A multi-mode hybrid control method based on lower limb exoskeleton robot

Abstract

The present invention discloses a multi-mode hybrid control method based on a lower limb exoskeleton robot, which belongs to the field of robot control. Specifically, the method comprises the following steps: first, for a patient, the motion trajectory of the patient's ankle joint in three-dimensional space is used as the desired motion trajectory; then, the robot's angle sensor is used to measure the robot's motion angle, and the actual position of the robot's ankle joint is obtained through kinematic calculation; the trajectory tracking error and the contour tracking error are calculated in the passive mode and the active mode respectively; finally, based on the dynamic model of the lower limb exoskeleton robot, the trajectory tracking error and the contour tracking error are used to calculate the torque values ​​of the controllers in the active control mode and the passive control mode respectively, so as to realize motion in different control modes. The two controllers of the present invention adopt a control architecture with exactly the same structure, but the error expression method is different. Therefore, different errors can be input based on different rehabilitation training needs to realize the switching of training modes.

Description

Technical Field

[0001] The present invention belongs to the field of robot control, and in particular relates to a multi-mode hybrid control method based on a lower limb exoskeleton robot. Background Art

[0002] The increasing aging of society and the high incidence of conditions like stroke have led to a yearly increase in the number of patients suffering from lower limb motor dysfunction. Lower limb exoskeleton robots, as a new type of rehabilitation training product, are gradually being applied in the medical rehabilitation field. Compared to traditional methods that rely on the individual experience and skills of rehabilitation therapists and are labor-intensive, rehabilitation robots provide patients with more efficient, targeted, and repeatable training guidance, as well as monitoring and evaluation capabilities.

[0003] Patients with stroke, joint injuries, and spinal cord injuries require targeted rehabilitation training tailored to their condition. As rehabilitation progresses and motor skills gradually recover, the program needs to be adjusted to suit each stage of the patient's condition. Therefore, it's necessary to study control strategies for exoskeleton robots tailored to each stage of the condition, with multimodal control being one of the best options.

[0004] Robot-assisted rehabilitation training is divided into two modes: passive and active. The passive mode is a trajectory tracking control mode, in which the patient's limbs are completely driven by the robot to complete lower limb rehabilitation training. In the active mode, the robot can interact with the patient based on the patient's movement intentions and provide necessary assistance. Passive control is suitable for the early stages of rehabilitation training, while active control is suitable for the middle and late stages of rehabilitation training. Therefore, control methods need to be designed for different control modes. For exoskeleton robots, not only do they need to have multiple different control modes, but they also need to switch between different modes according to the stage of the disease.

[0005] Existing control methods for lower-limb exoskeletons can achieve both passive and active control modes, but these modes are often independent of each other, meaning the system is designed for either passive or active control. There are also multi-mode control approaches that combine both passive and active control, but the control frameworks for these two modes are not analogous, making it impossible to switch between them. Summary of the Invention

[0006] In response to the problems existing in the on-demand gait assistance of lower limb exoskeleton robots, the present invention provides a multi-mode hybrid control method based on lower limb exoskeleton robots. First, according to the characteristics of different control modes, a representation method of trajectory error in passive mode and contour error in active mode is proposed. Then, based on the dynamic model of the lower limb exoskeleton robot, a controller with dynamic compensation is designed to realize motion control under different control modes.

[0007] The on-demand auxiliary control method based on the lower limb exoskeleton robot has the following specific steps:

[0008] Step 1: For the patient, the motion trajectory of the ankle joint in three-dimensional space is used as the expected motion trajectory;

[0009] Right now

[0010] Where s∈[0,100] represents the percentage of the current movement time t relative to the gait period T, represents a three-dimensional Euclidean space.

[0011] Step 2: Use the angle sensor on the robot to measure the robot's movement angle, and obtain the actual position P of the robot's ankle joint through kinematic calculation. a (t).

[0012] Step 3: In passive mode, the robot drives the patient's limbs to move, and uses the patient's expected motion trajectory to calculate the trajectory tracking error e p1 .

