Exoskeleton robot impedance control method based on differential flatness

Abstract

The invention discloses an exoskeleton robot impedance control method based on differential flatness. The two-degree-of-freedom lower limb skeleton robot structure includes a patient lower limb kinetic parameter acquisition and identification part for identifying lower limb inertial parameters of a patient and providing accurate data support for personalized rehabilitation training. Comprising the following steps: constructing a kinetic model of the two-degree-of-freedom lower limb exoskeleton robot A differential flatness theory is introduced and is used for processing nonlinearity and coupling characteristics of the lower limb exoskeleton robot; a nonlinear disturbance observer is provided for estimating disturbance terms in the system and improving the robustness of the system, a fuzzy impedance controller is designed, and impedance parameters are adjusted through fuzzy logic to adapt to the interaction force of different wearers so as to ensure the comfort of human-computer interaction and the trajectory tracking performance.

Description

Technical Field

[0001] This invention relates to an impedance control method for exoskeleton robots based on differential flatness, belonging to the field of exoskeleton control technology. Background Technology

[0002] With an increasingly aging population, functional impairments caused by cardiovascular and cerebrovascular diseases, stroke, and hemiplegia are becoming a growing social concern. Against this backdrop, lower limb exoskeleton robots have emerged as an important tool for assisting rehabilitation therapy. These robots can assist patients in a series of rehabilitation exercises, from sitting up to upper and lower limb movements, significantly improving their self-care ability and quality of life. Compared to traditional rehabilitation methods, lower limb rehabilitation robots demonstrate higher safety and therapeutic effectiveness. To accommodate the interaction forces of different wearers, this invention develops an impedance control method for exoskeleton robots based on differential flatness to ensure comfortable human-machine interaction and trajectory tracking performance. Summary of the Invention

[0003] The technical solution of this invention to solve the aforementioned technical problem is to design an impedance control method for exoskeleton robots based on differential flatness, characterized by the following operation steps:

[0004] Step 1: Provide a lower limb skeletal robot structure: control system, waist support, multiple torque sensors, multiple drive devices, multiple linkages, etc.;

[0005] Step 2: Establish a two-degree-of-freedom dynamic model of the lower limb exoskeleton based on the Lagrange equation;

[0006] Step 3: Differential flatness theory is used to handle the nonlinearity and coupling characteristics of lower limb exoskeleton robots;

[0007] Step 4: Design a nonlinear disturbance observer to estimate and compensate for disturbance terms in the subsystem;

[0008] Step 5: Acquisition and identification of patient's lower limb dynamic parameters, identifying the patient's lower limb inertial parameters;

[0009] Step 6: Design a fuzzy impedance controller to dynamically adjust impedance parameters through fuzzy logic to adapt to changes in the interaction force of different wearers;

[0010] Step 7: In the simulation, the total uncertainty perturbation of the system was added, and experiments were conducted in MATLAB / Simulink.

[0011] Experimental results validate the effectiveness of the proposed differentially flattened impedance control method for exoskeleton robots in improving trajectory tracking accuracy, enhancing system robustness, and improving human-robot interaction comfort. These results provide strong support for the practical application of this control method in rehabilitation-assisted therapy. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the structure of a lower limb skeletal robot provided in an embodiment of the present invention.

[0013] Figure 2 This is an overall structural diagram provided for an embodiment of the present invention.

[0014] Figure 3 The image shows the trajectory tracking effect (hip) with impedance control and interaction force provided in an embodiment of the present invention.

[0015] Figure 4 The image shows the trajectory tracking effect of impedance control with interactive force (knee) provided in an embodiment of the present invention. Detailed Implementation

[0016] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The purpose of providing embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0018] This invention provides an impedance control method for exoskeleton robots based on differential flatness (hereinafter referred to as the method), characterized by the following operation steps:

[0019] Step 1: Provide a lower limb skeletal robot structure: control system, waist support, multiple torque sensors, multiple drive devices, multiple linkages, etc.;

[0020] example Figure 1 The schematic diagram of a lower limb skeletal robot provided by the present invention, used to combine the control algorithm of the present invention, includes: a control system 100, a hip joint motor drive device 200, a leg guard 300, a knee joint motor drive device 400, a lower leg link 500, a waist guard 600, a torque sensor 700, and a thigh link 800.

[0021] The control system 100 is used to house the control system, communication system, and power system of the exoskeleton robot.

[0022] Hip joint motor drive device 200, Figure 1 The location of the hip joint motor drive device 200 is only shown as an example; it is actually distributed on both sides of the user's hip joint.

[0023] Leg Guard 300 is used to secure the exoskeleton robot to the thigh and leg.

