A rehabilitation robot adaptive interaction control method integrated with controllable damping

Abstract

This invention relates to the field of rehabilitation robot control technology, specifically an adaptive interactive control method for rehabilitation robots incorporating controllable damping. To simultaneously achieve joint position control within limits and adaptive variable-intensity human-machine interactive training for the rehabilitation robot, a hysteresis-fuzzy-PD (MD-Fuzzy-PD) inner-loop controller is designed by establishing a human upper limb resistance model. This enables the rehabilitation robot to accurately track the target position, ensuring response speed and stability while reducing position overshoot, avoiding dangerous postures, and ensuring the safety of exercise training. For variable-intensity training requirements, an adaptive trajectory generator based on Lyapunov functions is designed to correct the reference trajectory through interactive forces. By estimating model parameters online, the desired trajectory is calculated and corrected in real time, and then the inner-loop controller completes real-time tracking of the desired trajectory, ultimately achieving active and compliant human-machine interactive training.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation robot control technology, specifically an adaptive interactive control method for rehabilitation robots incorporating controllable damping. Background Technology

[0002] Exoskeleton rehabilitation robots are important medical devices that assist patients in the reconstruction of damaged motor nerves. Their human-computer interaction modes are generally divided into passive training and active training. Passive training is mainly aimed at problems such as muscle relaxation, joint atrophy, and lack of motor ability in the affected limb. The rehabilitation device drives the affected limb to move along the target trajectory. For patients whose muscle strength has recovered to a certain extent, active training is required in combination with their movement intentions to improve patient participation and accelerate the rehabilitation process. During the exercise training, the trajectory tracking performance of the rehabilitation robot is crucial to improving the rehabilitation effect of patients with upper limb dysfunction. Summary of the Invention

[0003] To simultaneously achieve joint position control within limits and adaptive variable intensity human-computer interaction training for rehabilitation robots, this paper proposes an adaptive interactive control method for rehabilitation robots incorporating controllable damping.

[0004] By adopting a structure mode in which a magnetorheological damper and a joint motor are connected in parallel, electromagnetic damping torque is introduced into the dynamic equation of the rehabilitation robot. A resistance dynamic model of the human upper limb is established, and a hysteresis fuzzy PD (MD-Fuzzy-PD) inner loop controller is designed. This controller takes joint angular velocity and angular acceleration as inputs and calculates and adjusts the excitation current value and joint torque gain coefficient of the damping torque in real time. This enables the rehabilitation robot to accurately track the target position, ensuring response speed and stability while reducing position overshoot, thereby avoiding falling into dangerous postures and ensuring the safety of exercise training.

[0005] To address the need for variable intensity training, an adaptive trajectory generator based on Lyapunov functions was designed as an outer loop controller, which corrects the reference trajectory through interactive force. This controller takes interactive force, joint angular velocity, and angular acceleration as inputs, calculates and corrects the desired trajectory in real time, and achieves adaptive and variable intensity training of human-computer interaction force by adjusting the stiffness and position parameters of the adversarial model. By estimating the adversarial model parameters online, the desired trajectory is calculated and corrected in real time, and then the inner loop controller completes the real-time tracking of the desired trajectory, ultimately realizing active and compliant human-computer interaction training.

[0006] Compared with the prior art, the present invention has the following characteristics:

[0007] This method can achieve variable intensity training simply by adjusting the stiffness parameters, while simultaneously enabling the rehabilitation robot to accurately track the target position. It ensures response speed and stability while reducing position overshoot, avoiding dangerous postures, and guaranteeing the safety of motion training. By estimating model parameters online, the desired trajectory is calculated and corrected in real time, ultimately achieving adaptive human-machine interaction force and active compliant human-machine interaction training. Attached Figure Description

[0008] Figure 1 A human-computer interaction model for a two-degree-of-freedom rehabilitation robot;

[0009] Figure 2 Diagram of an adaptive impedance control strategy incorporating controllable damping;

[0010] Figure 3 The simulation results show the elbow joint angle step response.

[0011] Figure 4 The simulation results show the step response of the shoulder joint angle.

[0012] Figure 5 The simulation results show the elbow joint angle.

[0013] Figure 6 Simulation results of elbow joint angle tracking error;

[0014] Figure 7 Simulation results of shoulder joint angle;

[0015] Figure 8 Simulation results of shoulder joint angle tracking error;

[0016] The labels in the attached diagram are explained as follows:

[0017] 1. Joint motor; 2. Magnetorheological damper. Detailed Implementation Plan

[0018] The following is in conjunction with the appendix Figure 1-8 The present invention will be further described as follows:

[0019] Upper limb rehabilitation robots primarily target the training of upper arm and forearm muscles. Therefore, this study focuses on a typical shoulder-elbow two-degree-of-freedom rehabilitation robot, establishing its resistance model, such as... Figure 1 As shown, the joint motor 1 and the magnetorheological damper 2 are connected in parallel, and the superimposed output torque of the two is used as the total output torque of the joint of the rehabilitation robot.

