CN117325169A - A control method for lower limb rehabilitation robot with initial state learning - Google Patents

Abstract

The invention discloses a control method for a lower limb rehabilitation robot with initial state learning, and relates to the technical field of auxiliary medical rehabilitation robot control. It includes: constructing an iterative learning controller of the lower limb rehabilitation robot; constructing an output error model; and collecting hip joints of the lower limb rehabilitation robot. and the discrete data information of the knee joint; construct the calculation equation of the iterative learning control law, and perform iterative learning, and obtain the final iterative learning control law as the first system input condition; construct the calculation equation of the initial deviation, and perform iterative learning to obtain the final The initial deviation is used as the input condition of the second system; the input condition of the first system and the input condition of the second system are input into the iterative learning controller, and the control strategy with initial state learning is output. The invention can well handle the problem of initial state deviation caused by different starting positions or repetitive training, correct the initial deviation, and enable the output trajectory of the lower limb rehabilitation robot system to track the desired trajectory even when there is an initial deviation.

Description

A control method for lower limb rehabilitation robot with initial state learning

Technical field

The invention relates to the technical field of auxiliary medical rehabilitation robot control, and specifically to a control method for a lower limb rehabilitation robot with initial state learning.

Background technique

At present, the number of patients with brain injuries and spinal cord injuries is increasing, but most patients still have walking dysfunction after saving their lives. For patients' gait rehabilitation, traditional treatment requires a medical staff to assist a patient in performing high-intensity and repetitive gait training. The rehabilitation effect can only be judged by the naked eye, and real-time feedback data cannot be obtained, and there is a serious shortage of rehabilitation therapists. As a product of the combination of medicine and engineering, rehabilitation robots can not only feed back gait data and quantify rehabilitation indicators in real time, but are also suitable for high-intensity and repetitive training work, freeing the hands of rehabilitation therapists. Lower limb rehabilitation robots assist patients in repetitive gait function training, stimulate and reshape the nervous system, restore patients to normal walking ability, and improve the walking function of patients with skeletal-related diseases such as stroke, spinal cord injury, brain trauma, and lower limb muscles.

In view of the repeatability, nonlinearity and inaccurate model of the lower limb rehabilitation robot system, the iterative learning control algorithm is widely used because it can learn through repetition, is suitable for nonlinearity and does not require accurate models, and is widely used. For example, the iterative learning control algorithm is used to establish the follow-up control model of the lower limb rehabilitation robot system and track the expected trajectories of the lower limb hip and knee joints to improve the follow-up performance of the human-machine system; iterative learning of the lower limb rehabilitation robot trajectory tracking in advance of the sampling time Control method; model-free adaptive iterative learning control algorithm; adaptive iterative learning control algorithm, etc.

However, the above control methods are all carried out under the condition that the initial state is consistent with the expected initial state, and the initial state deviation is not considered. If there is an error in the initial state, as the number of iterations increases, although the final tracking curve trend is consistent with the expected The target trajectory is roughly the same, but there is always a deviation from the expected target trajectory, and the control effect is poor. This situation can easily cause secondary harm to the patient.

Contents of the invention

The invention is an iterative learning control method for a lower limb rehabilitation robot with initial state learning, which solves the problems of initial angle deviation and low system convergence speed when patients use the lower limb rehabilitation robot system for training, and effectively realizes the lower limb rehabilitation robot system with In the case of initial errors, the desired motion trajectory can still be tracked quickly and accurately, improving the rehabilitation effect.

The technical solutions adopted by the present invention are as follows:

A control method for lower limb rehabilitation robots with initial state learning, including the following steps:

S1. Construct the dynamic equation of lower limb rehabilitation robot

Constructing an iterative learning controller for lower limb rehabilitation robots based on the dynamic equations of lower limb rehabilitation robots

in, is the derivative of x _k (t), x _k (t) represents the system state at the k-th iteration, u _k (t) represents the system input at the k-th iteration, y _k (t) represents the k-th iteration system output;

D(θ) is the inertia matrix; is the centrifugal force and Coriolis force matrix; G(θ) is the gravity matrix; T _θ represents the joint moment matrix; T _d is the error and disturbance;/> is the joint angle of the lower limb, θ ₁ and θ ₂ are the hip joint angle and knee joint angle respectively;/> is the joint angular velocity of the lower limbs,/> are the hip joint angular velocity and knee joint angular velocity respectively;/> is the joint angular acceleration of the lower limbs,/> are the hip joint angular acceleration and knee joint angular acceleration respectively;

