Moving base robot arm end load mass online estimation method

Abstract

The application discloses a kind of dynamic base mechanical arm end load mass online estimation method, belong to robot control technical field.The method of the application includes: the estimated value of load mass is handled to obtain six-dimensional virtual generalized force, six-dimensional pose feedback signal of trunk is collected and handled to obtain trunk actual stress, thrust and trunk actual stress are compared, the estimated value of current load mass is handled to obtain the estimated value of new load mass using load mass adaptive iterative speed sliding mode control model, until trunk actual stress and thrust are equal.The dynamic base mechanical arm end load mass online estimation method provided by the application can estimate the load at the end of the arm in real time based on the dynamic base thrust constructed and applied based on the current parameters and the actual thrust prediction value of the trunk to adjust the controller model parameters and improve the system control performance.

Description

Technical Field

[0001] This invention relates to an online estimation method for the end-effector load mass of a moving-base robotic arm, belonging to the field of robot control technology. Background Technology

[0002] The booming development of the robotics industry is injecting strong momentum into economic and social development, and is profoundly changing human production and lifestyles. Service robots and special-purpose robots will be widely used in warehousing and logistics, education and entertainment, and cleaning services. Compared with industrial robotic arms, mobile robotic arms have the advantages of a wider range of motion and more flexible operating space, making them a focus of next-generation robot development. However, the complexity of the operating space and the diversity of tasks pose challenges to robot control. When a mobile robotic arm operates, the interaction force between the robotic arm and the external environment is a crucial input for the robot controller. The movement of the moving base should be planned according to the magnitude of the interaction force to achieve stable control of the robot. Therefore, obtaining the precise value of the interaction force is a key issue in the control of mobile robotic arms.

[0003] Numerous studies have been conducted on end-effector force estimation for industrial robotic arms. The most intuitive method is to install force sensors at the end-effector for interactive force measurement. However, this method has limitations: First, sensors that meet the requirements of accuracy and reliability are too expensive; second, the installation location of force sensors is limited, and it cannot be guaranteed that positive pressure will be generated in every contact, thus failing to obtain reliable contact force data; third, force sensors estimate contact force through their own deformation, affecting the accuracy of end-effector position control. Therefore, researchers have proposed various force estimation methods to replace force sensors. One such method, based on a dynamic model, is the interactive force estimation method published in Robotics and Computer-Integrated Manufacturing in 2021, titled "Sensorless force estimation for industrial robots using disturbance observer and neural learning of friction approximation." This method uses neural network learning to compensate for and correct errors in the robot's dynamic model, and then constructs a disturbance observer based on Kalman filtering to achieve end-effector interactive force estimation for a fixed-base robotic arm. The paper "Precision motion control of a 6-DoFs industrial robot with accurate payload estimation", published in IEEE / ASME Transactions on Mechatronics in 2020, proposes an adaptive control law based on generalized momentum, which realizes online payload estimation.

[0004] However, for manipulators with moving bases (wheeled mobile manipulators, humanoid dual-arm robots, quadrupedal single-arm / dual-arm robots, etc.), the end-effector load mass estimation methods applicable to industrial robots cannot be directly transferred and applied. Furthermore, the interaction forces at the manipulator's end directly affect the control of the moving base; therefore, estimating these interaction forces is even more crucial. A research team at the Italian Institute of Technology added a manipulator to a quadruped robot to give HyQ operational capabilities. Their paper, "Towards a multi-legged mobile manipulator," published at the 2016 IEEE International Conference on Robotics and Automation, assumes the torso is a fixed base and estimates the end-effector force based on a precise dynamic model of the manipulator. This method still follows the end-effector force estimation methods used in industrial robots. However, this method has a drawback: it ignores the influence of the end-effector force on the base motion, resulting in inaccurate end-effector force estimation. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned shortcomings of the prior art and provide an online estimation method for the end-effector load mass of a moving-base robotic arm.

