objects to be grabbed, and the environment through the Internet of Things, quantifies the arm health risk index (integrates joint load rate, vibration entropy, and other mechanical arm state parameters) and the target grabbing risk index (integrates grabbing offset, stacking stability, and other task-related parameters) through multiple dimensions, comprehensively captures and simulates the potential risks in the grabbing process of each grabbing area, dynamically determines the optimal grabbing area based on the comprehensive risk index, and guides the mechanical arm to grab in different grabbing sequences. In order to improve the grabbing efficiency (reduce invalid operations), the service life of the mechanical arm is prolonged, and the safety and efficiency are optimized.

[0045] Embodiment two: an intelligent grabbing method of a mechanical arm based on the Internet of Things, the steps are as follows:

[0046] Step one: before the mechanical arm works, use the physical scene data collected by the Internet of Things sensor to build a grabbing scene twin model;

[0047] Step two: divide the working space of the mechanical arm into working areas, and mark the grabbing areas where there are objects to be grabbed;

[0048] Step three: for each grabbing area, calculate the arm health risk index and the target grabbing risk index in real time;

[0049] Step four: calculate the optimal grabbing area, and control the mechanical arm to grab the objects to be grabbed in the optimal grabbing area.

[0050] The above method is based on the dynamic optimization area judgment mechanism of the comprehensive risk index, realizes the intelligent scheduling of the grabbing sequence, preferentially selects the low-risk area, reduces the grabbing failure and mechanical arm failure (such as invalid retry and joint wear aggravation) caused by high-risk operation, and achieves the multiple optimization goals of "controllable risk, efficiency improvement, and life extension".

[0051] The above formulas are dimensionless values calculated, the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0052] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center and the like containing one or more available medium sets. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD) or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0053] It should be understood that in various embodiments of the present application, the size of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0054] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0055] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0056] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are merely schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0057] The functions if realized in the form of software function units and sold or used as independent products can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art or the part of the technical solutions of the present application can be embodied in the form of software product, and the computer software product is stored in a storage medium, including a plurality of instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0058] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent grabbing method of an Internet of Things based mechanical arm, characterized in that, The steps are as follows: Step one: before the mechanical arm works, the physical scene data collected by the Internet of Things sensor is used to construct a grasping scene twin model; Step two: the working space of the mechanical arm is divided into working areas, and the grasping area where the to-be-grasped object exists is marked; Step three: for each grasping area, the arm health risk index and the target grasping risk index are calculated in real time; The acquisition process of the arm health risk index of the grasping area: select a grasping area, determine all possible poses of the end of the mechanical arm in the grasping area in the grasping scene twin model, obtain the joint load rate and vibration entropy of each possible pose, set the joint load rate threshold, when the joint load rate of the possible pose is higher than the joint load rate threshold, increase the high load possibility times once, compare all possible poses two by two, calculate the absolute difference value of the joint load rate of the two compared possible poses, calculate the joint load fluctuation rate, calculate the average joint load fluctuation rate by summing all joint load fluctuation rates, calculate the average vibration entropy by summing all possible poses, and calculate the average vibration entropy based on the high load possibility times, the average joint load fluctuation rate and the average vibration entropy to calculate the arm health risk index of the grasping area; The acquisition process of the target grasping risk index of the grasping area: select a grasping area, determine all first-graspable objects in the grasping area, further obtain the grasping offset of each first-graspable object, calculate the average grasping offset by summing the grasping offsets of all first-graspable objects, obtain the grasping stack stability index of each first-graspable object, set the grasping stack stability threshold index, when the grasping stack stability index of the first-graspable object is higher than the grasping stack stability threshold index, mark the first-graspable object as a reasonable grasping object, and calculate the target grasping risk index of the grasping area based on the average grasping offset and the total number of reasonable grasping objects; Step four: calculate the optimal grasping area, and control the mechanical arm to grasp the to-be-grasped object in the optimal grasping area.

2. The intelligent grabbing system of the mechanical arm based on the Internet of Things, applied to the intelligent grabbing method of the mechanical arm based on the Internet of Things, characterized in that, The grasping scene twin unit is used to construct a grasping scene twin model before the mechanical arm starts working; The grasping area risk analysis unit is used to divide the working space of the mechanical arm in the grasping scene twin model into multiple working areas, mark the working area where the to-be-grasped object exists as a grasping area, and define the grasping comprehensive risk index of each grasping area in real time during the working process of the mechanical arm; The definition process of the grasping comprehensive risk index of the grasping area: select a grasping area, obtain the arm health risk index and the target grasping risk index of the grasping area, and calculate the grasping comprehensive risk index of the grasping area based on the arm health risk index and the target grasping risk index; The mechanical arm grasping control unit marks the grasping area with the smallest grasping comprehensive risk index value as the optimal grasping area, and controls the mechanical arm to preferentially grasp the to-be-grasped object in the optimal grasping area. 3.The intelligent grabbing system of the mechanical arm based on the Internet of Things according to claim 2, characterized in that, The grasping scene twin model includes a mechanical arm digital twin model, a to-be-grasped object digital twin model and a grasping environment digital twin model, and the grasping scene twin model is constructed by fusing the mechanical arm digital twin model, the to-be-grasped object digital twin model and the grasping environment digital twin model.

4. The intelligent grasping system of mechanical arm based on Internet of Things according to claim 2, characterized in that, The acquisition process of the joint load rate of the possible posture: select a possible posture, input the pose parameters of the possible posture by using KDL, and reverse the joint angle combination that satisfies the pose by inverse kinematics algorithm. Each joint angle included in the joint angle combination is input into the multi-body dynamics model of the robot arm, the torque output by each joint is calculated, and the load rate of each joint is further calculated. The load rates of each joint are weighted and summed to calculate the joint load rate of the possible posture.

5. The intelligent grasping system of mechanical arm based on Internet of Things according to claim 2, characterized in that, The acquisition process of the grasp offset of the first-graspable object: select a first-graspable object, simulate grasping the first-graspable object in the grasp area by the gripper of the robot arm based on the multi-body dynamics model, obtain the center of gravity coordinates of the first-graspable object and the center coordinates of the gripper after the simulation, and calculate the grasp offset of the first-graspable object based on the center of gravity coordinates of the first-graspable object and the center coordinates of the gripper.

6. The intelligent grasping system of mechanical arm based on Internet of Things according to claim 2, characterized in that, The acquisition process of the grasp stack stability index of the first-graspable object: select a first-graspable object, simulate grasping the first-graspable object in the grasp area by the gripper of the robot arm based on the multi-body dynamics model, identify the stacking relationship of all remaining graspable objects after the simulation, calculate the stability coefficient of each remaining graspable object, and calculate the grasp stack stability index by summing and averaging the stability coefficients of all remaining graspable objects.

Images