Flexible joint preset performance function design method and control system for man-machine interaction scene

Abstract

The invention relates to the technical field of robot control, and discloses a man-machine interaction scene-oriented flexible joint preset performance function design method and control system, and the method comprises the steps: building a collaborative robot flexible joint rigid-flexible coupling dynamic model; defining a tracking error, adding a Gaussian kernel function on a traditional preset performance function, and forming a Gaussian kernel-preset performance function as a constraint condition; performing error transformation according to the tracking error and a Gaussian kernel-preset performance function to obtain an unconstrained error; according to the unconstrained error, a virtual control law and an actual control law of a flexible joint nonlinear position controller are designed through a dynamic surface control method; and estimating external interference through the biradial basis function neural network, and replacing an external interference item in the nonlinear position controller with an estimated value to obtain a corrected virtual control law and a corrected actual control law of the nonlinear position controller.

Description

Technical Field

[0001] The present invention relates to the field of robot control technology, and in particular to a method for designing a preset performance function of a flexible joint for a human-machine interaction scenario and a control system. Background Art

[0002] At present, various position control algorithms focus on satisfying the steady-state performance of the system, but pay less attention to the transient performance of the system. Prescribed Performance Control (PPC) is mainly proposed to ensure the transient performance and steady-state performance of the system; preset performance refers to the artificially set performance envelope for the state (error) of the controlled system, and the preset performance function is usually designed as a gradually decaying function. If the tracking error can always be within the preset performance function envelope in the process of changing over time, then the purpose of quantitatively designing the transient performance of the tracking error (such as approach rate, upward adjustment, downward adjustment, etc.) and steady-state performance (such as control accuracy) is achieved.

[0003] The specification with publication number CN 114509949 A discloses a robot predetermined performance control method, comprising the following steps: using a dynamic surface control method to recursively design virtual control variables, actual control inputs and adaptive update laws; constructing a funnel error quantity that simultaneously satisfies state constraints and given steady-state error performance indicators; constructing a Lyapunov function of the system; taking the derivative of the Lyapunov function, substituting the designed virtual control variables, actual control inputs and adaptive update laws into the derivative of the Lyapunov function, judging that all closed-loop system variables are ultimately uniformly bounded according to the Lyapunov stability theory, proving that some state constraints are not violated, and obtaining deterministic expressions of tracking error and convergence time, thereby ensuring that the control performance indicators meet the design requirements.

[0004] The specification with the publication number of CN 116225043 B discloses a method for controlling the predetermined performance of a quad-rotor drone based on a disturbance observer, comprising the following steps: establishing a six-degree-of-freedom nonlinear mathematical model of the quad-rotor drone; organizing the six-degree-of-freedom nonlinear mathematical model into a state equation, and converting the virtual control law of the position subsystem into the required lift command and the attitude command of the attitude subsystem of the quad-rotor drone through algebraic transformation; introducing an error conversion function and a performance constraint function, and converting the predetermined performance control problem of the tracking error into the stability problem of the transformed error through error transformation; designing an adaptive predetermined performance controller; designing a disturbance observer for external disturbances at the same time, and obtaining an external disturbance estimation value; integrating the control law and the external disturbance estimation value to obtain the final composite control law. Although the adaptive control strategy in the prior art can respond to external disturbances, the integrated disturbance estimation requires an accurate dynamic model and an effective disturbance observation technology, and the effectiveness of the predetermined performance control strategy will be directly affected by the performance of the disturbance observer.

[0005] The specification with publication number CN 118286034 A discloses a hip exoskeleton robot control method for transient performance enhancement, including: establishing a dynamic model of the hip exoskeleton; constructing a system state space equation based on the dynamic model of the hip exoskeleton; improving the traditional performance function to obtain an improved performance function; based on the improved performance function and the system state space equation, designing a controller based on the improved performance function. The improved performance function in the prior art is a hyperbolic tangent performance function with an auxiliary function added, which does not have an advantage in the case where large errors need to be processed quickly and effectively, especially in applications with high requirements for fast convergence and strong performance recovery; and the performance function used in the prior art is relatively complex in form, and the parameter adjustment is not flexible enough. The strictness and response speed of the performance function cannot be flexibly adjusted according to the control requirements or the dynamic characteristics of the system. Summary of the invention

[0006] The purpose of the present invention is to provide a method for designing preset performance functions of flexible joints, which is aimed at the human-computer interaction scenarios of collaborative robots, avoids controller instability caused by occasional sudden changes in external torque, enhances the transient performance of predetermined performance control, improves the response speed and accuracy of the control system, and realizes accurate tracking of human-computer interaction trajectories.