[0020] In free space, the dynamic equations of the rehabilitation robot are established using the Newton-Euler recursive method: In the formula, D(θ) is the robot's inertia matrix. Here are the centrifugal force matrix and Coriolis torque, and G(θ) is the gravity term. This is the term for joint friction. These represent the angular displacement, angular velocity, and angular acceleration of each joint, respectively, with Γ representing the control torque. e Torque for various interfering factors.

[0021] When there is an interaction force with the upper limb of the human body, press Figure 1 As shown in the annotation, For the desired position; It represents the actual position of the exoskeleton and characterizes the actual movement trajectory of the rehabilitation equipment; M d B d ,K d ∈R n These are the mass, damping, and stiffness matrices of the controlled object; K e ∈R n It is the muscle stiffness coefficient of the patient's upper limb, B e ∈R n It is the muscle damping coefficient matrix, X e ∈R n ,X e =Θ d The position of the upper limbs can be obtained through kinematic calculations using θ1 and θ2; F e ∈R n It is the resistance force of muscle stiffness, K e ∈R n The parameter characterizes the training intensity in the active mode of rehabilitation training. F d ∈R n It is an interactive force actively applied by the patient. According to Hooke's Law, the resistance dynamics model is as follows: With the force and desired position unchanged, the robot's actual trajectory can be altered by changing the stiffness coefficient and damping coefficient.

[0022] By utilizing the control torque output from the parallel connection of a joint motor and a magnetorheological damper, inner-loop trajectory tracking is achieved, and a hysteresis fuzzy PD control law for the joint motion of the rehabilitation robot is established: u = γ i τ PD -λ i M t , where i=1,2,τ PD M t These are the joint control torque and the damping torque, γ i ≥0,λ i ≥0 represents the control torque gain coefficient, which is the value when the joint position of the exoskeleton robot is close to the desired position, i.e., θ. d -θ=θ s →0,γ i →0,λ i →1, weaken τ PD The control effect, while enhancing M tThe hysteresis effect is utilized to ensure system response speed while reducing system overshoot.

[0023] The control law for the joint motor is established as follows:

[0024]

[0025] In the formula, K p0 K d0 Here are the PD parameters from the last tuning, k is the sampling sequence, and e, e c These are joint angle deviation and deviation change rate, respectively.

[0026] Establish the relationship between excitation current I and joint angular acceleration Mapping between: Where T vt M is the viscous torque. t J is the damping torque of the magnetorheological damper. M K represents the maximum rotational inertia of the end-effector load of the rehabilitation machine relative to each joint. t β is the mapping coefficient, and β is the material property coefficient; the differential equation of motion of the rehabilitation robot after incorporating magnetorheological damping torque: After applying the damping torque of the magnetorheological damper, the system can quickly and stably track the desired trajectory while suppressing overshoot and oscillation. This magnetorheological damping torque can reduce the overshoot M of the fuzzy PD controller. p Further simplification yields: Where R2 is the maximum outer diameter of the magnetorheological fluid surface that generates the rheological effect, and R3 is the inner diameter of the magnetorheological damper, the excitation current I and the joint motion angular acceleration are obtained. and joint motion angular velocity The relationship between overshoot M p It can be expressed as: M p =(θ max -θ(∞)) / θ(∞), where, θ max Let θ(∞) be the peak value of the joint displacement, and θ(∞) be the joint displacement value in the steady state. Further, from: θ s →0, that is, θ→θ d , θ≤θ d ,get: In the formula, M RP To incorporate the overshoot after controllable damping, the hysteresis characteristics of the magnetorheological damper are incorporated, and the actual joint displacement continuously approaches the desired value. This suppresses the peak value of the joint displacement in the early stage of the joint response, thereby reducing the overshoot of the system.

[0027] The commonly used ideal impedance control model expression is: In the formula, M d B d ,K d ∈R2 These are the impedance parameters of the control model; These are the actual acceleration, velocity, and displacement vector of the end effector of the rehabilitation robot; It is the corresponding desired motion trajectory vector; F e ∈R 2 This is the resistance vector exerted by the robot's end effector on the patient's upper limb during traction. Considering the impedance control of the rehabilitation robot system in a single direction, and setting the position error e = xx... d Substituting into the expression of the ideal impedance control model, we get: Force F e With interaction force F d The difference is the force correction error e f The position deviation E is calculated using an impedance controller and then compared with the desired trajectory x. d The trajectory error is corrected by comparison, and finally the trajectory of the rehabilitation robot end effector is tracked by the inner loop controller, thus achieving dynamic coordination between human-machine interaction force and the end effector position of the rehabilitation equipment.