S2. Construct an output error model based on the target trajectory of the lower limb rehabilitation robot and the control output of the lower limb rehabilitation robot;

S3. Collect discrete data information of the hip joint and knee joint of the lower limb rehabilitation robot, including lower limb joint angle data of the lower limb rehabilitation robot at different time points;

S4. According to the lower limb joint angle of the lower limb rehabilitation robot, the calculation equation of the iterative learning control law is obtained based on the iterative learning controller and the output error model. Based on the calculation equation of the iterative learning control law, iterative learning is performed. After the number of iterations is reached, the final iteration is obtained. Learning the control law as the first system input condition;

S5. According to the initial angle of the lower limb joint of the lower limb rehabilitation robot, the calculation equation of the initial deviation is obtained based on the iterative learning controller and the output error model. Based on the calculation equation of the initial deviation, iterative learning is performed. After the number of iterations is reached, the final initial deviation is obtained as Second system input conditions;

S6. Input the first system input condition and the second system input condition into the iterative learning controller, and output the control strategy with initial state learning that controls the operation of the lower limb rehabilitation robot.

In a preferred implementation of the present invention, the output error model is:

e _k (t)＝y _d (t)-y _k (t)

Among them, e _k (t) represents the tracking error of the lower limb rehabilitation robot in the k-th iteration, y _d (t) represents the target trajectory of the lower limb rehabilitation robot, and y _k (t) represents the control output of the k-th operation of the lower limb rehabilitation robot system. .

In a preferred embodiment of the present invention, in step S3, it is necessary to use piecewise cubic polynomials to fit the discrete data of the hip joint and knee joint of the lower limb rehabilitation robot to gait data to obtain continuous curve functions of the two joints.

In a preferred embodiment of the present invention, the calculation equation of the iterative learning control law obtained in step S4 is:

Among them, e _k+1 (t) and Respectively represent the tracking error and its error derivative at the k+1 iteration of the lower limb rehabilitation robot system, K and L are iterative learning gain matrices with corresponding dimensions, u _k+1 (t) is the k+1 iterative learning After the system input, u _k (t) is the system input during the k-th iterative learning. When k = 0, u _k (t) represents the system input during the initial non-iterative learning. The system input includes the hip joint and knee of the lower limbs. joint torque.

In a preferred embodiment of the present invention, step S4 specifically includes the following steps:

S41. Collect lower limb joint angle data during the k+1th run of the lower limb rehabilitation robot;

S42. Based on the target trajectory of the lower limb rehabilitation robot, the control output and output error model of the kth operation of the lower limb rehabilitation robot, calculate the tracking error e _k+1 (t) and its error in the k+1 iteration of the lower limb rehabilitation robot system. Derivative

S43. Set the iterative learning gain matrices K and L;

S44. Combine the gain matrices K and L, the tracking error e _k+1 (t) of the lower limb rehabilitation robot system at the k+1 iteration and its error derivative And the system input u _k (t) during the k-th iterative learning is substituted into the calculation equation of the first iterative learning control law to obtain the iterative learning control law;

S45. Loop steps S41 to S44 until the specified number of iterations is reached, that is, the final iterative learning control law is obtained. The final iterative learning control law is used as the first system input condition. During the loop process, each iteration uses The joint angles of the lower limbs are all different.

In a preferred embodiment of the present invention, the calculation equation of the initial deviation obtained in step S5 is:

x _k+1 (0)＝x _k (0)+M(0)Le _k+1 (0)

Among them, x _k+1 (0) is the k+1 iteration initial state of the system, x _k (0) is the k-th iteration initial state of the system, and M (0) is when t=0 The obtained value, L is the iterative learning gain matrix of the system, and e _k+1 (0) is the initial value of the k+1 iteration tracking error of the system.

In a preferred embodiment of the present invention, step S5 specifically includes the following steps:

S51. Collect the iteration initial state angle during the k+1th run of the lower limb rehabilitation robot;

S52. Based on the target trajectory of the lower limb rehabilitation robot, the control output and output error model of the kth operation of the lower limb rehabilitation robot, calculate the initial value of the k+1 iteration tracking error of the lower limb rehabilitation robot system;

S53. Set the iterative learning gain matrix L;

S54. Substitute the gain matrix L, the initial value of the k+1 iteration tracking error of the lower limb rehabilitation robot system, and the iterative initial state angle at the k+1th run into the calculation equation of the initial deviation, and output the result of the lower limb rehabilitation robot system. The initial deviation obtained at the k+1 iteration;

S55. Loop steps S51 to S54 until the specified number of iterations is reached. The final initial deviation is obtained and used as the input condition of the second system. The initial angles of the lower limb joints used in each iteration are different.