[0007] A method for designing a preset performance function of a flexible joint for a human-computer interaction scenario, comprising:

[0008] Step 1: Establish a rigid-flexible coupling dynamic model of the flexible joints of the collaborative robot;

[0009] Step 2: Define the tracking error, add a Gaussian kernel function to the traditional preset performance function, and form a Gaussian kernel-preset performance function as a constraint condition;

[0010] Step 3: According to the tracking error and the Gaussian kernel-preset performance function, the error is transformed to obtain the unconstrained error;

[0011] Step 4: According to the unconstrained error, the nonlinear position controller of the flexible joint is designed by the dynamic surface control method;

[0012] Step 5: Estimate external disturbances through a dual radial basis function neural network, replace the external disturbance terms in the nonlinear position controller with the estimated values, and obtain the virtual control law and the actual control law of the nonlinear position controller.

[0013] Furthermore, in step 1, the process of establishing the rigid-flexible coupling dynamic model of the flexible joint is specifically as follows:

[0014] (1-1) Inertia J of the motor end of the flexible joint m and the connecting rod end inertia J l Separated by joint flexibility, J m and J l The motor end viscous damping B m and connecting rod end viscous damping B l Therefore, the rigid-flexible coupling dynamic model of the flexible joint can be expressed as:

[0015]

[0016] Among them, τ m is the input torque of the motor end of the flexible joint to be determined, that is, the actual control law after correction of the nonlinear position controller; N is the reduction ratio of the harmonic reducer, τ ext is the external torque measured by the external torque sensor, K is the joint stiffness, m is the load mass, g is the weight acceleration, l is the connecting rod length; E1 is the non-matching interference, E2 is the matching interference; x1 represents the motor end position, x2 represents the motor end speed, x3 represents the connecting rod end position, and x4 represents the connecting rod end speed.

[0017] Furthermore, in step 2, the process of designing the Gaussian kernel-preset performance function is specifically as follows:

[0018] (2-1) Define the reference motion trajectory of the connecting rod as x 1r , the external torque τ ext Under the action of Among them, M d is the virtual inertia, B d is the virtual damping, K d is the virtual stiffness, and s is a complex variable from the time domain to the frequency domain.

[0019] The expected motion trajectory of the connecting rod is obtained as x 1d =x1c +x 1r , define the tracking error at the output end of the connecting rod as e1 = x1-x 1d ; The tracking accuracy of the tracking error e1 should satisfy in, δ and is a positive constant, and ρ(t) is the Gaussian kernel - the default performance function.

[0020] (2-2) The Gaussian kernel-preset performance function can be expressed as:

[0021] ρ(t)=(ρ0-ρ ∞ -ρ g (t))exp(-γt)+ρ ∞ +ρ g (t)

[0022] Among them, ρ0, ρ ∞ ,γ is a preset positive constant, ρ g (t) is a Gaussian kernel function, which can be expressed as:

[0023]

[0024] Where n is the number of external torque mutations during human-computer interaction. The specific value can be obtained by the external torque τ ext The differential signal is determined; h j 、a j 、b j They are The height and center width of the .

[0025] ρ g (t) is the relaxation term in the Gaussian kernel-preset performance function. When a sudden change in external torque occurs during human-computer interaction, the Gaussian kernel function can relax the boundary of the preset performance function to avoid current saturation of the position controller due to overly tight constraints, thereby avoiding controller instability.

[0026] (2-3) Furthermore, the Gaussian kernel function ρ g The differential of (t) with respect to time is expressed as:

[0027]

[0028] The differential with respect to time is expressed as:

[0029]

[0030] Furthermore, in step 3, the process of performing error transformation to obtain the unconstrained error is as follows:

[0031] (3-1) A strictly monotonically increasing hyperbolic tangent function is selected as the error conversion function φ(ε1) to convert the actual constrained tracking error e1 into a controllable unconstrained error ε1:

[0032]

[0033] The range of the error conversion function is within a preset interval Inside.

