Tracking control method for cyclic switching of passive and active training modes of rehabilitation robot

Abstract

This invention discloses a tracking control method for the cyclical switching of active and passive training modes in a rehabilitation robot. Its features include: constructing a switching dynamics model for the rehabilitation robot based on the different interference environments generated on the human-machine system during active and passive training by the patient; designing a random configuration network with switching characteristics to estimate the interference environment; establishing a tracking error system for active and passive training, enabling the human-machine system to achieve cyclical switching between active and passive training while meeting the condition of minimum average training time; and using an STM32F411 series microcontroller to output PWM signals to the motor drive module, allowing the robot to assist the patient in performing the doctor-specified active-passive cyclical training. This invention not only provides an adaptive walking movement mode for patients to enter comprehensive active training but also avoids the safety impact of frequent manual switching between active and passive training modes.

Description

Technical Field

[0001] This invention relates to the control of wheeled rehabilitation robots, and more particularly to a tracking control method for the cyclic switching of active and passive training modes of a rehabilitation robot. Background Technology

[0002] Traffic accidents and an aging population have led to a year-on-year increase in patients with gait disorders. Due to a shortage of professional rehabilitation personnel, these patients cannot receive timely and effective exercise training, resulting in a gradual loss of walking function and an inability to live independently. The application of rehabilitation walking robots in rehabilitation centers and nursing homes has effectively addressed the shortage of rehabilitation personnel. In practical applications, robot-assisted rehabilitation using a combination of active and passive training modes can significantly improve training effectiveness. Existing rehabilitation robots require shutdown before selecting between active and passive training, which can easily lead to vibrations, collisions, and other hazards due to inertia, seriously affecting the safety of the human-machine system. Therefore, researching control methods for the cyclical switching of active and passive training modes in rehabilitation robots is of great significance for improving the training effect for rehabilitation patients and the safety of the human-machine system.

[0003] In recent years, digital and intelligent rehabilitation robots have been helping people with lower limb motor dysfunction regain their ability to walk independently. Rehabilitation physicians need to develop scientific and differentiated walking training programs for patients to improve the effectiveness of rehabilitation training. To date, there have been many research results on the working modes and control methods of rehabilitation robots; however, the mode and control method for the cyclical switching of active and passive training in rehabilitation robots have not been addressed. This invention proposes a new mode for the cyclical switching of active and passive training in rehabilitation robots from a novel perspective. This mode allows for switching between active and passive training without stopping the robot, preventing the patient's forearm from detaching from the handrail and potentially causing danger when adjusting the training mode. This enriches the training modes of rehabilitation robots and ensures the safety of patients. Summary of the Invention

[0004] This invention addresses the shortcomings of existing technologies by providing a tracking control method for the cyclical switching of active and passive training modes in a rehabilitation robot. This method can improve the rehabilitation effect for trainees and ensure the safety and stability of the human-machine system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a tracking control method for cyclically switching between active and passive training modes of a rehabilitation robot:

[0006] 1) Based on the different interference environments generated on the human-machine system during active and passive training of the rehabilitation patients, a switching dynamics model of the rehabilitation robot was constructed, and the interference environment was estimated using SSCN.

[0007] 2) A tracking error system for active and passive training was established, and a tracking control method for cyclic switching between active and passive training was proposed. Under the condition of meeting the shortest average training time, the human-machine system can realize the cyclic switching between active and passive training.

[0008] The steps are as follows:

[0009] Step 1) Based on the different interference environments generated on the human-machine system during active and passive training by the rehabilitation patient, a switching dynamics model of the rehabilitation robot was constructed, and the interference environment was estimated using SSCN. The dynamics model of the system is described as follows:

[0010]

[0011] in

[0012]

[0013]

[0014]

[0015]

[0016] Where M represents the mass of the rehabilitation robot, and m represents the mass of the user. a and M b Let X(t) represent the coefficient matrix, X(t) represent the motion trajectory of the rehabilitation robot in the x, y, and rotation angle directions, u(t) represent the control input force of the four wheels of the rehabilitation robot, and I0 represent the rotational inertia of the rehabilitation robot. The moment of inertia represents the user's rotation, and θ represents the angle between the horizontal axis and the line connecting the robot's center and the center of the first wheel, i.e., θ = θ1. This can be seen from the structure of the rehabilitation robot. θ3=θ+π, l i (i = 1, 2, 3, 4) represents the distance from the system's center of gravity to the center of each wheel, and r0 represents the distance from the center to the center of gravity. This represents the x′ axis and the l corresponding to each wheel. i The angle between them.

