Dynamic parameter synchronous identification method considering installation inclination angle of robot base

Abstract

A dynamic parameter synchronous identification method considering a robot base installation inclination angle comprises the steps that parameters are initialized, sensor signals are collected, acceleration of all joints is extracted, and then a Jacobian matrix of residual errors of base force sensing information about the robot installation inclination angle is obtained through calculation according to a base model observation matrix and a force spinor vector; after the installation inclination angle of the robot, the coupling matrix parameters and the kinetic model parameters are updated, whether the installation inclination angle of the robot converges or not is judged by judging whether the result difference obtained through two times of iteration is smaller than a given threshold value or not; and when convergence is carried out, the gravitational acceleration vector and the Jacobian matrix of the gravitational acceleration vector under the robot base coordinates corresponding to the robot installation inclination angle are calculated again, and otherwise, the correctness of the model obtained through identification is verified through different tracks. An industrial robot system with a base and an additional force / torque sensor is utilized, the installation inclination angle of the robot is expanded and identified, coupling parameters of the force sensor are calibrated in situ, and kinetic parameter identification of the robot at any installation inclination angle is achieved. And the whole machine motion control performance of the industrial robot under different installation inclination angles based on the dynamical model is obviously improved.

Description

Technical Field

[0001] The present invention relates to a technology in the field of intelligent manufacturing, in particular to a method for synchronously identifying dynamic parameters taking into account the installation inclination angle of a robot base. Background Art

[0002] The most important theoretical basis for robot dynamic model identification is that the machine dynamic model is linearly related to its inertia parameters, so the model parameters can be identified by the least squares method. However, when the robot has an installation inclination and the inclination is unknown, it cannot be identified by the most basic linearization parameters. Currently, there is no research on identifying the robot's inclination while identifying the dynamic parameters. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention proposes a method for synchronously identifying the dynamic parameters of the robot base taking into account the installation inclination angle of the robot base. By utilizing an industrial robot system with an additional force / torque sensor on the base, the installation inclination angle of the robot is expanded and identified, the coupling parameters of the force sensor are calibrated in situ, and the dynamic parameter identification of the robot at any installation inclination angle is realized, which significantly improves the overall motion control performance of the industrial robot based on the dynamic model at different installation inclination angles.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a method for synchronously identifying dynamic parameters taking into account the installation inclination angle of a robot base, comprising:

[0006] Step 1: Initialize parameters, collect sensor signals and extract the acceleration of each joint, including:

[0007] 1.1 Collect the joint encoder position feedback, joint velocity feedback, and base force / torque sensor signals of each joint, filter the feedback signals with a zero-phase delay filter, and obtain the acceleration of each joint through numerical differentiation;

[0008] 1.2 Initialize the robot installation inclination angle ζ = [φ ρ] T , the installation inclination angle ζ is initialized to [0 0] T ;

[0009] 1.3 Initialize the dynamic model parameters π with additional mechanical parameters of the robot base and the coupling matrix parameters γ of the force sensor set on the robot base.

[0010] Step 2: Calculate the Jacobian matrix of the residual of the base force sensing information with respect to the robot installation inclination angle according to the base model observation matrix and the force spinor vector, including: the Jacobian matrix of the residual with respect to the robot installation inclination angle Where: D is the Frechet derivative operator, and the base model observation matrix composed of the relevant data points of each robot motion and g is The gravity acceleration vector of the robot base coordinates corresponding to the robot installation inclination angle g0 is the gravitational acceleration, χ is the observation matrix of a single data point of the robot based on the base information, q(i) is the external force sensor information of the ith data point, and the intermediate value of the dynamic model parameters with the mechanical parameters of the robot base for The generalized inverse of Force spin matrix of each data point of the base force sensor o(i) is the external force sensor information of the i-th data point, o i is the i-th element of the external force sensor information, E6 is the sixth-order unit matrix, and is the intermediate value of the coupling matrix parameters of the force sensor set at the robot base. in: λ>0 is the regularization constant, E is the unit matrix, π is the dynamic model parameter, and the force spin vector is calculated and pieced together from the relevant data points of each robot motion and the robot mass. m0 is the total mass of the robot.