[0034] Invert the error conversion function to obtain the unconstrained error ε1:

[0035]

[0036] (3-2) Furthermore, the differential of the unconstrained error ε1 with respect to time is expressed as follows:

[0037]

[0038] in,

[0039]

[0040] The differential with respect to time is expressed as follows:

[0041]

[0042] in,

[0043] The conditions for using ε1 for designing nonlinear position controllers are obtained: when ε1 converges, the performance preset by e1 can be achieved.

[0044] Furthermore, in step 4, the process of designing the nonlinear position controller is specifically as follows:

[0045] (4-1) According to the Define the first virtual control law of x2 for:

[0046]

[0047] Where c1 is a positive constant.

[0048] (4-2) The First Virtual Control Law is the input of the first low-pass filter, and the output of the first low-pass filter is defined as x 2d , the first low-pass filter is as follows:

[0049]

[0050] Where r2 is the time constant of the first low-pass filter.

[0051] (4-3) Define the first virtual tracking error as z2 = x2-x 2d , differentiate the first virtual tracking error, expressed as:

[0052]

[0053] Among them, the first external interference term Ω1=E1.

[0054] (4-4) Define the second virtual control law of x3 for:

[0055]

[0056] Among them, c2 is a positive constant, and w is an introduced adjustment factor to reduce the influence of filtering error on control accuracy.

[0057] (4-5) The Second Virtual Control Law The input of the second low-pass filter is defined as x. 3d , the second low-pass filter is as follows:

[0058]

[0059] Where r3 is the time constant of the second low pass filter.

[0060] (4-6) The second virtual tracking error is defined as Differentiate the second virtual tracking error and express it as:

[0061]

[0062] (4-7) Define the third virtual control law of x4 for:

[0063]

[0064] Where c3 is a positive constant.

[0065] (4-8) The third virtual control law As the input of the third low-pass filter, x 4d The output of the third low-pass filter is as follows:

[0066]

[0067] Wherein, r4 is the time constant of the third low-pass filter.

[0068] (4-9) Define the third virtual tracking error as Differentiate the third virtual tracking error and express it as:

[0069]

[0070] Among them, the second external interference term

[0071] (4-10) Actual control law τ of the flexible joint m0 It is expressed as:

[0072]

[0073] Where c4 is a positive constant.

[0074] Furthermore, in step 5, the process of replacing the external interference term with the estimated value is specifically as follows:

[0075] (5-1) The first radial basis function neural network is used to estimate the first external disturbance Ω1, and the first estimated value is obtained. It is expressed as:

[0076]

[0077] in, is the estimate of the weight parameter vector, ψ1 is the output of the hidden layer Gaussian basis function, and the total number of nodes in the hidden layer is Take ξ=[x1,x2,x3,x4] T As the input of the first radial basis neural network model.

[0078] (5-2) Based on the first estimated value Modified second virtual control law It is expressed as:

[0079]

[0080] (5-8) Using the second radial basis function neural network to estimate the second external disturbance Ω2, the second estimated value is obtained. It is expressed as:

[0081]

[0082] in, is the estimate of the weight parameter vector, ψ2 is the output of the hidden layer Gaussian basis function, and the total number of nodes in the hidden layer is Take ξ=[x1,x2,x3,x4] T As the input of the second radial basis neural network model.

[0083] (5-9) Based on the second estimated value The modified actual control law of the flexible joint τm It is expressed as:

[0084]

[0085] The present invention also provides a flexible joint preset performance control system for human-computer interaction scenarios, comprising:

[0086] Outer loop module: including the expected admittance model module, for a designed reference trajectory x 1r When performing human-computer interaction tasks, the external force is converted into the interaction trajectory x through the expected admittance model 1c The interaction trajectory is superimposed on the reference trajectory to form the desired trajectory x 1d ;

[0087] Inner loop module: including error conversion module, Gaussian kernel-preset performance function module and nonlinear position controller module;

[0088] The error conversion module is used to convert the actual constrained tracking error e1 into a controllable unconstrained error ε1; the Gaussian kernel-preset performance function module is used to use the designed Gaussian kernel-preset performance function as a constraint condition of the error conversion module; the nonlinear position controller module is used to track the desired trajectory x 1d ;

[0089] The error conversion module, Gaussian kernel-preset performance function module and nonlinear position controller module adopt the design method of the flexible joint preset performance function for human-computer interaction scenarios;

[0090] Interaction module: includes flexible joints and external interaction environment, used to perform human-computer interaction tasks.