[0017] During active training, patients typically need to reduce stress on the robot, stand at the center of the human-machine system, and increase walking speed. The coefficient matrix M is then used to determine the patient's influence on the robot's movement. a And B(θ) is represented as: M a =M1+ΔM1, B(θ)=B1(θ); During passive training, the patient relies on the robot's support board to reduce body weight and walks with the robot's assistance. Typically, the human-machine system experiences a center of gravity shift, affecting the coefficient matrix M.a And B(θ) is represented as: M a =M2+ΔM2, B(θ)=B2(θ)+ΔB2(θ). Thus, model (1) can be transformed into the following switching expression:

[0018]

[0019] Where Ξ1(t) and Ξ2(t) represent the system's uncertain parameter matrices, and

[0020]

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] Where L represents the distance from the center of the robot to each wheel, and λ0 is a constant physical quantity.

[0027] When the rehabilitation robot switches between active and passive training, model (2) can be transformed into the following form:

[0028]

[0029] Where σ(t) :=[0,+∞)→W={1,2,...,w}, W represents the training task set, σ(t) represents the switching training signal, and σ(t)=r means that the r-th sub-training task is activated, where r∈W represents the r-th sub-training task. According to the robot's training mode, w=2, σ(t)=1 indicates that the human-machine system is performing active training, and σ(t)=2 indicates that the human-machine system is performing passive training.

[0030] As shown in model (3), the switching dynamics model for active and passive training of the rehabilitation robot is as follows:

[0031]

[0032] in This represents the interference environment of the human-machine system during the active-passive training cycle switching mode. From a physical perspective, Θ... σ(t) (t) is bounded.

[0033] Next, the SSCN method is used to analyze the interference environment Θ. σ(t) (t) performs switching estimation based on the robot's motion trajectory and velocity. Input to the network, and input weights. and threshold Connected to the hidden layer, the hidden layer outputs ψ σ(t) It can be obtained through the Gaussian function θ σ(t) get

[0034]

[0035] in For the j-th hidden layer σ(t) (j σ(t) = the output of (1, 2, ..., n) nodes, and

[0036]

[0037] in For the dth σ(t) (d σ(t) =1,2,...,6) input layer connections to the j-th layer σ(t) The input weights of each hidden layer node, For the j-th hidden layer σ(t) The threshold of each node. The SSCN hidden layer outputs weights. Connecting to the output layer provides the network output for the system's interference environment. as follows:

[0038]

[0039] in

[0040] When the number of hidden layer nodes is n σ(t) When -1, let the network output of the model be The obtained human-machine system interference environment estimation error Random configuration of the nth σ(t) The parameters of each hidden layer node are set to satisfy the following inequality:

[0041]

[0042] in, It is a sequence of non-negative real numbers and

[0043] Next, we will further explain how, under the constraint of inequality (7), the estimation error of the disturbance environment can be made to approach zero.

[0044] make

[0045]

[0046] From equations (7) and (8), we can see that the following inequalities hold.

[0047]

[0048] From equation (9), we can see that This allows for an increase in the number of hidden layer nodes. Thus, the estimation of the switching of the interference environment in the active and passive training modes of the rehabilitation robot was realized.