[0006] When the end effector of the moving-base robotic arm is subjected to external interaction forces, the torso needs to provide a suitable counterforce to resist these external forces, as well as the gravity, inertial forces, and inertial torques of the robotic arm's components. The fundamental source of the thrust of the moving-base robotic arm is the interaction force generated by the foot-ground contact. Mapping the generalized pushing force of the moving-base robotic arm's leg-like chassis onto the torso, we can represent a six-dimensional generalized force W at the torso's center of mass. leg At the same time, the torso is also subjected to interactive forces with the robotic arm, W a Based on the Newton-Euler formula for a single rigid body, the dynamic model of the torso is represented by an analytical solution:

[0007]

[0008] Among them, M b Let P be the trunk inertia matrix. b This is the six-dimensional pose of the torso; therefore... For trunk pose acceleration, G b J is the torso gravity matrix. a The force Jacobian matrix is ​​the mapping of the interaction force between the robotic arm and the torso to the torso's center of mass.

[0009] And W a This can be expressed as the mass m of the robotic arm link. a Load mass m obj It presents a linear relationship:

[0010]

[0011] in, These are the linear regression analytical expressions for the effects of the end-effector mass and the mass of the robotic arm links on the interaction force between the robotic arm and the torso, respectively. q, These represent the spatial position, velocity, and acceleration of the robotic arm joints, respectively.

[0012] Substituting equation (2) into equation (1), we obtain the torso dynamics model related to the mass of the robotic arm links and the load mass:

[0013]

[0014] Based on formula (3), a dynamic base thrust control model is constructed, enabling the leg-type chassis to adjust the thrust according to the current motion state of the torso. A virtual model control method can be used here, constructing a six-dimensional virtual generalized force W at the torso's center of mass based on the error between the desired and actual pose of the torso. d :

[0015]

[0016] Among them, W d For a six-dimensional virtual generalized force, K p K is a positive definite proportional gain matrix. d P is a positive definite differential gain matrix. d For the desired pose of the torso, P is the desired velocity of the torso. b This represents the actual position of the torso. G represents the actual velocity of the torso. b J is the torso gravity matrix. a The force Jacobian matrix is ​​the mapping of the interaction forces between the robotic arm and the torso to the torso's center of mass. The estimated end-effector mass is given by the linear regression analytical expression of the robotic arm-torso interaction force. This is the linear regression analysis formula for the mass of the robotic arm's linkages with respect to the interaction force between the robotic arm and the torso. m is an estimate of the load mass. a For the mass of the robotic arm components.

[0017] It can be seen that when Not equal to m obj The thrust increase directly affects the control of the torso posture. The thrust calculated based on the estimated load mass is denoted as u. fb =W d .if When a thrust u is applied fbWhen the torso is in motion, the torso should move along the desired trajectory. If the actual movement of the torso deviates from the desired trajectory, the estimated load mass is inaccurate.

[0018] To estimate the actual forces acting on the torso, firstly, the state variables of the torso, namely the real-time six-dimensional pose of the torso, are determined. fb During the retrieval process, due to the fluctuating motion of the torso, the retrieved real-time pose is subject to significant noise interference, leading to errors in the estimation of the actual forces acting on the torso. Therefore, the motion state sampling frequency was set to 7.69Hz (i.e., the update frequency of the estimated load mass is 7.69Hz), effectively filtering out high-frequency disturbances. The torso velocity was obtained through differentiation. With acceleration Based on the dynamic model shown in formula (3), the following equation can be obtained:

[0019]

[0020] Among them, u d This is the input construction error caused by the load quality error (the difference between the estimated and actual load quality). At this point, u d The following relationship exists between the error in the estimation of load quality and the following:

[0021]

[0022] According to formula (5), the adaptive iterative speed sliding mode control model for load mass is constructed as follows:

[0023]

[0024] in, For adaptive iteration speed of load quality, The estimated load mass is expressed by the linear regression analysis of the robotic arm-torso interaction force. for The reciprocal of , diag means to arrange the elements of the column vector within the parentheses into a diagonal matrix, u d The input construction error caused by load quality error, u fb For thrust, u fb =W d , This expression calculates the average of the elements of the vector within the parentheses, where `sat` is the saturation function. The `sat` saturation function effectively suppresses... Significant fluctuations; the specific form of SAT is:

[0025]

[0026] Here, in, The threshold for adaptive iteration of load quality. The estimated load quality. The update is performed using the following formula, and the value asymptotically stabilizes at the actual value m. obj The estimated load quality after iteration. Construction of the torso pose control model for the next time step, calculation of W d .

