Three-link control method for lower extremity exoskeleton based on dynamic motion primitives

Abstract

The application discloses a lower limb exoskeleton three-loop control method based on dynamic motion primitives, which is applied to the field of exoskeleton robots and aims at the problem of poor compliance of existing exoskeleton man-machine coupling. According to the body characteristic parameters of a wearer, a relatively reasonable gait trajectory is selected from a gait trajectory library as an initial training trajectory of DMP, the output trajectory of DMP is used as the input trajectory of a lead-lag control loop, the output trajectory of the lead-lag control loop is used as a reference trajectory, and the reference trajectory is used as a tracking target of an inner-loop position controller. The establishment of a man-machine coupling model needs to be considered in the design of the inner-loop position controller. The actual trajectory of the exoskeleton output by the position controller is used as the input trajectory of DMP in the next round of training, and the process is repeated to complete the closed-loop active control of the exoskeleton, and the control target is that the man-machine coupling effect is as small as possible, and the work task of the man-machine cooperation is completed.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of exoskeleton robots, and particularly relates to a lower limb exoskeleton control technology. BACKGROUND

[0002] As a typical wearable robot, an exoskeleton robot, in combination with a human body, forms a man-machine coupling system, and has the advantages of high mechanical strength, high load power of the robot and environmental perception and task analysis ability of the human body, and has wide application value in medical rehabilitation, industrial production and the aid of the old and the disabled. As an auxiliary walking device, a lower limb exoskeleton robot couples the mechanical structure of the exoskeleton and the human legs together, and enables the operator who is inconvenient to move or unable to walk to walk autonomously through human control and external power supply, and different gaits and speeds can be designed to adapt to patients in different situations.

[0003] The lower limb exoskeleton mainly consists of the following parts:

[0004] (1) Mechanical structure: The lower limb exoskeleton usually adopts a hip+knee+ankle or hip+knee structure according to its functional requirements. For example, a rehabilitation type exoskeleton robot is usually used for patients, and needs to reduce the movement of the joints, so the latter structure is usually adopted. The material used for the mechanical structure of the exoskeleton should have the characteristics of light weight, high strength and fatigue resistance, such as carbon fiber, aluminum alloy, titanium alloy and nanometer material.

[0005] (2) Power system: The power system mainly provides power for the assistance of the exoskeleton, and the provided mode can be hydraulic, pneumatic and motor, etc. If a motor is used for driving, the motor drives the device to complete the corresponding man-machine coupling task according to the real-time control instruction.

[0006] (3) Sensor system: The sensor system of the exoskeleton is mainly used to measure and perceive the real-time running state of the exoskeleton prototype, acquire various signals in man-machine interaction, judge the human gait or motion intention, and formulate the exoskeleton control strategy and algorithm. Commonly used sensors include three-dimensional force sensors, absolute encoders and torque sensors.

[0007] (4) Control system: After the control algorithm and related methods proposed are realized by using software such as Matlab / Simulink, they are downloaded into the corresponding hardware controller.

[0008] At present, in the prior art, the exoskeleton is difficult to control according to the intention of the wearer, so that the compliance of the man-machine coupling is poor, the man-machine interaction force is also too large, and there is a certain risk of injury for some rehabilitation patients, which is not conducive to assisting patients to complete rehabilitation training. SUMMARY

[0009] To solve the above technical problems, the application provides a lower limb exoskeleton three-loop control method based on dynamic movement primitives, adopts DMP (Dynamic Movement Primitives) for learning and planning an initial trajectory of an exoskeleton from a teaching trajectory, adopts an admittance controller to ensure the compliance of human-machine coupling and guide the exoskeleton to complete a specific task, and adopts a position controller to track an ideal trajectory and improve the control accuracy of the exoskeleton, thereby completing the human-machine coupling task.

[0010] The technical scheme adopted by the application is as follows: a lower limb exoskeleton three-loop control method based on dynamic movement primitives, comprising:

[0011] S1, selecting a gait trajectory from a gait trajectory library as an initial training trajectory q of an improved DMP algorithm module according to the body characteristic parameters of a wearer; init ;

[0012] S2, taking the output trajectory of the improved DMP algorithm module as an input trajectory of an admittance controller;

[0013] S3, taking the output trajectory of the admittance controller as a reference trajectory, and taking the reference trajectory as a tracking target of an exoskeleton model;

[0014] S4, taking the actual trajectory of the exoskeleton output by a position controller in a current exoskeleton model tracking process as a DMP input trajectory for next round of training, and completing closed-loop active control of the exoskeleton.