[0049] Step 2) A tracking error system for active and passive training was established, and a tracking control method for cyclically switching between active and passive training was proposed. Under the condition of satisfying the shortest average training time, the human-machine system can realize the cyclical switching between active and passive training. Let x1(t) = X(t) represent the actual motion trajectory of the robot. To represent the robot's actual speed, model (4) is simplified to the following form:

[0050]

[0051] During the active training phase, the actual motion trajectory X(t) of the rehabilitation robot and the active training trajectory X specified by the doctor are used. d,1 (t), trajectory tracking error e 1,1 (t) and velocity tracking error e 2,1 (t) are respectively:

[0052]

[0053] From equations (10) and (11), the active training tracking error system can be defined as follows:

[0054]

[0055] During the passive training phase, the actual motion trajectory X(t) of the rehabilitation robot and the passive training trajectory X specified by the doctor are used. d,2 (t), trajectory tracking error e 1,2 (t) and velocity tracking error e 2,2 (t) are respectively:

[0056]

[0057] From equations (10) and (13), the passive training tracking error system can be defined as follows:

[0058]

[0059] To achieve the switching between active and passive training cycles, the tracking controller is designed as follows:

[0060]

[0061] in and They are B σ(t) (θ) and The generalized inverse matrix, P σ(t) and T σ(t) I is a positive definite matrix, and I is the identity matrix.

[0062] To implement the tracking controller (15), it is necessary to design the output weights for the disturbance environment estimation. make The optimal value is The interfering environment can then be represented as and

[0063]

[0064] Therefore, the output weight estimation error can be obtained. The adaptive law for weighting is designed as follows:

[0065]

[0066] Among them l σ(t) It is a normal number.

[0067] Consider the following multi-Lyapunov function:

[0068]

[0069] The Lyapunov function of the system under the r-th sub-training task is as follows:

[0070]

[0071] Differentiating equation (18) and substituting equations (15) and (16) into it, we can obtain

[0072]

[0073] According to Young's inequality, we know that...

[0074]

[0075] Substituting equation (20) into equation (19) yields

[0076]

[0077] Where γ=min{2,2,l r},

[0078] To illustrate that the tracking error of the human-machine system and the output weight estimation error of the SSCN are bounded under the condition of satisfying the shortest average training time, we construct the function Z(t) = e γt Vσ(t) (χ(t)), then from the above formula we can obtain

[0079]

[0080] in

[0081] Integrating equation (22), and when the active and passive training tasks switch, i.e., t∈[t j ,t j+1 )∪[t j+1 ,t j+2 ), we can get

[0082]

[0083] Where t j Let t0 represent the time when the robot performs the j-th (j=0,1,2,...) switch, and t0=0.

[0084] When t j When sub-training task l switches to sub-training task r, according to equation (18), we know that

[0085]

[0086] Where υ>1 is the adjustment parameter, σ(t) j ) = r, left limit value l∈W and l≠r, that is, when l=1, r=2, and when l=2, r=1.

[0087] From equations (23) and (24), we can see that

[0088]

[0089] From j=0 to j=N c Repeating the above calculation process for (T,0)-1, we can obtain:

[0090]

[0091] Where T represents the total training time, N c (T,0) represents the number of switching times within the time interval [0,T).

[0092] Design a switching training signal σ(t) such that the average training time π a satisfy Then for any It can be known

[0093]

[0094] Based on average training time π a The relationship with the number of switching can be seen

[0095]

[0096] Where N0≥0 is the minimum number of switching operations required by the design.

[0097] Combining equations (27) and (28), we can obtain:

[0098]

[0099] According to the above formula, from N c (T,0)-j≤1+N c (T,t j+1 We can obtain:

[0100]

[0101] Depend on achievable

[0102]

[0103] Combining equations (26), (30), and (31), we can obtain

[0104]

[0105] From the above formula, we can obtain:

[0106]

[0107] From equation (18), it can be seen that there exists and Make

[0108] a||χ(t)|| 2 ≤V r (χ(t))≤b||χ(t)|| 2 (34)

[0109] Combining equations (33) and (34), we can obtain

[0110]

[0111] From the above formula, we can obtain

[0112]

[0113] Average training time Under the given conditions, it can be seen from the above formula that

[0114]

[0115] Therefore, human-machine systems meet the average training time requirement. Under certain conditions, both the tracking error and the SSCN output weight estimation error are bounded, and the rehabilitation robot can achieve active and passive cyclic switching training.