[0091] Compared with the prior art, the present invention adds a Gaussian kernel function as a relaxation term to the traditional exponential preset performance function, relaxes the boundary of the preset performance function, enhances the stability of the position controller when abrupt external torque occurs, improves the transient and steady-state performance of the tracking error in the flexible joint dynamics model, improves the response speed and accuracy of the collaborative robot control system, and achieves accurate tracking of the human-computer interaction trajectory. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 It is a design process of a preset performance function of a flexible joint for a human-computer interaction scenario in an embodiment of the present invention;

[0093] Figure 2 is a block diagram of a flexible joint control system based on a Gaussian kernel-preset performance function in an embodiment of the present invention;

[0094] Figure 3 is the sudden torque applied to the joint during human-computer interaction in the embodiment of the present invention;

[0095] Figure 4 is the initial reference trajectory and the expected trajectory under the influence of the sudden torque in the embodiment of the present invention;

[0096] Figure 5 It is a comparison diagram of tracking errors based on Gaussian kernel-preset performance function (GK-PPF) and traditional preset performance function (PPF) in an embodiment of the present invention. DETAILED DESCRIPTION

[0097] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0098] like Figure 1 As shown, a design process of a preset performance function of a flexible joint for a human-computer interaction scenario includes:

[0099] S1. Establish a rigid-flexible coupling dynamic model of the flexible joints of collaborative robots;

[0100] S2. Define the tracking error, add a Gaussian kernel function to the traditional preset performance function, and form a Gaussian kernel-preset performance function as a constraint condition;

[0101] S3, performing error transformation according to the tracking error and the Gaussian kernel-preset performance function to obtain an unconstrained error;

[0102] S4. Design a nonlinear position controller for the flexible joint using a dynamic surface control method based on the unconstrained error;

[0103] S5. Estimate external disturbances through a dual radial basis function neural network, replace the external disturbance term in the nonlinear position controller with the estimated value, and obtain a corrected virtual control law and a corrected actual control law of the nonlinear position controller.

[0104] The process of establishing the rigid-flexible coupling dynamics model of the flexible joint is as follows:

[0105] Inertia J of the motor end of the flexible joint m and the connecting rod end inertia J l Separated by joint flexibility, J m and J l The motor end viscous damping B m and connecting rod end viscous damping B l Therefore, the rigid-flexible coupling dynamic model of the flexible joint can be expressed as:

[0106]

[0107] Among them, τm is the input torque of the motor end of the flexible joint to be determined, that is, the actual control law after correction of the nonlinear position controller; N is the reduction ratio of the harmonic reducer, τ ext is the external torque measured by the external torque sensor, K is the joint stiffness, m is the load mass, g is the weight acceleration, l is the connecting rod length; E1 is the non-matching interference, E2 is the matching interference; x1 represents the motor end position, x2 represents the motor end speed, x3 represents the connecting rod end position, and x4 represents the connecting rod end speed.

[0108] In this embodiment, the parameters of the flexible joint dynamics model are shown in Table 1:

[0109] Table 1 Parameters of the flexible joint dynamics model

[0110] parameter <![CDATA[Motor-side inertia J m > <![CDATA[Motor terminal damping B m > Joint stiffness K Gravitational acceleration g Numeric <![CDATA[2.4×10 -5 ]]> 0.0003 1388 9.81 unit <![CDATA[kg·m 2 ]]> N·m·s / rad N·m / rad <![CDATA[m / s 2 ]]> parameter <![CDATA[Inertia J of the connecting rod end l > <![CDATA[Connecting rod end damping B l > Load mass m Connecting rod length l Numeric 0.0477 0.05 2.8 0.14 unit <![CDATA[kg·m 2 ]]> N·m·s / rad kg m

[0111] like Figure 2 As shown, a flexible joint controller based on Gaussian kernel-preset performance function, wherein the design process of the Gaussian kernel-preset performance function is as follows:

[0112] Define the reference motion trajectory of the connecting rod as x 1r , the external torque τ ext Under the action of Among them, M d is the virtual inertia, B d is the virtual damping, K d is the virtual stiffness, and s is a complex variable from the time domain to the frequency domain.