[0027]

[0028] The present invention adopts the following technical solution:

[0029] A method for online estimation of the end-effector load mass of a moving-base robotic arm includes the following steps:

[0030] S1. Obtain initial values ​​as an estimate of load quality;

[0031] S2. The estimated value of the load mass is processed according to the dynamic base thrust control model to obtain a six-dimensional virtual generalized force;

[0032] S3. Apply thrust to perform corresponding actions, wherein the thrust is equal to a six-dimensional virtual generalized force;

[0033] S4. Acquire six-dimensional pose feedback signals of the torso;

[0034] S5. Based on the trunk force estimation model, the six-dimensional pose feedback signal of the trunk is processed to obtain the actual force on the trunk;

[0035] S6. Record the thrust exerted in this operation as the feedback force; compare the feedback force with the actual force exerted on the torso to obtain the comparison relationship between the feedback force and the actual force exerted on the torso:

[0036] S7. Process the estimated load mass based on the comparison between the feedback force and the actual force on the torso:

[0037] When the actual force on the torso is equal to the feedback force, the estimated value of the current load mass is the actual value of the load mass.

[0038] When the actual force on the torso is less than or greater than the feedback force, the load mass adaptive iterative speed sliding mode control model is used to process the current load mass estimate, obtain a new load mass estimate, and continue to execute S2-S7 until the actual force on the torso is equal to the feedback force.

[0039] In step S1, the initial value can be randomly assigned by the robot's control center and can be any value, such as 0. In step S4, the sampling frequency can be set to 7.69Hz.

[0040] The dynamic base thrust control model mentioned in step S2 is as follows:

[0041]

[0042] Among them, W d For a six-dimensional virtual generalized force, K p K is a positive definite proportional gain matrix. d P is a positive definite differential gain matrix. d For the desired pose of the torso, P is the desired velocity of the torso. b This represents the actual position of the torso. G represents the actual velocity of the torso. b J is the torso gravity matrix. a The force Jacobian matrix is ​​the mapping of the interaction forces between the robotic arm and the torso to the torso's center of mass. The estimated end-effector mass is given by the linear regression analytical expression of the robotic arm-torso interaction force. Let q be the linear regression analytical expression of the mass of the robotic arm link with respect to the interaction force between the robotic arm and the torso. These are the spatial position, velocity, and acceleration of the robotic arm, respectively. m is an estimate of the load mass. a For the mass of the robotic arm components.

[0043] The trunk force estimation model described in step S5 is as follows:

[0044]

[0045] Among them, M b For the actual force on the torso, For trunk pose acceleration, G b For the torso gravity matrix, W leg J is the force exerted by the legs on the torso. a The force Jacobian matrix is ​​the mapping of the interaction forces between the robotic arm and the torso to the torso's center of mass. The estimated load mass is expressed by the linear regression analysis of the robotic arm-torso interaction force. Let m be the linear regression analytical expression of the mass of the robotic arm link with respect to the interaction force between the robotic arm and the torso. obj Take an estimated value of the load quality. m a For the mass of the robotic arm link, q, These are the spatial position, velocity, and acceleration of the robotic arm, respectively.

[0046] The load quality adaptive iterative speed sliding mode control model mentioned in step S7 includes the load quality adaptive iterative speed formula and the load quality update formula;

[0047] The formula for the adaptive iteration speed of load quality is: in, For the adaptive iteration speed of load quality, q, These are the spatial position, velocity, and acceleration of the robotic arm, respectively. The estimated load mass is expressed by the linear regression analysis of the robotic arm-torso interaction force. for The reciprocal of , diag means to arrange the elements of the column vector within the parentheses into a diagonal matrix, u d The input construction error caused by load quality error, u fb For thrust, u fb =W d , This expresses the expression for calculating the average of the elements of the vector within the parentheses, where `sat` is the saturation function; the specific form of `sat` is:

[0048]

[0049] Here, in, A threshold for adaptive iteration of the estimated load quality;

[0050] Load quality update formula: Among them, the left side of the formula For the new

[0051] This method is based on the assumption that the moving base (wheeled, tracked, or legged, etc.) can drive the torso to accurately servo the desired pose under conditions of no load mass or precise acquisition of load mass. The assumptions of this method are easy to meet, simple to understand, and highly operable.

[0052] The fundamental source of the thrust of the moving-base robotic arm is the interaction force generated by the foot-ground contact. By simultaneously mapping the leg-type chassis and the output force of the robotic arm's end effector onto the torso force space, and performing comparative analysis of input and state variables in this space, the algorithm flow is greatly simplified. The algorithm flow consists of three parts: execution, observation, and correction; the specific flowchart is shown below. Figure 1 As shown; the specific interaction process is as follows: Figure 2 As shown.