[0116] Step 3) Based on the STM32F411 series microcontroller, the output PWM signal is provided to the motor drive module, enabling the robot to assist the rehabilitation patient in the active-passive cyclic switching training specified by the doctor. The STM32F411 series microcontroller is the main controller. The input of the main controller is connected to the motor speed measurement module, and the output is connected to the motor drive module. The motor drive module is connected to the DC motor. The power supply system supplies power to all electrical devices. The main controller control method involves reading the feedback signal from the motor encoder and the command signal given by the main controller to calculate the error signal. Based on the error signal, the main controller calculates the control quantity of the motor according to a predetermined control algorithm and sends it to the motor drive module. The motor rotates, driving the wheels to maintain its balance and move in the specified manner.

[0117] Compared with the prior art, the present invention has the following advantages.

[0118] This invention relates to a tracking control method for the cyclical switching of active and passive training modes in rehabilitation robots. It has the following advantages: This invention employs the SSCN method to estimate the interference environment and constructs a switching dynamics model for the rehabilitation robot; it proposes a new cyclical switching mode for active and passive training of the rehabilitation robot, which can improve rehabilitation effectiveness, allowing patients to adapt to active training in advance and preventing danger to patients from directly entering active training; and it proposes a tracking control method for the cyclical switching of active and passive training modes in rehabilitation robots, ensuring the safety of trainees and the stability of the system, enabling the human-machine system to achieve cyclical switching between active and passive training. Attached Figure Description

[0119] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The scope of protection of the present invention is not limited to the following description.

[0120] Figure 1 This is a block diagram of the controller of the present invention.

[0121] Figure 2 This is a coordinate diagram of the system of the present invention.

[0122] Figure 3 This invention relates to the minimum system of the STM32F411 microcontroller.

[0123] Figure 4 This is the peripheral circuit for the MPU9250 of this invention.

[0124] Figure 5 This is the peripheral circuit of the motor drive module of the present invention.

[0125] Figure 6 This is the overall hardware principle circuit of the present invention. Detailed Implementation

[0126] The present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments.

[0127] like Figure 1-6 As shown, the tracking control method for the cyclic switching of active and passive training modes of the rehabilitation robot includes:

[0128] 1) Based on the different interference environments generated on the human-machine system during active and passive training of the rehabilitation patients, a switching dynamics model of the rehabilitation robot was constructed, and the interference environment was estimated using SSCN.

[0129] 2) A tracking error system for active and passive training was established, and a tracking control method for cyclic switching between active and passive training was proposed. Under the condition of meeting the shortest average training time, the human-machine system can realize the cyclic switching between active and passive training.

[0130] The steps are as follows:

[0131] Step 1) Based on the different interference environments generated on the human-machine system during active and passive training by the rehabilitation patient, a switching dynamics model of the rehabilitation robot was constructed, and the interference environment was estimated using SSCN. The dynamics model of the system is described as follows:

[0132]

[0133] in

[0134]

[0135]

[0136]

[0137]

[0138] Where M represents the mass of the rehabilitation robot, and m represents the mass of the user. a and M b Let X(t) represent the coefficient matrix, X(t) represent the motion trajectory of the rehabilitation robot in the x, y, and rotation angle directions, u(t) represent the control input force of the four wheels of the rehabilitation robot, and I0 represent the rotational inertia of the rehabilitation robot. The moment of inertia represents the user's rotation, and θ represents the angle between the horizontal axis and the line connecting the robot's center and the center of the first wheel, i.e., θ = θ1. This can be seen from the structure of the rehabilitation robot. θ3=θ+π, l i(i = 1, 2, 3, 4) represents the distance from the system's center of gravity to the center of each wheel, and r0 represents the distance from the center to the center of gravity. This represents the x′ axis and the l corresponding to each wheel. i The angle between them.

[0139] During active training, patients typically need to reduce stress on the robot, stand at the center of the human-machine system, and increase walking speed. The coefficient matrix M is then used to determine the patient's influence on the robot's movement. a And B(θ) is represented as: M a =M1+ΔM1, B(θ)=B1(θ); During passive training, the patient relies on the robot's support board to reduce body weight and walks with the robot's assistance. Typically, the human-machine system experiences a center of gravity shift, affecting the coefficient matrix M. a And B(θ) is represented as: M a =M2+ΔM2, B(θ)=B2(θ)+ΔB2(θ). Thus, model (1) can be transformed into the following switching expression:

[0140]

[0141] Where Ξ1(t) and Ξ2(t) represent the system's uncertain parameter matrices, and

[0142]

[0143]

[0144]

[0145]

[0146]

[0147]

[0148] Where L represents the distance from the center of the robot to each wheel, and λ0 is a constant physical quantity.