[0113] The expected motion trajectory of the connecting rod is obtained as x 1d =x 1c +x 1r , define the tracking error at the output end of the connecting rod as e1 = x1-x 1d ; The tracking accuracy of the tracking error e1 should satisfy Among them, δ and is a positive constant, ρ(t) is the Gaussian kernel-preset performance function, which can be expressed as:

[0114] ρ(t)=(ρ0-ρ ∞ -ρ g (t))exp(-γt)+ρ ∞ +ρ g (t)

[0115] Among them, ρ0, ρ ∞ ,γ is a preset positive constant, ρ g (t) is a Gaussian kernel function, which can be expressed as:

[0116]

[0117]

[0118] Where n is the number of external torque mutations during the human-computer interaction process in this embodiment. The specific value can be obtained by the external torque τ ext The differential signal is determined; h j 、a j 、b j They are The height, center point, and width of the

[0119] ρ g (t) is the relaxation term in the Gaussian kernel-preset performance function. When a sudden change in external torque occurs during human-computer interaction, the Gaussian kernel function can relax the boundary of the preset performance function to avoid current saturation of the position controller due to overly tight constraints, thereby avoiding controller instability.

[0120] ρ g The differential of (t) with respect to time is expressed as:

[0121]

[0122] The differential with respect to time is expressed as:

[0123]

[0124] A strictly monotonically increasing hyperbolic tangent function is selected as the error conversion function φ(ε1) to convert the actual constrained error e1 into a controllable unconstrained error ε1:

[0125]

[0126] The range of the error conversion function is within a preset interval Inside.

[0127] Invert the error conversion function to obtain the unconstrained error ε1:

[0128]

[0129] The time derivative of ε1 is expressed as follows:

[0130]

[0131] in,

[0132] The differential with respect to time is expressed as follows:

[0133]

[0134] in,

[0135] The conditions for using ε1 for designing nonlinear position controllers are obtained: when ε1 converges, the performance preset by e1 can be achieved.

[0136] exist Figure 2 In the design process of the nonlinear position controller, the specific process is as follows:

[0137] According to the

[0138] Define the first virtual control law of x2 for:

[0139]

[0140] Where c1 is a positive constant.

[0141] The first virtual control law is the input of the first low-pass filter, and the output of the first low-pass filter is defined as x 2d , the first low-pass filter is as follows:

[0142]

[0143] Where r2 is the time constant of the first low-pass filter.

[0144] Define the first virtual tracking error as z2=x2-x 2d , differentiate the first virtual tracking error, expressed as:

[0145]

[0146] Among them, the first external interference term Ω1=E1.

[0147] Define the second virtual control law of x3 for:

[0148]

[0149] Among them, c2 is a positive constant, and w is an introduced adjustment factor to reduce the influence of filtering error on control accuracy.

[0150] The second virtual control law The input of the second low-pass filter is defined as x. 3d , the second low-pass filter is as follows:

[0151]

[0152] Where r3 is the time constant of the second low pass filter.

[0153] The second virtual tracking error is defined as Differentiate the second virtual tracking error and express it as:

[0154]

[0155] Define the third virtual control law of x4 for:

[0156]

[0157] Where c3 is a positive constant.

[0158] The third virtual control law As the input of the third low-pass filter, x 4d The output of the third low-pass filter is as follows:

[0159]

[0160] Wherein, r4 is the time constant of the third low-pass filter.

[0161] Define the third virtual tracking error as Differentiate the third virtual tracking error and express it as:

[0162]

[0163] Among them, the second external interference term

[0164] The actual control law of the flexible joint τ m0 It is expressed as:

[0165]

[0166] Where c4 is a positive constant.

[0167] The process of replacing the external disturbance term of the nonlinear position controller with the estimated value is as follows:

[0168] The first radial basis neural network is used to estimate the first external disturbance Ω1, and the first estimated value is It is expressed as:

[0169]

[0170] Among them, the total number of nodes in the hidden layer is The input is ξ=[x1,x2,x3,x4] T ;

[0171] is an estimate of the weight parameter vector, which is updated by tracking the error z1. The weight parameter adaptive update rate is It can be expressed as in, is a positive constant, Γ1 is a positive definite diagonal matrix;

[0172] is the output of the hidden layer Gaussian basis function, where ψ 1i It can be expressed as Among them, d i =[d i1 ,d i2 ,d i3 ,d i4 ] T is 1i The center vector, p i is 1i Width.