Compared with the prior art, the beneficial effects of the present invention are:

It is an iterative learning control method with initial state learning, which can well handle the problem of initial state deviation caused by different starting positions or repetitive training, correct the initial deviation, and make the lower limb rehabilitation robot system when there is an initial state deviation. The output trajectory can also track the desired trajectory, reducing secondary damage caused by initial value deviation and improving the user experience.

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, embodiments of the present invention are given below and described in detail with reference to the accompanying drawings.

Description of drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and therefore do not It should be regarded as a limitation of the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without exerting creative efforts.

Figure 1 is a model diagram of the lower limb rehabilitation robot system used in the present invention;

Figure 2 is a control block diagram of a lower limb rehabilitation robot with initial state learning according to the present invention;

Figure 3 is a diagram showing changes in the tracking error of the hip joint and knee joint angles of the lower limbs using iterative learning control with initial state learning in the present invention;

Figure 4 is the angular trajectory tracking diagram and angle error curve diagram of the hip joint and knee joint of the lower limb after 20 iterations in the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, rather than all embodiments.

Please refer to Figures 1 and 2. The present invention provides a lower limb rehabilitation robot control method with initial state learning, which specifically includes the following steps:

Step 1: Collect the discrete data information of the hip joint O and knee joint P of the human lower limbs, use piecewise cubic polynomials to fit the data, and obtain the curve functions of the two joints, namely formula (1) and formula (2), and calculate The fitted curve function is subjected to error analysis. If the error is too large, data fitting is performed again. The traditional fitting method is a trigonometric function polynomial, and the fitting is rough. The present invention uses piecewise polynomials for fitting, and the error lower than before. Through error analysis, it is ensured that the fitting functions of the hip joint and knee joint are suitable for the trajectory tracking target curve of the lower limb rehabilitation robot system.

Among them, t represents time, the joint angle of the lower limb includes the hip joint angle and the knee joint angle, θ ₁ represents the hip joint angle, and θ ₂ represents the knee joint angle.

Step 2: Establish a lower limb rehabilitation robot system model, as shown in Figure 1. O and P represent the hip joint and knee joint respectively. It is stipulated that horizontally to the right is the positive direction of the x-axis; vertically downward is the positive direction of the y-axis; the intersection of the two axes O represents the coordinate origin, that is, the hip joint; P represents the knee joint; l ₁ and l ₂ represent the lengths of the thigh link and calf link respectively; A and B are the positions of the center of mass of the two links respectively; d ₁ and d ₂ are respectively The length from the center of mass of the connecting rod to the hip joint and knee joint; θ ₁ and θ ₂ are the angles of the hip joint and knee joint respectively; according to the forward kinematics and inverse kinematics, the dynamic equation of the lower limb rehabilitation robot system model is obtained as Equation (3) is shown, and the system state equation of the lower limb rehabilitation robot system model is derived as shown in Equation (4).

in,

D(θ) is the inertia matrix; is the centrifugal force and Coriolis force matrix; G(θ) is the gravity matrix; u(t) represents the joint moment matrix; T _d is the error and disturbance;/> is the joint angle of the lower limb, including the hip joint angle θ ₁ and the knee joint angle θ ₂ ;/> is the joint angular velocity of the lower limbs, including the hip joint angular velocity/> and knee joint angular velocity/> is the joint angular acceleration of the lower limbs, including the hip joint angular acceleration/> and knee joint angular acceleration/>

Step 3: Design an iterative learning controller for the problem that the lower limb rehabilitation robot system cannot track the expected trajectory due to the initial state error. It aims to realize that the actual output trajectory of the lower limb rehabilitation robot system can still be fast even if there is an initial state deviation. Accurately track the upper target curve. The iterative learning controller is obtained according to equation (4), as shown in equation (5).

Among them, the subscript k represents the k-th iteration learning control of the lower limb rehabilitation robot system, is the derivative of x _k (t), x _k (t) represents the system state at the k-th iteration, that is, the angle and angular velocity of the hip joint and knee joint, u _k (t) represents the system input at the k-th iteration, y _k (t) represents the system output at the k-th iteration.