[0013] Trajectory tracking error e p1 Expressed as:

[0014] e p1 =P a -f(s)

[0015] Step 4: In active mode, calculate the point f(s) closest to the patient's current actual position on the patient's expected motion trajectory f(s). * ) to the actual position P a (t), that is, the contour tracking error e p2 ;

[0016] Right now

[0017] e p2 =P a -f(s * )

[0018] Use the following controller to calculate the point f(s) closest to the current actual position * ):

[0019]

[0020] k and λ are positive constants, k Ψ is a function of s, ∑ is the sliding surface function; α is the order, α≥1, and Ψ is a variable describing the distance projection.

[0021] Step 5: Based on the dynamic model of the lower limb exoskeleton robot, use the trajectory tracking error e p1 and contour tracking error ep2 Design a controller with dynamic model and velocity error estimation to realize motion control under different control modes.

[0022] The dynamic equation of the robot is:

[0023]

[0024] Where M is the inertia matrix; C is the Coriolis force and centripetal force; G is the gravitational term; F is the friction force; τ ext is the interaction force between the human and the robot, that is, the force applied by the patient to the robot; q = [θ h θ k ] T is a generalized variable, where θ h and θ k are the angles of the hip joint and knee joint respectively, and τ is the control law, which is designed as:

[0025]

[0026] J is Jacobi for robots, K d is the speed gain, is the estimated value of the velocity error, is the estimated value of the kinetic model, F a is the torque term, in passive control mode:

[0027] F a =-K p e p

[0028] K p is the position gain, e p is the error, and at this time e p =e p1 ;

[0029] In active control mode: F a =k1ω1F ac +k2ω2F tr

[0030]

[0031] k1 and k2 are control gains used to adjust the amplitude of the output torque; ω1 and ω2 are the weights of the tangential component and the normal component respectively; r is the value of the attitude profile error e p A variable to adjust the relative weights of the two components;

[0032] F ac and F tr is the unit vector that applies the adjustment torque direction at the nearest point, and the calculation formula is:

[0033] F ac =-(n+b) / ||n+b||=-e p / ||e p ||

[0034] F tr =t

[0035] n and b represent the normal vector and binormal vector at the nearest position, respectively. p is the error, at this time e p =e p2 ; t is the tangent vector at the nearest position.

[0036] The advantages of the present invention are:

[0037] 1) A multi-mode hybrid control method for a lower-limb exoskeleton robot is designed to meet the training needs of different rehabilitation stages. A passive controller based on tracking error and an active controller based on contour error are designed. The two controllers use the same control architecture, but the error is expressed differently. Therefore, different errors can be input based on different rehabilitation training needs to achieve switching between training modes. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of a multi-mode hybrid control method based on a lower limb exoskeleton robot according to the present invention;

[0039] Figure 2 Schematic diagram of a parameterized curve C in three-dimensional Euclidean space of the present invention;

[0040] Figure 3 This invention proves that f(s * ) is a flow chart of the nearest point on the desired motion trajectory f(s);

[0041] Figure 4 This is the overall block diagram of the control system of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be further described below with reference to the accompanying drawings and examples.

[0043] The on-demand auxiliary control method based on the lower limb exoskeleton robot is as follows: Figure 1 The specific steps are as follows:

[0044] Step 1: For the patient, the motion trajectory of the lower limb end point, i.e., the ankle joint in three-dimensional space is used as the expected motion trajectory;

[0045] Right now:

[0046]

[0047] Where s∈[0,100] represents the percentage of the current movement time t relative to the gait period T; represents a three-dimensional Euclidean space.

[0048] Step 2: Use the angle sensor on the robot to measure the robot's movement angle, and obtain the actual end point of the robot, that is, the actual position P of the ankle joint through kinematic calculation. a (t).

[0049] Step 3: In passive mode, the robot drives the patient's limbs to move, and uses the patient's expected motion trajectory to calculate the trajectory tracking error e p1 .