[0173] Based on the first estimate Modified second virtual control law It can be expressed as:

[0174]

[0175] The second radial basis neural network is used to estimate the second external disturbance Ω2, and the second estimated value is It is expressed as:

[0176]

[0177] Among them, the total number of nodes in the hidden layer is The input is ξ=[x1,x2,x3,x4] T ;

[0178] is the estimate of the weight parameter vector, and the weight parameters are updated by tracking the error z4. The weight parameter adaptive update rate It can be written as in, is a positive constant, Γ2 is a positive definite diagonal matrix;

[0179] is the output of the hidden layer Gaussian basis function, where ψ 2j It can be expressed as Among them, d j =[d j1 ,d j2 ,d j3 ,d j4 ] T is 2j The center vector, p j is2j Width.

[0180] Based on the second estimate The modified actual control law of the flexible joint τ m It is expressed as:

[0181]

[0182] based on Figure 2 A flexible joint preset performance control system for human-computer interaction scenarios is simulated and verified in Simulink. The control system includes:

[0183] Outer loop module: including the expected admittance model module, for a designed reference trajectory x 1r When performing human-computer interaction tasks, the external force is converted into the interaction trajectory x through the expected admittance model 1c The interaction trajectory is superimposed on the reference trajectory to form the desired trajectory x 1d ;

[0184] Inner loop module: including error conversion module, Gaussian kernel-preset performance function module and nonlinear position controller module;

[0185] The error conversion module is used to convert the actual constrained tracking error e1 into a controllable unconstrained error ε1; the Gaussian kernel-preset performance function module is used to use the designed Gaussian kernel-preset performance function as a constraint condition of the error conversion module; the nonlinear position controller module is used to track the desired trajectory x 1d ;

[0186] The error conversion module, Gaussian kernel-preset performance function module and nonlinear position controller module adopt the design method of the flexible joint preset performance function for human-computer interaction scenarios;

[0187] Interaction module: includes flexible joints and external interaction environment, used to perform human-computer interaction tasks.

[0188] In this embodiment, the virtual inertia M of the desired admittance model d 、Virtual damping B d , virtual stiffness K d Set to M d =1,B d =5, K d =3.

[0189] like Figure 3 As shown, in the human-computer interaction process of this embodiment, two sudden external torques are artificially applied.

[0190] In this embodiment, the control system parameters based on Gaussian kernel-preset performance function (GK-PPF) are set as follows: positive constants c1=150, c2=450, c3=100, c4=330; time constants r2=0.01, r3=0.01, r4=0.02, w=1 / 4388; where the parameters of GK-PPF are set as δ =1, ρ0=0.03,ρ ∞ =0.005, γ=2, h1=h2=0.01, a1=12.345, a2=15.501, b1=0.4, b2=0.4.

[0191] As a comparison, a control system based on a traditional preset performance function (PPF) is simulated, and the parameters are set as follows: normal constants c1 = 150, c2 = 450, c3 = 100, c4 = 330; time constants r2 = 0.01, r3 = 0.01, r4 = 0.02, w = 1 / 4388; among which the parameters of PPF are set as δ =1, ρ0=0.03,ρ ∞ =0.015,γ=2.

[0192] like Figure 4 As shown, the reference trajectory x is given 1r The desired trajectory x is formed by superimposing the trajectory corrected by the admittance controller on the reference trajectory 1d , x 1d Used for tracking of nonlinear position controllers.

[0193] like Figure 5 As shown, compared with the tracking error in the comparative example, the inner loop position tracking error of the control system in this embodiment not only has a smaller fluctuation range of the steady-state error, but also has improved transient performance of the tracking error when a sudden change occurs in the external torque.

[0194] Table 2 shows the statistics of tracking error indicators of control methods based on different preset performance functions. It can be seen from Table 2 that the mean tracking error of the control method based on GK-PPF is reduced by 35.05% compared with the control method based on PPF, which illustrates the applicability of this method.

[0195] Table 2: Tracking error indicators of control methods based on different preset performance functions

[0196] Preset performance functions Maximum value (rad) Mean(rad) Root mean square error (rad) GK-PPF 0.0088 <![CDATA[7.90×10 -4 ]]> 0.0013 PPF 0.0138 0.0012 0.0021

[0197] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the embodiments of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by those skilled in the art without departing from the scope and essence of the embodiments of the present invention shall be within the scope of the rights of the embodiments of the present invention.