[0015] The application has the following beneficial effects: the application provides a three-loop control scheme, the outer loop DMP is used for trajectory learning, the middle loop admittance controller is used to ensure the compliance of human-machine coupling, and the inner loop position controller is used to track an ideal trajectory. The control scheme adopted by the application is used for motor driving of an exoskeleton device, and can effectively improve the response capability and tracking precision of the exoskeleton device; the method of the application can make the exoskeleton motion trajectory gradually approach the real trajectory of the wearer, improve the compliance of the human-machine coupling system, reduce the human-machine interaction torque, realize stable control of the lower limb exoskeleton system, and ensure the completion of the specified task. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The application is designed for an exoskeleton three-loop control block diagram. DETAILED DESCRIPTION

[0017] To facilitate the technical personnel in the art to understand the technical content of the application, the content of the application is further explained below in combination with the drawings.

[0018] As Figure 1The control block diagram shown primarily consists of a human standard gait trajectory library, an improved DMP algorithm module, a position controller, an exoskeleton model, a human model, and an admittance controller. The human standard gait trajectory library contains walking data of normal individuals with different body characteristics. In the initial training phase, a relatively reasonable gait trajectory can be selected from the library based on the wearer's body characteristic parameters as the initial training trajectory q for the DMP. init The reference trajectory q is planned using the improved DMP algorithm module. r As the input to the admittance controller, Δq is used to determine the reference trajectory q. r After making corrections, the true desired trajectory q of the position controller is obtained. d .Depend on Figure 1 It is known that admittance controllers need to be used in conjunction with position controllers, and existing experience also shows that the performance of position controllers largely determines the performance and stability of admittance controllers.

[0019] △q represents the difference between the desired trajectory and the reference trajectory.

[0020] The method of the present invention includes the following processes:

[0021] 1. Based on the wearer's body characteristic parameters, select a relatively reasonable gait trajectory from the gait trajectory database as the initial training trajectory q for the DMP. init The output trajectory of the improved DMP algorithm module serves as the input trajectory of the admittance controller;

[0022] Those skilled in the art will understand that the different physical characteristics referred to here are an individual's height and weight, and the gait trajectories stored in the gait trajectory database.

[0023] In this embodiment, through a given teaching trajectory The model parameters of DMP can be obtained. The DMP algorithm, based on a stable dynamic system, modulates the system by introducing a nonlinear forcing function, ultimately bringing the system to the desired attractor state. A dynamic model is established using a spring-damped model:

[0024]

[0025] Among them, y demo , These represent the position, velocity, and acceleration of the teaching trajectory, respectively, with y generally representing the position. and These represent velocity and acceleration, respectively, where g represents the system's expected target, and α... y and β y τ is the gain coefficient, and f is the introduced nonlinear forcing term. τ acts as the time scaling factor for the control system; adjusting this constant modifies the velocity at which the trajectory converges to the target g.

[0026] Since the classic DMP cannot fit the curve with very close start and end points, in order to solve this problem, in the present application, an improved DMP is adopted:

[0027]

[0028] Wherein, y0 represents the initial state, K 00 Take a constant, in order to keep consistent with the system, take K 00 = a y , x comes from a first-order system, f is a radial basis function, and the expression is:

[0029]

[0030] N represents the number of basis functions; x represents a time-independent quantity

[0031] The basis function ψ i is a Gaussian function with c i as the center and h i as the variance, and ψ i = exp(-h i (x-c i ) 2 ).