[0149] When the rehabilitation robot switches between active and passive training, model (2) can be transformed into the following form:

[0150]

[0151] Where σ(t) :=[0,+∞)→W={1,2,...,w}, W represents the training task set, σ(t) represents the switching training signal, and σ(t)=r means that the r-th sub-training task is activated, where r∈W represents the r-th sub-training task. According to the robot's training mode, w=2, σ(t)=1 indicates that the human-machine system is performing active training, and σ(t)=2 indicates that the human-machine system is performing passive training.

[0152] As shown in model (3), the switching dynamics model for active and passive training of the rehabilitation robot is as follows:

[0153]

[0154] in This represents the interference environment of the human-machine system during the active-passive training cycle switching mode. From a physical perspective, Θ... σ(t) (t) is bounded.

[0155] Next, the SSCN method is used to analyze the interference environment Θ. σ(t) (t) performs switching estimation based on the robot's motion trajectory and velocity. Input to the network, and input weights. and threshold Connected to the hidden layer, the hidden layer outputs ψ σ(t) It can be obtained through the Gaussian function θ σ(t) get

[0156]

[0157] in For the j-th hidden layer σ(t) (j σ(t) = the output of (1, 2, ..., n) nodes, and

[0158]

[0159] in For the dth σ(t) (d σ(t) =1,2,...,6) input layer connections to the j-th layer σ(t) The input weights of each hidden layer node, For the j-th hidden layer σ(t) The threshold of each node. The SSCN hidden layer outputs weights. Connecting to the output layer provides the network output for the system's interference environment. as follows:

[0160]

[0161] in

[0162] When the number of hidden layer nodes is n σ(t) When -1, let the network output of the model be The obtained human-machine system interference environment estimation error Random configuration of the nth σ(t) The parameters of each hidden layer node are set to satisfy the following inequality:

[0163]

[0164] in, It is a sequence of non-negative real numbers and

[0165] Next, we will further explain how, under the constraint of inequality (7), the estimation error of the disturbance environment can be made to approach zero.

[0166] make

[0167]

[0168] From equations (7) and (8), we can see that the following inequalities hold.

[0169]

[0170] From equation (9), we can see that This allows for an increase in the number of hidden layer nodes. Thus, the estimation of the switching of the interference environment in the active and passive training modes of the rehabilitation robot was realized.

[0171] Step 2) A tracking error system for active and passive training was established, and a tracking control method for cyclically switching between active and passive training was proposed. Under the condition of satisfying the shortest average training time, the human-machine system can realize the cyclical switching between active and passive training. Let x1(t) = X(t) represent the actual motion trajectory of the robot. To represent the robot's actual speed, model (4) is simplified to the following form:

[0172]

[0173] During the active training phase, the actual motion trajectory X(t) of the rehabilitation robot and the active training trajectory X specified by the doctor are used. d,1 (t), trajectory tracking error e 1,1 (t) and velocity tracking error e 2,1 (t) are respectively:

[0174]

[0175] From equations (10) and (11), the active training tracking error system can be defined as follows:

[0176]

[0177] During the passive training phase, the actual motion trajectory X(t) of the rehabilitation robot and the passive training trajectory X specified by the doctor are used. d，2 (t), trajectory tracking error e 1,2 (t) and velocity tracking error e 2,2 (t) are respectively:

[0178]

[0179] From equations (10) and (13), the passive training tracking error system can be defined as follows:

[0180]

[0181] To achieve the switching between active and passive training cycles, the tracking controller is designed as follows:

[0182]

[0183] in and They are B σ(t) (θ) and The generalized inverse matrix, P σ(t) and T σ(t) I is a positive definite matrix, and I is the identity matrix.

[0184] To implement the tracking controller (15), it is necessary to design the output weights for the disturbance environment estimation. make The optimal value is The interfering environment can then be represented as and

[0185]

[0186] Therefore, the output weight estimation error can be obtained. The adaptive law for weighting is designed as follows:

[0187]

[0188] Among them l σ(t) It is a normal number.