Claims

1. A method for designing a preset performance function of a flexible joint for a human-computer interaction scenario, comprising: Step 1: Establish a rigid-flexible coupling dynamic model of the flexible joints of the collaborative robot; Step 2: Define the tracking error, add a Gaussian kernel function to the traditional preset performance function, and form a Gaussian kernel-preset performance function as a constraint condition; Step 3: According to the tracking error and the Gaussian kernel-preset performance function, the error is transformed to obtain the unconstrained error; Step 4: According to the unconstrained error, the virtual control law and the actual control law of the nonlinear position controller of the flexible joint are designed by the dynamic surface control method; Step 5: Estimate the external disturbance through a dual radial basis function neural network, replace the external disturbance term in the nonlinear position controller with the estimated value, and obtain a corrected virtual control law and a corrected actual control law of the nonlinear position controller.

2. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 1, characterized in that: In step 1, the rigid-flexible coupling dynamics model of the flexible joint is expressed as: Among them, τ m is the input torque of the motor end of the flexible joint to be determined, that is, the actual control law after correction of the nonlinear position controller; J m is the motor end inertia, J l is the connecting rod end inertia, B m is the motor end viscous damping, B l is the viscous damping at the connecting rod end; N is the reduction ratio of the harmonic reducer, τ ext is the external torque, K is the joint stiffness, m is the load mass, g is the weight acceleration, and l is the connecting rod length; E1 is the non-matching interference, and E2 is the matching interference; x1 represents the motor end position, x2 represents the motor end speed, x3 represents the connecting rod end position, and x4 represents the connecting rod end speed.

3. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 1, characterized in that: In step 2, the tracking error is defined as: e1=x1-x 1d , Among them, x 1d is the desired motion trajectory of the connecting rod, δ and is a positive constant, ρ(t) is the Gaussian kernel-preset performance function, and t represents time.

4. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 3, characterized in that: The Gaussian kernel-preset performance function ρ(t) is expressed as: ρ(t)=(ρ0-ρ ∞ -r g (t))exp(-γt)+ρ ∞ +r g (t) Among them, ρ0, ρ ∞ ,γ is a positive constant, ρ g (t) is a Gaussian kernel function; The Gaussian kernel function ρ g (t) is expressed as: Where n is the number of external torque mutations during human-computer interaction, h j 、a j 、b j Respectively The height, center point, and width of the 5. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 1, characterized in that: In step 3, the error transformation function φ(ε1) used in the error transformation is expressed as: Among them, ε1 represents the unconstrained error.

6. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 5, characterized in that: The condition for achieving the preset performance of the tracking error e1 is that the unconstrained error ε1 converges.

7. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 1, characterized in that: In step 4, the virtual control law of the nonlinear position controller is for: Among them, c2 is a positive constant, w is the introduced adjustment factor, and Ω1 is the external interference term; In step 4, the actual control law τ of the nonlinear position controller m0 It is expressed as: Among them, c4 is a positive constant, z4 is the virtual tracking error, and Ω2 is the external interference term.

8. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 1, characterized in that: In step 5, the virtual control law of the final nonlinear position controller is And the actual control law τ m It is expressed as: in, and are the estimated values ​​of external disturbance terms Ω1 and Ω2, respectively.

9. The method for designing a preset performance function of a flexible joint for a human-computer interaction scenario according to claim 8, characterized in that: The external disturbance estimation by dual radial basis function neural network is expressed as: in, are the estimates of the weight parameter vectors of the two radial basis neural networks, and ψ1 is the output of the Gaussian basis function of the hidden layers of the two radial basis neural networks; the total number of nodes in the hidden layers of the two radial basis neural networks is and Take ξ=[x1,x2,x3,x4] T As input to two radial basis neural network models.

10. A flexible joint preset performance control system for human-computer interaction scenarios, comprising: Outer loop module: including the expected admittance model module, for a designed reference trajectory x 1r When performing human-computer interaction tasks, the external force is converted into the interaction trajectory x through the expected admittance model 1c The interaction trajectory is superimposed on the reference trajectory to form the desired trajectory x 1d ; Inner loop module: including error conversion module, Gaussian kernel-preset performance function module and nonlinear position controller module; The error conversion module is used to convert the actual constrained tracking error e1 into a controllable unconstrained error ε1; the Gaussian kernel-preset performance function module is used to use the designed Gaussian kernel-preset performance function as a constraint condition of the error conversion module; the nonlinear position controller module is used to track the desired trajectory x 1d ; The error conversion module, Gaussian kernel-preset performance function module and nonlinear position controller module adopt the design method of the flexible joint preset performance function of any one of claims 1 to 9; Interaction module: includes flexible joints and external interaction environment, used to perform human-computer interaction tasks.

Images