[0050] What is controlled at this time is the trajectory tracking error, and the expected motion point is

[0051] P d1 (t) = f(s) (2)

[0052] At this time, s is

[0053]

[0054] Is a quantity related to the current running time t. The trajectory tracking error is expressed as

[0055] Trajectory tracking error e p1 Expressed as:

[0056] e p1 =P a -P d1 =P a -f(s) (4)

[0057] P d1 (t) is the expected movement point when the robot drives the patient's limbs to move;

[0058] Step 4: In active mode, calculate the point f(s) closest to the patient's current actual position on the patient's expected motion trajectory f(s). * ) to the actual position P a (t), that is, the contour tracking error e p2 ;

[0059] Right now

[0060] e p2 =P a -P d2 =P a -f(s * ) (5)

[0061] P d2 The point on the desired motion trajectory closest to the current actual position is the distance from this point to the actual position Pa (t) distance;

[0062] In order to solve the nearest point f(s) on the desired motion trajectory f(s) * ), using the following method:

[0063] If the mapping Defined three-dimensional Euclidean space A parameterized curve C on the actual position and The controller then calculates as follows:

[0064]

[0065] The point f(s) closest to the actual position on the entire curve C can be found * ),like Figure 2 As shown; where k and λ are positive constants, k Ψ is a function of s, ∑ is a sliding surface function, and its form will be shown in the proof process. The proof process is as follows:

[0066] For the Frenet frame at s∈[0,l], it is {f(s); t(s), n(s), b(s)} that satisfies

[0067]

[0068] in

[0069] t(s)=f s (s) / ||f s (s)|| (8)

[0070] definition:

[0071]

[0072] From the definition, we know that Ψ describes the s The projection length, when at the nearest point s * Satisfaction

[0073]

[0074] At the same time, we get

[0075]

[0076] Using (7) and (11) we can get

[0077]

[0078] definition but Define a reaching law as

[0079]

[0080] Substituting into (6) we get

[0081]

[0082] Select the Lyapunov function as

[0083]

[0084] thereby

[0085]

[0086] The calculation process of the algorithm is as follows Figure 3 shown.

[0087] Step 5: Based on the dynamic model of the lower limb exoskeleton robot, use the trajectory tracking error e p1 and contour tracking error e p2 Design a controller with dynamic model and velocity error estimation to realize motion control under different control modes.

[0088] The dynamic equation of the robot is:

[0089]

[0090] Where M is the inertia matrix, C is the Coriolis force and centripetal force, G is the gravitational term, F is the friction force, τ ext is the interaction force between the human and the robot, that is, the force applied by the patient to the robot; q = [θ h θ k ] T is a generalized variable, where θ h and θ k are the angles of the hip and knee joints, respectively, and τ is the control law;

[0091] Need to meet:

[0092] Property 1: M is a positive definite symmetric matrix;

[0093] Property 2: M and C satisfy:

[0094]

[0095] Define e = qq d ,therefore That is, the Jacobi relationship between the two is satisfied, then

[0096]

[0097] in Dynamics and friction are difficult to model accurately.

[0098] In practical systems, for the first-order derivative, whether still It is difficult to measure directly, and It is also difficult to find the solution, but is bounded, exists Therefore, a first-order filter is used for estimation, that is,

[0099]

[0100]

[0101] Introducing a measurable auxiliary signal s

[0102]

[0103] thereby

[0104]

[0105] The control law is designed as

[0106]

[0107] J is Jacobi for robots, K d is the speed gain, is the estimated value of the velocity error, is the estimated value of the kinetic model;

[0108] in As an estimate of D, The estimation error is bounded and exists Using RBF neural network approximation processing, we can get

[0109]

[0110] definition is the estimated value, and the estimated optimal value is recorded as W * , and satisfies Selected from the robot's dynamic equations The update law is

[0111]

[0112] The weight estimation error is Defining filtering error According to the filter definition, we can get

[0113]

[0114]

[0115] According to different control modes, the torque term F in (24) a They are:

[0116] Passive control mode

[0117] F a =-K p e p (29)

[0118] K p is the position gain, e p is the error, and at this time e p =e p1

[0119] In active control mode, based on the closest point, the unit vectors of the two directions of applying the adjustment torque are found as follows:

[0120] F ac =-(n+b) / ||n+b||=-e p / ||e p || (30)

[0121] F tr =t (31)

[0122] n and b represent the normal vector and binormal vector at the nearest position, respectively. p is the error, at this time e p =e p2 , t is the tangent vector at the nearest position.