Step 4: Construct the output error model as shown in Equation (6):

e _k (t)＝y _d (t)-y _k (t) (6)

Among them, e _k (t) represents the tracking error of the lower limb rehabilitation robot system in the k-th iteration, Represents the target trajectory of the lower limb rehabilitation robot system. The target trajectory curve here has been given in step 1, that is, formula (1) and formula (2). y _k (t) represents the control of the lower limb rehabilitation robot system during the kth run. The output, both given in steps 2 and 3.

Step 5: Design the first iterative learning control law as system input:

Among them, e _k+1 (t) and Respectively represent the tracking error and its error derivative at the k+1 iteration of the lower limb rehabilitation robot system, K and L are iterative learning gain matrices with corresponding dimensions, u _k+1 (t) is the k+1 iterative learning After the system input, u _k (t) is the system input during the k-th iterative learning. When k = 0, u _k (t) represents the system input during the initial non-iterative learning. The system input includes the hip joint and knee of the lower limbs. joint torque.

The following uses an iterative learning control law for control to ensure that the lower limb rehabilitation robot system can track the target trajectory. The specific process is as follows:

Step 5.1: Collect the lower limb joint angles θ _1(k+1) and θ _2(k+1) during the k+1th operation of the lower limb rehabilitation robot;

Step 5.2: According to equation (6), derive e _k+1 (t) = y _d (t)-y _k+1 (t), where As given in step 1, calculate the tracking error e _k+1 (t) and its error derivative at the k+1 iteration of the lower limb rehabilitation robot system/>

Step 5.3: Set the iterative learning gain matrices K and L;

Step 5.4: Combine the gain matrices K and L in step 5.3, the tracking error e _k+1 (t) of the lower limb rehabilitation robot system at the k+1 iteration in step 5.2 and its error derivative And the system input u _k (t) during the k-th iterative learning is substituted into equation (7) to obtain the iterative learning control law;

Step 5.5: Loop the above steps until the specified number of iterations is reached, that is, the final iterative learning control law u _k+1 (t) is obtained, and the final iterative learning control law u _k+1 (t) is used as the first system input Condition, during the loop process, the lower limb joint angles used in each iteration are different.

Step 6: While running step 5, the initial state of each iteration of the lower limb rehabilitation robot system (i.e. t=0) is also learned using the initial deviation to ensure that the lower limb rehabilitation robot system can still track even if the initial state shifts. upper target trajectory curve. The calculation equation to construct the initial deviation is:

x _k+1 (0)＝x _k (0)+M(0)Le _k+1 (0) (8)

Among them, x _k+1 (0) is the k+1 iteration initial state of the system, x _k (0) is the k-th iteration initial state of the system, and M (0) is when t=0 The obtained value, L is the iterative learning gain matrix of the system, and e _k+1 (0) is the initial value of the k+1 iteration tracking error of the system.

The specific iteration process is as follows:

Step 6.1: Collect the iterative initial state angles θ _1(k+1)(0) and θ _2(k+1) (0) during the k+1th run of the lower limb rehabilitation robot;

Step 6.2: According to equation (6), derive e _k+1 (0)=y _d (0)-y _k+1 (0), where y _d (t) has been given in step 1. Calculate the initial value of the k+1 iteration tracking error e _k+1 (0) of the lower limb rehabilitation robot system;

Step 6.3: Set the iterative initial state learning gain matrix L;

Step 6.4: Combine M(0), the gain matrix L in step 6.3, the initial tracking error value e _k+1 (0) of the k+1 iteration of the lower limb rehabilitation robot system in step 6.2, and the k-th iteration of the system. The initial value x _k (0) is substituted into equation (8) to obtain the initial deviation x _k+1 (0) obtained at the k+1 iteration of the system;

Step 6.5: Loop the above steps until the specified number of iterations is reached. The initial deviation x _k+1 (0) obtained by the final iteration is obtained, and is used as the input condition of the second system. The initial state of the lower limb joints used in each iteration is The angles are all different.