[0024] Knee joint motor drive device 400, Figure 1 The location of the knee joint motor drive unit 400 is only shown as an example; it is actually distributed on both of the user's knee joints. It is used to connect the thigh link 800 and drive the lower leg link 500.

[0025] The 500 lower leg link is used to fix the exoskeleton robot to the lower leg.

[0026] The Waist Support 600 connects the robot to the user's waist, thus securing the robot in place.

[0027] Torque sensor 700, Figure 1 The location of the torque sensor 700 is only shown as an example; in reality, it is distributed on both sides of the user's hip joint and on both knee joints. The torque sensor 700 acquires the force exerted by the user on the thigh link 800 and the lower leg link 500, determines the torque, and sends the torque to the control system 100. Then, the control system 100 controls the hip joint motor drive device 200 and the knee joint motor drive device 400 according to the torque, driving the thigh link 800 and the lower leg link 500 to move the user's lower limbs.

[0028] Thigh link 800 is used to fix the exoskeleton robot to the thigh. Hip joint motor drive device 200 drives thigh link 800.

[0029] Step 2: Establish a two-degree-of-freedom dynamic model of the lower limb exoskeleton based on the Lagrange equation;

[0030]

[0031] Where M(q)∈R 2×2 It is the inertia matrix. It is a Coriolis matrix, G(q)∈R 2×1 It is the gravity matrix. It is the friction force matrix, d(t)∈R 2×1 This is the interference matrix under consideration. The driving torque of the two motors is τ = [τ1, τ2]. T The interaction torque between the two joints is τ int ∈R 2×1 .

[0032] M(q), G(q), It can be represented as follows:

[0033] M(q)=M0(q)+M e (q),

[0034] G(q)=G0(q)+G e (q),

[0035] Where M0(q), G0(q), M is the nominal term identified by the model. e (q), G e (q), Let this be the modeling error term. Substituting it into the above equation, we get:

[0036]

[0037] Step 3: Using the differential flatness theory, the nonlinearity and coupling characteristics of the lower limb exoskeleton robot are addressed:

[0038] Differential flatness describes a system that is equivalent to a linear system through a special type of feedback called endogenousness. The state and input variables can be directly represented by the flat output and its finite number of derivatives, without integrating any differential equations.

[0039] According to the theory of differential flatness, equation (2) can be written as:

[0040]

[0041] Introduce the system's output vector y = [q1, q2]. T The state vector is x1 = [q1, q2]. T , ξ = [x1, x2] T The following state-space equations are obtained:

[0042]

[0043] The complete form of the invertible matrix M0 can be written as:

[0044]

[0045] In the above formula, j1, j2, and X are inertial parameters, which are related to the mass of the robot's exoskeleton, the length of the links, and the moment of inertia of the motors. For ease of subsystem decomposition, Expressed as follows:

[0046]

[0047] According to the theory of differential flatness, the state variables of the subsystem, x, should also be introduced. 11 =q1, x 21 =q2, The dynamic equations of a robotic exoskeleton can be written as the following subsystem:

[0048]

[0049] Rewritten as:

[0050]

[0051]

[0052] Step 4: Design a nonlinear perturbation observer to estimate and compensate for the perturbation terms in the subsystem, and use Lyapunov theory to prove that it is transformed into the stability of two subsystems; since the two subsystems have the same structure, the first subsystem is studied now. To estimate the lumped perturbation terms, the nonlinear perturbation observer is designed as follows:

[0053]

[0054] Where Z is the state variable of the observer, and L > 0 is the adjustment parameter of the observer. This is the estimated perturbation value from the observation. The observer is globally asymptotically stable. Observation error is defined as follows:

[0055]

[0056] Under the disturbance estimation term, to achieve the output trajectory position x 11 Tracking the desired trajectory x 1r The tracking error is:

[0057] z1=x 11 -x 1r

[0058] z2=x 12 -α

[0059]

[0060] Find the virtual control rate When K > 0, it represents the feedback gain. Construct the Lyapunov function for the total error:

[0061]

[0062]

[0063]

[0064] Find the control law S1 and rewrite it as follows:

[0065]

[0066] Step 5: An initial motion trajectory is given to the lower limb rehabilitation robot, and the patient performs the initial motion under the traction of the robot. Pressure sensors located at the end effector of the lower limb rehabilitation robot and angle sensors located at the rotation joints of the robot collect the interaction force between the robot's end effector and the patient's foot, as well as the rotation angle of the robot's end effector. The collected data is filtered to obtain high-precision values, which are then used as input parameters to the human-machine system dynamics model to calculate the end effector's real-time position, velocity, and acceleration. Based on the regression equation for the inertial parameters of Chinese adults, estimated values ​​of the patient's lower limb inertial parameters (leg length, mass, center of mass, and moment of inertia) are obtained from the patient's height and weight data. The binary regression equation is:

[0067] y (m，c) =B0+B1g+B2l

[0068] I (x，y，z) =K0+K1g+K2l

[0069] Where y (m，c) For the patient's lower limb mass or center of mass position, I (x，y，z) The moment of inertia of the patient's lower limbs about the frontal axis, sagittal axis, and vertical axis are given, g is the body weight, l is the height, and B0, B1, B2 and K0, K1, K2 are the regression equation coefficients provided in the standard library.

[0070] The estimated moment of inertia of the patient's lower limb about the rotational joint can be obtained using the parallel axis theorem:

[0071] I = I (x，y，z) +y m y c 2

[0072] Based on patient motion data, a cost function is constructed to calculate the interaction force and expected interaction force between the lower limb rehabilitation robot end effector and the patient's foot, and constraints are introduced regarding the estimated values ​​of the patient's lower limb inertial parameters:

[0073]

[0074] Where m represents the m sets of interactive forces generated by the lower limb rehabilitation robot under a given initial motion trajectory, and τ e τ represents the actual interaction force collected. d This represents the expected interaction force between the lower limb rehabilitation robot and the patient, calculated from the current updated values ​​of the patient's lower limb parameters using a dynamic model. k represents the inertial parameters of the patient's lower limb. e These represent the estimated values ​​of the patient's lower limb inertial parameters calculated using regression equations from the national standard library. The update equations for the patient's lower limb inertial parameters using the gradient descent method are as follows:

[0075]

[0076] Where k0 represents the inertial parameters of the patient's lower limbs before the update, and α is the weighting coefficient. Let be the partial derivative of the cost function with respect to the inertial parameters of each lower limb of the patient. The initial update values ​​of each inertial parameter are the estimated values ​​calculated using the regression equation from the national standard library. Through the above algorithm, the inertial parameters of the patient's lower limbs are iteratively updated until the cost function converges to its minimum value, thus yielding the optimal solution that best approximates the true values ​​of the patient's lower limb inertial parameters.

[0077] Step 6: Design a fuzzy impedance controller to dynamically adjust impedance parameters through fuzzy logic to adapt to changes in the interaction force of different wearers;

[0078] Human-computer interaction often results in interaction forces with significant range variations due to individual differences among wearers. To improve the system's universality, control accuracy, and compliance, adaptive fuzzy impedance control is employed, where the impedance parameter matrix is ​​variable. The fuzzy impedance model is as follows:

[0079]

[0080] Where, Δe=q d -q, q d For the desired joint angle, τ b The torque generated by impedance control. K, D, and M are the stiffness, damping, and mass matrices, respectively. Δe, These are position error, velocity error, and acceleration error, respectively.

[0081] Considering the low-speed application scenario of wearable exoskeletons studied in this paper, and to simplify the fuzzy control rules, the acceleration term is neglected. The impedance control model of the system is rewritten as follows:

[0082]

[0083] The fuzzy rules are designed based on the characteristics of the exoskeleton. Through numerous experiments, different forces were applied, resulting in varying position and velocity errors, which led to the design of impedance parameters. The establishment of the fuzzy rules depends on the characteristics of the exoskeleton, and once designed, the fuzzy rules can adapt to different wearers.

[0084] The fuzzy rule table is as follows:

[0085]

[0086] Step 7: In the simulation, the total uncertainty disturbance of the system was added, and three comparative experiments were carried out in MATLAB / Simulink: trajectory tracking with UAV interaction force, trajectory with interaction force and no impedance control, and trajectory with interaction force and impedance control.

[0087] In the simulation, the total uncertainty disturbance of the system is added. as follows:

[0088]

[0089] Figure 3 The trajectory tracking effect of fuzzy impedance control of the hip joint under human-computer interaction force. Figure 4 The trajectory tracking effect of fuzzy impedance control of the knee joint under human-computer interaction force.

Claims

1. An impedance control method for an exoskeleton robot based on differential flatness, characterized in that, The operation steps are as follows: S1. A lower limb skeletal robot structure is provided, including a control system, a waist support, multiple torque sensors, multiple drive devices, and multiple linkages. S2. Establish a two-degree-of-freedom dynamic model of the lower limb exoskeleton based on the Lagrange equation; S3, Differential Flatness Theory, is used to handle the nonlinearity and coupling characteristics of lower limb exoskeleton robots; S4. Design a nonlinear disturbance observer to estimate and compensate for disturbance terms in the subsystem; S5. The patient's lower limb dynamic parameters acquisition and identification section identifies the patient's lower limb inertial parameters; S6. Design a fuzzy impedance controller to dynamically adjust impedance parameters through fuzzy logic to adapt to the changes in interaction force among different wearers.