[0053] The beneficial effects of this invention are:

[0054] The online estimation method for the end-effector load mass of the dynamic base robotic arm of the present invention integrates the robotic arm links, the end-effector load mass, and the force vectors applied to the torso generated by dynamic motion into a torso-robotic arm interaction force, and describes it in a form that is linearly related to the robotic arm links and the end-effector load mass, thus avoiding the real-time calculation pressure caused by the complex analytical formulas of the overall dynamic model.

[0055] The online estimation method for the end-load mass of the moving base robotic arm of the present invention estimates the resultant force on the torso using the posture feedback value of the torso, so as to describe the total influence of the active thrust of the base and the interactive resultant force of the robotic arm on the torso, without having to calculate the interactive force on the torso by facing the base and the robotic arm, which greatly simplifies the estimation complexity of the interactive force.

[0056] The online estimation method for the end-effector load mass of the moving base robotic arm of the present invention constructs an adaptive iterative speed sliding mode control model for the load mass to update the iterative speed of the load mass, ensuring that the predicted value smoothly approaches the actual value.

[0057] This invention ensures the reliability of motion control for a moving base by accurately estimating the load mass of a robotic arm. The inaccuracy and real-time changes in the robotic arm's load mass affect the pose control of the moving base. This invention addresses the problem of inaccurately obtaining the robotic arm's load mass by addressing the interference of the load mass on pose control. The method is simple and easy to understand, and applicable to various types of moving base robotic arms.

[0058] The online estimation method for the end-effector load mass of the moving base robotic arm of the present invention can estimate the end-effector load mass in real time based on the predicted values ​​of the moving base thrust and the actual thrust of the torso constructed and applied with the current parameters, so as to adjust the controller model parameters and continuously improve the system control performance. Attached Figure Description

[0059] Figure 1 This is a flowchart of the online estimation method for the end-effector load mass of the moving base robotic arm according to the present invention;

[0060] Figure 2 This is an interactive process diagram of the online estimation method for the end-effector load mass of the moving base robotic arm of the present invention;

[0061] Figure 3 This is a schematic diagram of the topology of a quadruped robot equipped with a three-degree-of-freedom robotic arm, as described in a specific embodiment of the present invention.

[0062] Figure 4 This involves transforming the effect of the leg-based chassis on the torso into a six-dimensional generalized force acting on the torso. Figure 2 The diagram shows the forces acting on the robot's torso. Detailed Implementation

[0063] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0064] The structures, proportions, and sizes illustrated in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art in understanding and reading the invention. They are not intended to limit the scope of the invention and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, provided they do not affect the effectiveness or purpose of the invention, should still fall within the scope of the technical content disclosed herein. Furthermore, the terms "upper," "lower," "left," "right," "middle," and "one" used in this specification are merely for clarity and not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0065] This paper takes a quadruped robot with a moving base equipped with a three-degree-of-freedom robotic arm (hereinafter referred to as the robot) as an example to illustrate in detail the online estimation method of the end-effector load mass of the moving base robotic arm. A schematic diagram of the topology of this quadruped robot with a three-degree-of-freedom robotic arm is shown below. Figure 3 As shown; after transforming the effect of the leg-based chassis on the torso into a six-dimensional generalized force acting on the torso, Figure 3 The diagram showing the forces acting on the torso of the quadruped robot is as follows: Figure 4 As shown.

[0066] Apply a load of mass X kg to the end effector of the robot's robotic arm.

[0067] S1. The robot first obtains a random value as an estimate of the load mass; this random value is the initial value; the initial value can be 0.

[0068] S2. The robot processes the estimated value (initial value) of the load mass based on the dynamic base thrust control model to obtain the six-dimensional virtual generalized force W. d The dynamic base thrust control model is as follows:

[0069]

[0070] Among them, W d For a six-dimensional virtual generalized force, K p K is a positive definite proportional gain matrix. d P is a positive definite differential gain matrix. d For the desired pose of the torso, P is the desired velocity of the torso. b and These are the actual pose and actual velocity of the torso, obtained through real-time measurement. G b J is the torso gravity matrix. a The force Jacobian matrix is ​​the mapping of the interaction forces between the robotic arm and the torso to the torso's center of mass. The estimated end-effector mass is given by the linear regression analytical expression of the robotic arm-torso interaction force. The above four parts are analytical expressions related to the robot's physical parameters and joint motion states, where the mass of the robotic arm's linkages is linearly regressed with respect to the interaction force between the robotic arm and the torso. The physical parameters can be obtained through measurement during modeling, and the joint motion states q... These are the robotic arm's spatial position, velocity, and acceleration, obtained through real-time measurement. m is an estimate of the load mass. a For the mass of the robotic arm components.