Step 7: Combine the input conditions of the first system (the system input u _k+1 (t) after the k+1th iteration learning) and the input conditions of the second system (the initial state of the k+1 iteration x _k+1 ( 0)) Input the lower limb rehabilitation robot, that is, input formula (5), as shown in Figure 2, and obtain the iterative learning control output y _k+1 (t) of the k+1th operation of the lower limb rehabilitation robot system, that is, control the lower limb rehabilitation robot The running control strategy with initial state learning includes the lower limb hip joint angle θ ₁ and knee joint angle θ ₂ . In Figure 2, e _k+1 (t) represents the tracking error at the k+1 iteration of the lower limb rehabilitation robot system. ;K and L are iterative learning gain matrices with corresponding dimensions; Indicates the derivation of e _k+1 (t), that is, the tracking error derivative e k ₊₁ (t) at the k+1 iteration of the lower limb rehabilitation robot system; u _k (t) is the tracking error at the k-th iteration of learning. System input; u _k+1 (t) is the system input after the k+1 iteration learning; x _k+1 (0) is the k+1 iteration initial state of the system; x _k (0) is the system’s k+1 iteration initial state Initial state of k iterations; M(0) is when t=0 The obtained value; e _k+1 (0) is the initial value of the k+1 iteration tracking error of the system; y _d (t) represents the target trajectory of the lower limb rehabilitation robot system; y _k+1 (t) represents the lower limb rehabilitation robot The control output when the system runs for the k+1th time.

Next, a simulation experiment is conducted on the control method of the lower limb rehabilitation robot with initial state learning according to the present invention.

Simulate and analyze the lower limb rehabilitation robot system in the initial offset state to verify the effectiveness of the improved algorithm. Figure 3 is a graph showing changes in joint angle tracking errors of the lower limb rehabilitation robot under iterative learning control with initial state learning. The star line represents the angle tracking error curve of the hip joint, and the circle line represents the angle tracking error curve of the knee joint. Judging from the curve trend in the figure, the hip joint and knee joint angle tracking errors gradually decrease as the number of iterations increases, indicating that the improved algorithm can effectively suppress the impact of the initial state error. At the 8th iteration control point, the angle tracking The error is 0.2104°, and then the system output error gradually approaches zero. It is verified that the iterative learning control strategy with initial state learning can overcome the impact of initial state error.

Figure 4 shows the trajectory tracking effect of the lower limb rehabilitation robot system after 20 iterations of learning control. (a) shows the hip joint angle tracking chart; (b) shows the hip joint angle error curve; (c) shows the knee joint angle. Tracking graph; (d) graph shows the knee joint angle error curve. In the figures (a) and (c), the dotted lines represent the target curves of the hip and knee joints; the solid lines represent the actual output curves of the hip and knee joints of the lower limb rehabilitation robot system. It can be seen from these four pictures that the angles of the hip joint and the knee joint have tracked the target trajectory curve of the upper system, and the maximum angle error does not exceed 0.3°. The results verify that the iterative learning control strategy with initial state learning can overcome the impact of initial state deviation, effectively realizing that the lower limb rehabilitation robot system can still quickly and accurately track the desired motion trajectory despite initial errors, improving the rehabilitation effect. .

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (7)