[0011] Step 3: After updating the robot installation inclination angle, coupling matrix parameters and dynamic model parameters, determine whether the robot installation inclination angle converges by judging whether the difference between the results of the two iterations is less than a given threshold. If converged, return to step 2 to recalculate the gravity acceleration vector and its Jacobian matrix in the robot base coordinates corresponding to the robot installation inclination angle. Otherwise, use different trajectories to verify the correctness of the identified model, including:

[0012] 3.1 Update robot installation inclination ζ new =ζ old -α(J T J+ρE) -1 d, where: α, ρ are constants in the optimization algorithm, and the residual

[0013] 3.2 Update coupling matrix parameters in: and By ζ new Calculated.

[0014] 3.3 Update of dynamic model parameters

[0015] The present invention relates to a system for implementing the above method, comprising: a controller unit, an industrial robot unit, a six-dimensional base force sensor unit and an identification unit, wherein: the controller unit sends a target trajectory to the industrial robot unit according to input command information, the industrial robot unit performs corresponding movements according to the acquired command information sent by the controller, and outputs corresponding movement data, the six-dimensional base force sensor unit outputs corresponding base force data according to the movement of the robot, and the identification unit obtains the identification result according to the robot movement information and the base force information. Technical Effects

[0016] The present invention uses a six-dimensional force sensor to identify a robot with an installation inclination during installation, and can identify most of the dynamic parameters through the six-dimensional force sensor of the base without being affected by the joint friction force model; facing practical specific tasks, while identifying the robot's dynamic parameters, the coupling matrix of the six-dimensional force sensor at the bottom of the robot is calibrated in situ. Compared with the prior art, the present invention significantly improves the accuracy of robot dynamic model parameter identification and the versatility of dynamic parameter identification under different installation postures. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the present invention;

[0018] Figure 2 It is a schematic diagram of an application example;

[0019] Figure 3 The excitation trajectory and test trajectory diagram of the embodiment;

[0020] Figure 4 to Figure 5 It is a diagram of the convergence process of the embodiment;

[0021] Figures 6 to 11 It is a cross-validation error data diagram of the embodiment model;

[0022] Fig.12 This is a diagram of the linear calibration results of the force sensor of the embodiment;

[0023] Figure 13 to Figure 14 This is a diagram of the robot installation inclination calibration result of the embodiment. DETAILED DESCRIPTION

[0024] like Figure 1 As shown, this embodiment relates to a method for Figure 2 The method for synchronously identifying dynamic parameters of an industrial robot equipped with a rotary joint includes:

[0025] Step 1: Prepare data: collect data from external sensors and joint encoder feedback positions; let the robot track the pre-optimized excitation curve, record the data collected by external sensors and encoder feedback positions, and then use a zero-phase delay filter and a median difference filter to filter and differentiate the feedback positions to obtain joint velocity and joint acceleration, and a total of 50,000 data points can be obtained;

[0026] Step 2: Initialize the robot installation inclination angle ζ = [φ ρ] T , the installation inclination angle ζ is initialized to [0 0] T ;

[0027] Step 3, initializing the dynamic model parameters π of the additional mechanical parameters of the robot base and the coupling matrix parameters γ of the force sensor set on the robot base to zero vectors;

[0028] Step 4: Calculate the force spin matrix composed of the data points of the base force sensor in: o(i) is the external force sensor information of the i-th data point, o i is the i-th element of the external force sensor information, E6 is the sixth-order unit matrix;

[0029] The force spinor matrix is a matrix of size 3000 rows and 27 columns.

[0030] Step 5: Calculate the gravity acceleration vector of the robot base coordinates corresponding to the robot installation inclination angle Wherein: g0 is the magnitude of gravitational acceleration, which is 9.81 in this embodiment.

[0031] Step 6: Calculate the base model observation matrix composed of the relevant data points of each robot movement and g Where: χ is the observation matrix of a single data point of the robot based on the base information, q(i) is the external force sensor information of the i-th data point;

[0032] The base model observation matrix is a matrix of size 3000 rows and 43 columns.

[0033] Step 7: Calculate the force spin vector of each robot motion-related data point and robot mass calculation and splice Wherein: m0 is the total mass of the robot, which is 37.2 in this embodiment.

[0034] The force spinor vector is a vector of length 3000.