[0028] The steady-state error of the control system can be expressed as: e ss =F d +k d k e / (k d +k e (x) e -x d ), F ss =k d k e / (k d +k e (x) d -x e In the formula, e ss For steady-state error, F ss To eliminate contact force error, steady-state error, i.e., e, must be eliminated. ss →0, requires contact force error F ss →F d That is, when x d =F d / k eq +x e At that time, F ss →F d e ss →0, where k eq =k d k e / (k d +k e ) represents the system stiffness.

[0029] To achieve dynamic coordination between human-computer interaction force and the position of the rehabilitation equipment's end effector, the force F must first be realized.e Human-computer interaction force F d Stable tracking and online estimation of environmental parameter k e x e These estimates are used to calculate the desired reference position trajectory x in real time. d Thus achieving force F e Human-computer interaction force F d Following this, the patient's upper limbs can ultimately achieve smooth interaction with the rehabilitation equipment.

[0030] Define a positive Lyapunov equation: V = Ψ T ΠΨ, where Π is a 2×2 positive symmetric constant matrix, let but Depend on It can be seen that Π is a semi-negative definite matrix, and when t→∞... Established.

[0031] The parameter estimation algorithm for the outer loop impedance strategy of adaptive trajectory generation is as follows:

[0032]

[0033] The motion trajectory correction algorithm for rehabilitation robots is as follows: The desired trajectory in one direction for the adaptive trajectory generation outer loop impedance strategy is: in, These are respectively used as resistance stiffness, system stiffness, and position k. e k eq x e The estimated value, m d ,b d ,k d These are the impedance parameters of the control model, F d For interactive force, It is the contact force F e The estimated value, C dcp For force / position coordination gain, constants ξ1, ξ2 > 0.

[0034] Figure 2 The diagram shows an adaptive impedance control strategy incorporating controllable damping, consisting of two parts: an inner-loop position control strategy incorporating controllable damping and an outer-loop impedance strategy that adaptively generates the trajectory.

[0035] The dynamic response of joint displacement is shown in Table 1.

[0036] Table 1. Elbow / shoulder joint step response.

[0037]

[0038] From Table 1 and Figure 3-8It can be seen that, compared with traditional controllers and fuzzy controllers, the controller proposed in this invention has the smallest joint position control error for rehabilitation robots. Compared with traditional controllers, the error is reduced by 99.34% and 98.82% for shoulder and elbow joints, respectively. In addition, the controller achieves smaller joint angular displacement overshoot and shorter adjustment time. Compared with fuzzy controllers, the overshoot of shoulder and elbow joints is reduced by 97.5% and 96.2%, respectively, and the adjustment time is shortened by 67.2% and 76.6%, respectively. It achieves better position control effect and significantly improves response speed during the training phase of rehabilitation robots.

Claims

1. An adaptive interactive control method for a rehabilitation robot incorporating controllable damping, comprising an inner-loop position control strategy incorporating controllable damping and an outer-loop impedance control strategy for adaptive trajectory generation, characterized in that: By establishing a resistance model of the human upper limb, a hysteresis fuzzy PD (MD-Fuzzy-PD) inner-loop controller was designed to achieve precise tracking of the target position by the rehabilitation robot. This ensures response speed and stability while reducing position overshoot, avoiding dangerous postures, and guaranteeing the safety of exercise training. For variable-intensity training requirements, an adaptive trajectory generator based on Lyapunov functions was designed to correct the reference trajectory through interactive forces. This generator estimates the adversarial model parameters online, calculates and corrects the desired trajectory in real time, and then uses the inner-loop controller to complete real-time tracking of the desired trajectory, ultimately achieving active and compliant human-machine interactive training. The hysteresis fuzzy PD control law for the joint movements of the rehabilitation robot is as follows: In the formula , and These are the joint control torque and damping torque, respectively, and the control torque gain coefficient. The parameter estimation algorithm for the adaptive trajectory generation outer loop impedance strategy is as follows: ; The motion trajectory correction algorithm for rehabilitation robots is as follows The desired trajectory of the exoskeleton in one direction is: In the formula These are resistance stiffness, system stiffness, and position. The estimated value, These are the impedance parameters of the control model. It is interactive force. It is contact force The estimated value, It is the force / position coordination gain, a constant. ,constant .

2. The adaptive interactive control method for a rehabilitation robot incorporating controllable damping according to claim 1, characterized in that: The dynamic equation of the rehabilitation robot incorporating controllable damping is as follows In the formula It is the robot's inertia matrix. It consists of the centrifugal force matrix and the Coriolis torque. It is a gravity term. It is the term of joint friction. These are the angular displacement, angular velocity, and angular acceleration of each joint. It is the control torque. These are the torques of various interfering factors; when there is an interaction force with the upper limbs of the human body, the resistance model is as follows: ,in It is the muscle stiffness coefficient of the patient's upper limb. It is the muscle damping coefficient matrix. It is the desired position; The actual position of the exoskeleton is the force. When the desired position remains unchanged, the actual trajectory of the robot can be changed by altering the stiffness coefficient and damping coefficient.

Images