1. A control method for a lower limb rehabilitation robot with initial state learning, which is characterized by including the following steps: S1. Construct the dynamic equation of lower limb rehabilitation robot Constructing an iterative learning controller for lower limb rehabilitation robots based on the dynamic equations of lower limb rehabilitation robots in, is the derivative of x _k (t), x _k (t) represents the system state at the k-th iteration, u _k (t) represents the system input at the k-th iteration, y _k (t) represents the k-th iteration system output; D(θ) is the inertia matrix; is the centrifugal force and Coriolis force matrix; G(θ) is the gravity matrix; Tθ represents the joint moment matrix; T _d is the error and disturbance;/> is the joint angle of the lower limb, θ ₁ and θ ₂ are the hip joint angle and knee joint angle respectively; is the joint angular velocity of the lower limbs,/> are the hip joint angular velocity and knee joint angular velocity respectively;/> is the joint angular acceleration of the lower limbs,/> are the hip joint angular acceleration and knee joint angular acceleration respectively; S2. Construct an output error model based on the target trajectory of the lower limb rehabilitation robot and the control output of the lower limb rehabilitation robot; S3. Collect discrete data information of the hip joint and knee joint of the lower limb rehabilitation robot, including lower limb joint angle data of the lower limb rehabilitation robot at different time points; S4. According to the lower limb joint angle of the lower limb rehabilitation robot, the calculation equation of the iterative learning control law is obtained based on the iterative learning controller and the output error model. Based on the calculation equation of the iterative learning control law, iterative learning is performed. After the number of iterations is reached, the final iteration is obtained. Learning the control law as the first system input condition; S5. According to the initial angle of the lower limb joint of the lower limb rehabilitation robot, the calculation equation of the initial deviation is obtained based on the iterative learning controller and the output error model. Based on the calculation equation of the initial deviation, iterative learning is performed. After the number of iterations is reached, the final initial deviation is obtained as Second system input conditions; S6. Input the first system input condition and the second system input condition into the iterative learning controller, and output the control strategy with initial state learning that controls the operation of the lower limb rehabilitation robot. 2. The lower limb rehabilitation robot control method with initial state learning according to claim 1, characterized in that the output error model is: e _k (t)＝y _d (t)-y _k (t) Among them, e _k (t) represents the tracking error of the lower limb rehabilitation robot in the k-th iteration, y _d (t) represents the target trajectory of the lower limb rehabilitation robot, and y _k (t) represents the control output of the k-th operation of the lower limb rehabilitation robot system. . 3. The lower limb rehabilitation robot control method with initial state learning according to claim 2, characterized in that, in step S3, it is necessary to use a piecewise cubic polynomial to perform gait on the discrete data information of the hip joint and knee joint of the lower limb rehabilitation robot. The data is fitted to obtain the continuous curve function of the two joints. 4. The lower limb rehabilitation robot control method with initial state learning according to claim 2 or 3, characterized in that the calculation equation of the iterative learning control law obtained in step S4 is: Among them, e _k+1 (t) and Respectively represent the tracking error and its error derivative at the k+1 iteration of the lower limb rehabilitation robot system, K and L are iterative learning gain matrices with corresponding dimensions, u _k+1 (t) is the k+1 iterative learning After the system input, u _k (t) is the system input during the k-th iterative learning. When k = 0, u _k (t) represents the system input during the initial non-iterative learning. The system input includes the hip joint and knee of the lower limbs. joint torque. 5. The lower limb rehabilitation robot control method with initial state learning according to claim 4, characterized in that step S4 specifically includes the following steps: S41. Collect lower limb joint angle data during the k+1th run of the lower limb rehabilitation robot; S42. Based on the target trajectory of the lower limb rehabilitation robot, the control output and output error model of the kth operation of the lower limb rehabilitation robot, calculate the tracking error e _k+1 (t) and its error in the k+1 iteration of the lower limb rehabilitation robot system. Derivative S43. Set the iterative learning gain matrices K and L; S44. Combine the gain matrices K and L, the tracking error e _k+1 (t) of the lower limb rehabilitation robot system at the k+1 iteration and its error derivative And the system input u _k (t) during the k-th iterative learning is substituted into the calculation equation of the first iterative learning control law to obtain the iterative learning control law; S45. Loop steps S41 to S44 until the specified number of iterations is reached, that is, the final iterative learning control law is obtained. The final iterative learning control law is used as the first system input condition. During the loop process, each iteration uses The joint angles of the lower limbs are all different. 6. The lower limb rehabilitation robot control method with initial state learning according to claim 2 or 3, characterized in that the calculation equation of the initial deviation obtained in step S5 is: x _k+1 (0)＝x _k (0)+M(0)Le _k+1 (0) Among them, x _k+1 (0) is the k+1 iteration initial state of the system, x _k (0) is the k-th iteration initial state of the system, and M (0) is when t=0 The obtained value, L is the iterative learning gain matrix of the system, and e _k+1 (0) is the initial value of the k+1 iteration tracking error of the system. 7. The lower limb rehabilitation robot control method with initial state learning according to claim 6, characterized in that step S5 specifically includes the following steps: S51. Collect the iteration initial state angle during the k+1th run of the lower limb rehabilitation robot; S52. Based on the target trajectory of the lower limb rehabilitation robot, the control output and output error model of the kth operation of the lower limb rehabilitation robot, calculate the initial value of the k+1 iteration tracking error of the lower limb rehabilitation robot system; S53. Set the iterative learning gain matrix L; S54. Substitute the gain matrix L, the initial value of the k+1 iteration tracking error of the lower limb rehabilitation robot system, and the iterative initial state angle at the k+1th run into the calculation equation of the initial deviation, and output the result of the lower limb rehabilitation robot system. The initial deviation obtained at the k+1 iteration; S55. Loop steps S51 to S54 until the specified number of iterations is reached. The final initial deviation is obtained and used as the input condition of the second system. The initial angles of the lower limb joints used in each iteration are different.