[0071] S3. Apply thrust to perform the corresponding action, whereby the thrust is equal to the six-dimensional virtual generalized force. That is, the robot will obtain the six-dimensional virtual generalized force W. d As a propulsive force, the robot executes this force to perform actions. The robot's six-dimensional torso pose changes.

[0072] S4. Real-time acquisition of six-dimensional posture feedback signals of the torso, including the three-dimensional linear acceleration and three-dimensional angular acceleration of the torso.

[0073] S5. Based on the trunk force estimation model, the six-dimensional pose feedback signal of the trunk is processed to obtain the actual force on the trunk.

[0074] The trunk force estimation model described in step S5 is as follows:

[0075]

[0076] Among them, M b For the actual force on the torso, For trunk pose acceleration, G b For the torso gravity matrix, W leg J is the force exerted by the legs on the torso. a The force Jacobian matrix is ​​the mapping of the interaction forces between the robotic arm and the torso to the torso's center of mass. The estimated load mass is expressed by the linear regression analysis of the robotic arm-torso interaction force. Let m be the linear regression analytical expression of the mass of the robotic arm link with respect to the interaction force between the robotic arm and the torso. obj Take an estimated value of the load quality. m a For the mass of the robotic arm link, q, These are the spatial position, velocity, and acceleration of the robotic arm, respectively.

[0077] The estimated load mass is expressed by the linear regression analysis of the robotic arm-torso interaction force. This is a linear regression analytical expression for the mass of the robotic arm's linkages with respect to the interaction force between the robotic arm and the torso. The specific form of the analytical expression varies depending on the robot configuration; the linear regression analytical expression can be derived by organizing the dynamic model, which is common knowledge to those skilled in the art and will not be elaborated upon here.

[0078] S6. Record the thrust W of this execution. d The feedback force is equal to W. d ; Feedback force W d The actual force M on the torso b A comparison was made to obtain the relationship between the feedback force and the actual force on the torso:

[0079] S7. Process the estimated load mass based on the comparison between the feedback force and the actual force on the torso:

[0080] When the actual force M on the torso b With feedback force W d When they are equal, the estimated value of the current load quality That is, the actual value of the load mass m obj ;

[0081] When the actual force M on the torso b Less than or greater than the feedback force W d At that time, the load quality adaptive iterative speed sliding mode control model is used to estimate the current load quality. Process the data to obtain a new estimate of the load quality. Continue executing S2-S7 until the actual force M on the torso is reached. b With feedback force W d equal.

[0082] The load mass adaptive iterative speed sliding mode control model described in step S7 includes a load mass adaptive iterative speed formula and a load mass update formula; that is, firstly, the adaptive iterative speed of the load mass is obtained according to the load mass adaptive iterative speed formula, and then the adaptive iterative speed of the load mass is substituted into the load mass update formula; a new estimated value of the load mass is calculated based on the current estimated value of the load mass and the adaptive iterative speed of the load mass. Then, steps S2-S7 are repeated based on the new estimated value of the load mass until the actual force M on the torso is reached. b With feedback force W d Equal. When the actual force M on the torso b With feedback force W d When they are equal, the estimated load quality is equal to the actual load quality.

[0083] The formula for the adaptive iteration speed of load quality is: in, For the adaptive iteration speed of load quality, q, These are the spatial position, velocity, and acceleration of the robotic arm, respectively. The estimated load mass is expressed by the linear regression analysis of the robotic arm-torso interaction force. for The reciprocal of , diag means to arrange the elements of the column vector within the parentheses into a diagonal matrix, u d The input construction error caused by load quality error, u fd For thrust, u fd =W d , This expresses the calculation of the average of the elements of the vector within the parentheses, where `sat` is the saturation function; the specific form of `sat` is:

[0084]

[0085] Here, in, A threshold for adaptive iteration of the estimated load quality;

[0086] Load quality update formula: Among them, the left side of the formula For the new The update frequency for the estimated load quality is set to 7.69 Hz.