[0035] Step 8: Calculate the intermediate value of the coupling matrix parameters of the force sensor set on the robot base in: λ>0 is the regularization term constant, which is 100 in this embodiment, and E is the unit matrix.

[0036] Step 9: Calculate the intermediate values ​​of the dynamic model parameters with the mechanical parameters of the robot base in:

[0037] Step 10: Calculate the residual error of the base force sensor information under the parameters calculated in steps 8 and 9

[0038] Step 11: Calculate the Jacobian matrix of the residual with respect to the robot installation inclination angle Where: D is the Frechet derivative operator.

[0039] Step 12: Update the robot installation inclination angle new =ζ old -α(J T J+ρE) -1 d, where: α, ρ are constants in the optimization algorithm, and are 1 and 100 in this embodiment.

[0040] Step 13: Update the coupling matrix parameters of the force sensor set on the robot base in: and By ζ new Calculated.

[0041] Step 14: Update the dynamic model parameters with the robot base mechanical parameters

[0042] Step 15, determine whether the robot installation inclination angle ζ converges. If it converges, go to step 5, otherwise go to step 16. The convergence is determined by whether the difference between the results obtained by two iterations is less than a given threshold of 0.001.

[0043] Step 16, verify the identification parameters through the identified installation inclination angle and the six-dimensional force spinor of the model base. The specific operation is: compare the identified installation inclination angle with the actually measured installation inclination angle, use the identification parameters to calculate the six-dimensional force spinor of the base of the robot on the verification trajectory and compare it with the six-dimensional force spinor of the base actually measured after the robot moves on the verification trajectory.

[0044] After specific actual experiments, the above method was run with the above parameters, and the experimental data that can be obtained are: the installation inclination angle identification error is less than 0.5 degrees, and the relative error of the six-dimensional force spinor prediction of the base on the verification trajectory is less than 3%.

[0045] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A method for synchronously identifying dynamic parameters considering the inclination angle of the robot base installation, characterized in that: Initialize the parameters, collect sensor signals and extract the acceleration of each joint, then calculate the Jacobian matrix of the residual of the base force sensing information with respect to the robot installation inclination angle according to the base model observation matrix and the force spinor vector, then update the robot installation inclination angle, coupling matrix parameters and dynamic model parameters, and judge whether the robot installation inclination angle converges by judging whether the difference between the results of the two iterations is less than a given threshold. When converged, recalculate the gravity acceleration vector and its Jacobian matrix in the robot base coordinates corresponding to the robot installation inclination angle, otherwise use different trajectories to verify the correctness of the identified model.

2. The method for synchronously identifying dynamic parameters considering the robot base installation inclination angle according to claim 1 is characterized in that: The initialization parameters, collecting sensor signals and extracting the acceleration of each joint, specifically include: 1.1 Collect the joint encoder position feedback, joint velocity feedback, and base force / torque sensor signals of each joint, filter the feedback signals with a zero-phase delay filter, and obtain the acceleration of each joint through numerical differentiation; 1.2 Initialize the robot installation inclination angle ζ = [φ ρ] T , the installation inclination angle ζ is initialized to [0 0] T ; 1.3 Initialize the dynamic model parameters π with additional mechanical parameters of the robot base and the coupling matrix parameters γ of the force sensor set on the robot base.

3. The method for synchronously identifying dynamic parameters considering the robot base installation inclination angle according to claim 1 is characterized in that: The Jacobian matrix is ​​obtained in the following way: The Jacobian matrix of the residual with respect to the robot installation inclination angle Where: D is the Frechet derivative operator, and the base model observation matrix composed of the relevant data points of each robot motion and g is The gravity acceleration vector of the robot base coordinates corresponding to the robot installation inclination angle g0 is the gravitational acceleration, χ is the observation matrix of a single data point of the robot based on the base information, q(i) is the external force sensor information of the ith data point, and the intermediate value of the dynamic model parameters with the mechanical parameters of the robot base for The generalized inverse of Force spin matrix of each data point of the base force sensor o(i) is the external force sensor information of the i-th data point, o i is the i-th element of the external force sensor information, E6 is the sixth-order unit matrix, and is the intermediate value of the coupling matrix parameters of the force sensor set at the robot base. in: λ>0 is the regularization constant, E is the unit matrix, π is the dynamic model parameter, and the force spin vector is calculated and pieced together from the relevant data points of each robot motion and the robot mass. m0 is the total mass of the robot.