[0123] Thus, the force field is established as

[0124] F a =k1ω1F ac +k2ω2F tr (32)

[0125]

[0126] k1 and k2 are control gains used to adjust the amplitude of the output torque; ω1 and ω2 are the weights of the tangential component and the normal component respectively; r is the value of the attitude profile error e p A variable to adjust the relative weights of the two components;

[0127] Select the Lyapunov function as

[0128]

[0129] Then its first-order derivative is

[0130]

[0131] Substituting into the control law and using Property 2 we get

[0132]

[0133] because and Then there is

[0134]

[0135] at the same time

[0136]

[0137]

[0138]

[0139]

[0140] In the passive control mode, substituting (29) into (35) yields

[0141]

[0142] in Pick:

[0143]

[0144] Then there is

[0145]

[0146] By adjusting the parameters to ensure that λ1>0, the designed controller is stable and the system is robust.

[0147] In the active control mode, we get

[0148]

[0149] Substituting (45) into (35) yields

[0150]

[0151] in Also because

[0152]

[0153] Substituting into (46) we get

[0154]

[0155] Due to the error ep is bounded, that is, it satisfies ||e p ||≤e p , so K p Needs to be satisfied Pick

[0156]

[0157] Then there is

[0158]

[0159] By adjusting the parameters to ensure that λ2>0, the designed controller is stable, so that the system has robustness; the overall block diagram of the control system is as follows Figure 4 shown.

Claims

1. A multi-mode hybrid control method based on a lower limb exoskeleton robot, characterized in that: The specific steps are as follows: For the patient, the motion trajectory of the ankle joint in three-dimensional space is taken as the expected motion trajectory f(s); Then, the angle sensor on the robot is used to measure the movement angle of the robot, and the actual position P of the robot's ankle joint is obtained through kinematic calculation. a (t); Calculate the patient's limb movement in the passive mode and use the patient's expected motion trajectory to calculate the trajectory tracking error e p1 , and the point f(s) closest to the current actual position on the patient's expected motion trajectory f(s) in active mode * ) to the actual position P a (t), that is, the contour tracking error e p2 ; The trajectory tracking error e p1 Expressed as: e p1 =P a -f(s); Contour tracking error e p2 Expressed as: e p2 =P a -f(s * ); Among them, the point f(s) closest to the current actual position * ) is calculated as: k and λ are positive constants, k Ψ is a function of s, Σ is the sliding surface function; α is the order, α≥1, Ψ is the variable describing the distance projection; Finally, based on the dynamic model of the lower limb exoskeleton robot, the trajectory tracking error e p1 and contour tracking error e p2 Design a controller with dynamic model and velocity error estimation to realize motion control under different control modes; The dynamic equation of the robot is: Where M is the inertia matrix; C is the Coriolis force and centripetal force; G is the gravitational term; F is the friction force; τ ext is the interaction force between the human and the robot, that is, the force applied by the patient to the robot; q = [θ h θ k ] T is a generalized variable, where θ h and θ k are the angles of the hip joint and knee joint respectively, and τ is the control law, which is designed as: J is Jacobi for robots, K d is the speed gain, is the estimated value of the velocity error, is the estimated value of the kinetic model, F a is the torque term, in passive control mode: F a =-K p e p K p is the position gain, e p is the error, and at this time e p =e p1 ; In active control mode: F a =k1ω1F ac +k2ω2F tr k1 and k2 are control gains used to adjust the amplitude of the output torque; ω1 and ω2 are the weights of the tangential component and the normal component respectively; r is the value of the attitude profile error e p A variable to adjust the relative weights of the two components; F ac and F tr is the unit vector that applies the adjustment torque direction at the nearest point, and the calculation formula is: F ac =-(n+b) / ||n+b||=-e p / ||e p || F tr =t n and b represent the normal vector and binormal vector at the nearest position, respectively. p is the error, at this time e p =e p2 ; t is the tangent vector at the nearest position.

2. A multi-mode hybrid control method based on a lower limb exoskeleton robot as claimed in claim 1, characterized in that: The desired motion trajectory, i.e. l is the end point of the interval; Where s∈[0,100] represents the percentage of the current movement time t relative to the gait period T, represents a three-dimensional Euclidean space.

Images