[0087] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for online estimation of the end-effector load mass of a moving-base robotic arm, characterized in that, Includes the following steps: S1. Obtain initial values ​​as an estimate of load quality; S2. The estimated value of the load mass is processed according to the dynamic base thrust control model to obtain a six-dimensional virtual generalized force; The fundamental source of the thrust of the moving-base robotic arm is the interaction force generated by the foot-ground contact. The generalized force of the moving-base robotic arm's leg-like chassis pushing the torso can be mapped as a six-dimensional generalized force at the torso's center of mass. W leg At the same time, the torso is also subjected to interactive forces with the robotic arm. W a Based on the Newton-Euler formula for a single rigid body, the dynamic model of the torso is represented by an analytical solution: (1) in, M b For the torso inertia matrix, P b This is the six-dimensional pose of the torso; therefore... For trunk posture acceleration, G b For the torso gravity matrix, J a The force Jacobian matrix is ​​the mapping of the robotic arm-torso interaction force to the torso's center of mass. and W a Represented as the mass of the robotic arm link m a Load quality m obj It presents a linear relationship: (2) in, , These are the linear regression analytical expressions for the effects of the end-effector mass and the mass of the robotic arm linkage on the interaction force between the robotic arm and the torso, respectively; q, , These are the spatial position, velocity, and acceleration of the robotic arm joints, respectively. m obj Take an estimated value of the load quality. , m a For the mass of the robotic arm components; Substituting formula (2) into formula (1), we obtain the result related to the mass of the robotic arm link. m a Load quality m obj Related trunk dynamics models: (3) Based on formula (3), a dynamic base thrust control model is constructed so that the leg-type chassis adjusts the thrust according to the current motion state of the torso. Here, a virtual model control method is adopted, and a six-dimensional virtual generalized force at the torso's center of mass is constructed based on the error between the desired pose and the actual pose of the torso. W d : The dynamic base thrust control model is as follows: ； in, K p It is a positive definite proportional gain matrix. K d It is a positive definite differential gain matrix. P d For the desired pose of the torso, For the desired velocity of the torso, This is an estimate of the load quality; S3. Apply thrust to perform corresponding actions, wherein the thrust is equal to a six-dimensional virtual generalized force; S4. Acquire six-dimensional pose feedback signals of the torso; S5. Based on the trunk force estimation model, the six-dimensional pose feedback signal of the trunk is processed to obtain the actual force on the trunk; S6. Record the thrust exerted in this operation as the feedback force; compare the feedback force with the actual force exerted on the torso to obtain the comparison relationship between the feedback force and the actual force exerted on the torso: S7. Process the estimated load mass based on the comparison between the feedback force and the actual force on the torso: When the actual force on the torso is equal to the feedback force, the estimated value of the current load mass is the actual value of the load mass. When the actual force on the torso is less than or greater than the feedback force, the load mass adaptive iterative speed sliding mode control model is used to process the current load mass estimate, obtain a new load mass estimate, and continue to execute S2-S7 until the actual force on the torso is equal to the feedback force. The load quality adaptive iterative speed sliding mode control model includes a load quality adaptive iterative speed formula and a load quality update formula; The formula for the adaptive iteration speed of load quality is: ; in, For the adaptive iteration speed of load quality, q, , , , represent the spatial position, velocity, and acceleration of the robotic arm, respectively; The estimated load mass is expressed by the linear regression analysis of the robotic arm-torso interaction force. for The reciprocal, diag This means arranging the elements of the column vector within the parentheses into a diagonal matrix. u d The input construction error is caused by the load quality error. u fb For thrust, u fb = W d , This indicates calculating the average of the elements of the vector within the parentheses. sat It is a saturation function; sat The specific form is: , Here, ;in, A threshold for adaptive iteration of the estimated load quality; The load quality update formula is: ； Among them, the left side of the formula For the new , The adaptive iteration speed for load quality.

2. The online estimation method for the end-effector load mass of a moving-base robotic arm as described in claim 1, characterized in that, Set the update frequency of the estimated load quality in step S7 to 7.69 Hz.

3. The online estimation method for the end-effector load mass of a moving-base robotic arm as described in claim 1, characterized in that, The initial value can be any value.

Images