4. The method for synchronously identifying dynamic parameters considering the robot base installation inclination angle according to claim 1 is characterized in that: Whether the robot installation inclination angle converges is determined by judging whether the difference between the results of the two iterations is less than a given threshold, which specifically includes: 3.1 Update robot installation inclination ζ new =ζ old -α(J T J+ρE) -1 d, where: a, ρ are constants in the optimization algorithm, and the residual 3.2 Update coupling matrix parameters in: and By ζ new Calculated; 3.3 Update of dynamic model parameters 5. The method for synchronously identifying dynamic parameters considering the robot base installation inclination angle according to any one of claims 1 to 4, characterized in that: include: Step 1: Prepare data: collect data from various external sensors and feedback positions from joint encoders; Let the robot track the pre-optimized excitation curve, record the data collected by the external sensor and the encoder feedback position, and then use the zero-phase delay filter and median difference filter to filter and differentiate the feedback position to obtain the joint velocity and joint acceleration, and a total of 50,000 data points can be obtained; Step 2: Initialize the robot installation inclination angle ζ = [φ ρ] T , the installation inclination angle ζ is initialized to [0 0] T ; Step 3, initializing the dynamic model parameters π of the additional mechanical parameters of the robot base and the coupling matrix parameters γ of the force sensor set on the robot base to zero vectors; Step 4: Calculate the force spin matrix composed of the data points of the base force sensor in: o(i) is the external force sensor information of the i-th data point, o i is the i-th element of the external force sensor information, E6 is the sixth-order unit matrix; Step 5: Calculate the gravity acceleration vector of the robot base coordinates corresponding to the robot installation inclination angle Where: g0 is the magnitude of gravitational acceleration, which is taken as 9.81 in this embodiment; Step 6: Calculate the base model observation matrix composed of the relevant data points of each robot movement and g Where: χ is the observation matrix of a single data point of the robot based on the base information, q(i) is the external force sensor information of the i-th data point; Step 7: Calculate the force spin vector of each robot motion-related data point and robot mass calculation and splice Where: m0 is the total mass of the robot; Step 8: Calculate the intermediate value of the coupling matrix parameters of the force sensor set on the robot base in: λ>0 is the regularization term constant, which is 100 in this embodiment, and E is the unit matrix; Step 9: Calculate the intermediate values ​​of the dynamic model parameters with the mechanical parameters of the robot base in: Step 10: Calculate the residual error of the base force sensor information under the parameters calculated in steps 8 and 9 Step 11: Calculate the Jacobian matrix of the residual with respect to the robot installation inclination angle Where: D is the Frechet derivative operator; Step 12: Update the robot installation inclination angle new =ζ old -α(J T J+ρE) -1 d, where: α, ρ are constants in the optimization algorithm; Step 13: Update the coupling matrix parameters of the force sensor set on the robot base in: and By ζ new Calculated; Step 14: Update the dynamic model parameters with the robot base mechanical parameters Step 15, judging whether the robot installation inclination angle ζ converges, if converged, go to step 5, otherwise go to step 16, the convergence judgment is: whether the difference between the results obtained by two iterations is less than a given threshold of 0.001; Step 16, verify the identification parameters through the identified installation inclination angle and the six-dimensional force spinor of the model base. The specific operation is: compare the identified installation inclination angle with the actually measured installation inclination angle, use the identification parameters to calculate the six-dimensional force spinor of the base of the robot on the verification trajectory and compare it with the six-dimensional force spinor of the base actually measured after the robot moves on the verification trajectory.

6. A system for synchronously identifying dynamic parameters taking into account the robot base installation inclination angle for implementing any of the methods described in claims 1-5, characterized in that: include: A controller unit, an industrial robot unit, a six-dimensional base force sensor unit and an identification unit, wherein: the controller unit sends the target trajectory to the industrial robot unit according to the input command information; the industrial robot unit performs corresponding movements according to the acquired command information sent by the controller and outputs corresponding movement data; the six-dimensional base force sensor unit outputs corresponding base force data according to the movement of the robot; the identification unit obtains the identification result according to the robot movement information and the base force information.

Images