[0032] When the given demonstration trajectory is given, the target value f target of the radial basis function f is defined as:

[0033]

[0034] In this embodiment, for f the parameter ω is solved and obtained by using a locally weighted regression method (LWR), because the DMP method only needs to be learned once, and the LWR can realize fast learning, and each Gaussian kernel function learning process is independent. The nonlinear function to be solved can be expressed as shown in the following formula:

[0035]

[0036] The value of the nonlinear function can be obtained by giving the demonstration trajectory, since the nonlinear function is composed of basis functions by weighted superposition, the loss function J can be constructed by using the LWR optimization method, as shown in the formula:

[0037]

[0038] ξ(t) = x(t)(g-y0)

[0039] The loss function is minimized by an optimization method, where P represents the number of time steps of the trajectory, for discrete dynamic motion primitives, the parameter solving process is a weighted linear regression problem, and finally ω is shown as follows:

[0040]

[0041]

[0042] The admittance control loop part, the admittance controller makes the human-machine coupling action τ ext and the trajectory deviation e = Δq = q d -q r between them meet the designed second-order dynamic relationship, where q d is the true desired trajectory of the position controller, q r is the reference trajectory, and q r is the tracking target of the inner loop position control.

[0043] The linear second-order model set by the admittance controller is as follows:

[0044]

[0045] Where M, B, and K represent virtual inertia, virtual damping, and virtual stiffness parameters, respectively. According to e = Δq, the linear second-order model of the admittance controller can also be expressed as:

[0046]

[0047] 2. The output trajectory of the admittance control module is as the tracking target of the inner loop position controller;

[0048] The purpose of the intermediate loop admittance controller is to achieve the effect of human-machine following, and the input force information is the interaction force between the human and the machine, which can be obtained through the three-dimensional force sensor installed on the exoskeleton. The ideal trajectory q d output by the admittance controller is used as the input of the inner loop controller. The admittance controller needs to be used in combination with the position tracker, that is, the position tracking control of the joint needs to be realized in order to use the admittance control. In addition, existing experience shows that the performance of the underlying positioner largely determines the performance and stability of the admittance control.

[0049] According to Lagrange dynamics, the 2-DOF lower limb exoskeleton dynamics model considering human-machine coupling terms can be expressed as:

[0050]

[0051] In the formula, represent the double-joint angle, angular velocity and angular acceleration of the exoskeleton, R represents a real number set, and τ2∈R 2denotes the driving torque of the motor, τ ext ∈ R 2 denotes the human-robot coupling torque, M(q), and G(q) denote the inertia matrix, Coriolis matrix and gravity term of the system, respectively, f dis (t) denotes the lumped term composed of unknown disturbances outside the system. M(q), and G(q) contain unknown dynamic parameters.

[0052] The human-robot interaction torque calculation formula is:

[0053]

[0054] where q h denotes the real trajectory of the human body, K he denotes the spring stiffness coefficient, B he denotes the spring damping coefficient.

[0055] In this embodiment, the state variable of the exoskeleton system is set as x1 = [q1, q2] T , According to the dynamic model, the state space expression of the system is defined as:

[0056]

[0057]

[0058] where q1 denotes the angle of the exoskeleton hip joint, q2 denotes the angle of the exoskeleton knee joint, denotes the angular acceleration of the exoskeleton hip joint, denotes the angular acceleration of the exoskeleton knee joint, M is a short form of M(q), G is a short form of G(q), and C is a short form of ;

[0059] The state error of the exoskeleton is defined as: z1 = x1 - x d , z2 = x2 - a.

[0060] where x d denotes the set trajectory, and a is a virtual control variable. According to the two-degree-of-freedom exoskeleton robot, the following can be obtained

[0061]

[0062] The Lyapunov function V1 is designed as: The time derivative of V1 is:

[0063]

[0064] The following can be obtained:

[0065]

[0066] wherein, Design Lyapunov function The time derivative of V2 is:

[0067]

[0068] 3, the actual trajectory q of the exoskeleton output by the position controller is used as the DMP input trajectory for the next round of training;

[0069] The inner loop position controller uses a model-based backstepping controller to drive the exoskeleton model. The exoskeleton dynamics model parameters can be obtained by system identification methods, such as least squares method, particle swarm algorithm and intelligent optimization algorithm, etc.

[0070] If all the dynamic model parameters are known, the backstepping controller is designed as:

[0071]

[0072] It can be obtained that:

[0073]

[0074] wherein, K1, K2 represent positive gain matrices, and T represents the transpose of the matrix.

[0075] Therefore, the state errors z1 and z2 of the exoskeleton system tend to 0 when t→∞, that is, the system is globally asymptotically stable.