2. The lower limb skeletal robot as described in claim 1, characterized in that, Step S1 describes a lower limb skeletal robot structure that includes: a control system, a lumbar support, multiple torque sensors, multiple drive devices, multiple linkages, etc. The control method applicable to this invention.

3. The control method as described in claim 1, characterized in that, The dynamic model of the two-degree-of-freedom lower limb exoskeleton in step S2 is described as follows: Where M(q)∈R 2×2 It is the inertia matrix. It is a Coriolis matrix, G(q)∈R 2×1 It is the gravity matrix. It is the friction force matrix, d(t)∈R 2×1 This is the interference matrix under consideration. The driving torque of the two motors is τ = [τ1, τ2]. T The interaction torque between the two joints is τ int ∈R 2×1 . M(q), G(q), It can be represented as follows: M(q)=M0(q)+M e (q), G(q)=G0(q)+G e (q), Where M0(q), G0(q), M is the nominal term identified by the model. e (q), G e (q), Let this be the modeling error term. Substituting it into the above equation, we get:

4. The control method as described in claim 1, characterized in that, In step S3, the differential flatness theory is used to handle the nonlinearity and coupling characteristics of the lower limb exoskeleton robot. Differential flatness describes a system that is equivalent to a linear system through a special type of feedback called endogenousness. The state and input variables can be directly represented by the flat output and its finite number of derivatives, without integrating any differential equations. According to the theory of differential flatness, equation (2) can be written as: Introduce the system's output vector y = [q1, q2]. T The state vector is x1 = [q1, q2]. T , ξ = [x1, x2] T The following state-space equations are obtained: The complete form of the invertible matrix M0 can be written as: In the above formula, j1, j2, and X are inertial parameters, which are related to the mass of the robot's exoskeleton, the length of the links, and the moment of inertia of the motors. For ease of subsystem decomposition, Expressed as follows:

5. The control method as described in claim 1, characterized in that, In step S4, the nonlinear disturbance observer is designed as follows: Where Z is the state variable of the observer, and L > 0 is the adjustment parameter of the observer. This is the estimated perturbation value from the observation. The observer is globally asymptotically stable.

6. The control method as described in claim 1, characterized in that, In step S5, the identification of the patient's lower limb dynamic parameters uses a regression algorithm, introducing constraint terms regarding the estimated values ​​of the patient's lower limb inertial parameters: Where m represents the m sets of interactive forces generated by the lower limb rehabilitation robot under a given initial motion trajectory, and τ e τ represents the actual interaction force collected. d This represents the expected interaction force between the lower limb rehabilitation robot and the patient, calculated from the current updated values ​​of the patient's lower limb parameters using a dynamic model. k represents the inertial parameters of the patient's lower limb. e This represents the estimated values ​​of the patient's lower limb inertial parameters calculated using regression equations from the national standard library. The update equations for the patient's lower limb inertial parameters using the gradient descent method are as follows: Where k0 represents the inertial parameters of the patient's lower limbs before the update, and α is the weighting coefficient. Let be the partial derivative of the cost function with respect to the inertial parameters of each lower limb of the patient. The initial update values ​​of each inertial parameter are the estimated values ​​calculated using the regression equation from the national standard library. Through the above algorithm, the inertial parameters of the patient's lower limbs are iteratively updated until the cost function converges to its minimum value, thus yielding the optimal solution that best approximates the true values ​​of the patient's lower limb inertial parameters.

7. The control method as described in claim 1, characterized in that, In step S6, the fuzzy impedance model is as follows: Where, Δe=q d -q, q d For the desired joint angle, τ b The torque generated by impedance control. K, D, and M are the stiffness, damping, and mass matrices, respectively. Δe, These are position error, velocity error, and acceleration error, respectively. Considering the low-speed application scenario of wearable exoskeletons studied in this paper, and to simplify the fuzzy control rules, the acceleration term is neglected. The impedance control model of the system is rewritten as follows: The fuzzy rules are designed based on the characteristics of the exoskeleton. Through numerous experiments, different forces were applied, resulting in varying position and velocity errors, which led to the design of impedance parameters. The establishment of the fuzzy rules depends on the characteristics of the exoskeleton, and once designed, the fuzzy rules can adapt to different wearers. The fuzzy rule table is as follows:

Images