[0076] In this embodiment, the iterative control idea is to first establish a standard teaching gait trajectory library, select a relatively reasonable gait trajectory from the gait trajectory library as the initial training trajectory of DMP according to the body characteristic parameters of the wearer, use the output trajectory of DMP as the input trajectory of the admittance control loop, use the output trajectory of the admittance control loop as the reference trajectory, and use it as the tracking target of the inner loop controller. When designing the inner loop position controller, the establishment of the human-machine coupling model needs to be considered. The actual trajectory of the exoskeleton output by the position controller is used as the DMP input trajectory for the next round of training, and the process is repeated to complete the closed-loop active control of the exoskeleton. The control goal is to make the human-machine coupling as small as possible to complete the work task of the human-machine auxiliary.

[0077] Those skilled in the art will appreciate that the embodiments described herein are presented for purposes of illustration and understanding of the principles of the application and should not be construed as limiting the scope of the application to such specifically enumerated embodiments. Various modifications and changes can be made to the application by those skilled in the art which will be apparent from this disclosure without departing from the spirit and principles of the application. Any modifications, equivalent substitutions, improvements, etc. made to the application should be included within the scope of the application as defined in the following claims.

Claims

1. A three-loop control method for lower extremity exoskeleton based on dynamic motion primitives, characterized in that, The lower limb exoskeleton three-loop control system based on comprises: a gait trajectory library, an improved DMP algorithm module, an admittance controller, an exoskeleton model and a position controller. S1, first, according to the body characteristic parameters of the wearer, select a gait trajectory from the gait trajectory library as the initial training trajectory of the improved DMP algorithm module ; S2, the output trajectory of the improved DMP algorithm module is used as the input trajectory of the admittance controller; S3, the output trajectory of the admittance controller is used as the tracking target of the exoskeleton model; S4, the actual trajectory of the exoskeleton output by the position controller in the current exoskeleton model tracking process is used as the DMP input trajectory for the next round of training, and the closed-loop active control of the exoskeleton is completed.

2. The three-link control method of a lower extremity exoskeleton based on dynamic motion primitives according to claim 1, wherein, The expression of the improved DMP algorithm module in step S1 is: ； where y denotes the position, and are the velocity and acceleration, respectively, and g denotes the desired target of the system; and are gain coefficients; is the time scaling of the control system; denotes the initial state, is a constant, x comes from a first order system, and f is a radial basis function.

3. The three-link control method of a lower extremity exoskeleton based on dynamic motion primitives according to claim 2, wherein, By the given teaching trajectory , the target value of the radial basis function f is obtained , The expression is: ； wherein , , respectively represent the position, velocity, acceleration of the teaching trajectory.

4. The three-link control method of lower extremity exoskeleton based on dynamic motion primitives according to claim 3, wherein, The expression of the admittance controller is: ； wherein M, B, respectively represent virtual inertia, virtual damping and virtual stiffness parameters; represents the interaction force between human and robot; represents the trajectory deviation, , represents the derivative of first derivative, represents the second derivative of second derivative, represents the output trajectory of the admittance controller, represents the output trajectory of the improved DMP algorithm module.

5. The three-link control method of lower extremity exoskeleton based on dynamic motion primitives according to claim 4, wherein, The expression of the exoskeleton model is: ； wherein respectively represent the two joint angles, angular velocity and angular acceleration of the exoskeleton, represents the driving torque of the motor, , and respectively represent the inertia matrix, Coriolis matrix, gravity term of the system, represents the lumped term consisting of the unknown disturbance outside the system.

6. The three-link control method of a lower extremity exoskeleton based on dynamic motion primitives according to claim 5, wherein, In the 2-DOF lower limb exoskeleton, the following equations are used denotes the angle of the exoskeleton hip joint, denotes the angle of the exoskeleton knee joint, denotes the angular acceleration of the exoskeleton hip joint, denotes the angular acceleration of the exoskeleton knee joint, the state variables of the 2-DOF lower limb exoskeleton are , the superscript T denotes transposition; The state error of the exoskeleton is defined as: , ; wherein is represented as a set trajectory, is a virtual control variable; when and All When the time approaches 0, the three-ring control system of the 2-DOF lower limb exoskeleton is globally stable.

Images