[0189] Consider the following multi-Lyapunov function:

[0190]

[0191] The Lyapunov function of the system under the r-th sub-training task is as follows:

[0192]

[0193] Differentiating equation (18) and substituting equations (15) and (16) into it, we can obtain

[0194]

[0195] According to Young's inequality, we know that...

[0196]

[0197] Substituting equation (20) into equation (19) yields

[0198]

[0199] Where γ=min{2,2,l r},

[0200] To illustrate that the tracking error of the human-machine system and the output weight estimation error of the SSCN are bounded under the condition of satisfying the shortest average training time, we construct the function Z(t) = e γt V σ(t) (χ(t)), then from the above formula we can obtain

[0201]

[0202] in

[0203] Integrating equation (22), and when the active and passive training tasks switch, i.e., t∈[t j ,t j+1 )∪[t j+1 ,t j+2 ), we can get

[0204]

[0205] Where tj represents the time when the robot performs the j-th (j=0,1,2,...) switch, and t0=0.

[0206] When sub-training task l switches to sub-training task r at time tj, according to equation (18), we know that

[0207]

[0208] Where υ>1 is the adjustment parameter, σ(t) j ) = r, left limit value l∈W and l≠r, that is, when l=1, r=2, and when l=2, r=1.

[0209] From equations (23) and (24), we can see that

[0210]

[0211] From j=0 to j=N c Repeating the above calculation process for (T,0)-1, we can obtain:

[0212]

[0213] Where T represents the total training time, and Nc(T,0) represents the number of switching times within the time interval [0,T).

[0214] Design a switching training signal σ(t) such that the average training time π a satisfy Then for any It can be known

[0215]

[0216] Based on average training time π a The relationship with the number of switching can be seen

[0217]

[0218] Where N0≥0 is the minimum number of switching operations required by the design.

[0219] Combining equations (27) and (28), we can obtain:

[0220]

[0221] According to the above formula, from N c (T,0)-j≤1+N c (T,t j+1 We can obtain:

[0222]

[0223] Depend on achievable

[0224]

[0225] Combining equations (26), (30), and (31), we can obtain

[0226]

[0227] From the above formula, we can obtain:

[0228]

[0229] From equation (18), it can be seen that there exists and Make

[0230] a||χ(t)|| 2 ≤V r (χ(t))≤b||χ(t)|| 2 (34)

[0231] Combining equations (33) and (34), we can obtain

[0232]

[0233] From the above formula, we can obtain

[0234]

[0235] Average training time Under the given conditions, it can be seen from the above formula that

[0236]

[0237] Therefore, human-machine systems meet the average training time requirement. Under certain conditions, both the tracking error and the SSCN output weight estimation error are bounded, and the rehabilitation robot can achieve active and passive cyclic switching training.

[0238] Step 3) Based on the STM32F411 series microcontroller, the output PWM signal is provided to the motor drive module, enabling the robot to assist the rehabilitation patient in the active-passive cyclic switching training specified by the doctor. The STM32F411 series microcontroller is the main controller. The input of the main controller is connected to the motor speed measurement module, and the output is connected to the motor drive module. The motor drive module is connected to the DC motor. The power supply system supplies power to all electrical devices. The main controller control method involves reading the feedback signal from the motor encoder and the command signal given by the main controller to calculate the error signal. Based on the error signal, the main controller calculates the control quantity of the motor according to a predetermined control algorithm and sends it to the motor drive module. The motor rotates, driving the wheels to maintain its balance and move in the specified manner.

[0239] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Therefore, these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A tracking control method for cyclically switching between active and passive training modes of a rehabilitation robot, characterized in that: Includes the following steps: Step 1: Based on the different interference environments generated on the human-machine system during the active and passive training of the rehabilitation patients, a switching dynamics model of the rehabilitation robot was constructed, and the interference environment was estimated using SSCN. Step 2: A tracking error system for active and passive training was established, and a tracking control method for cyclic switching between active and passive training was proposed. Under the condition of meeting the shortest average training time, the human-machine system can realize the cyclic switching between active and passive training. In step 1, the dynamic model of the human-machine system is as follows: （1） in ， ， ， in Indicating the quality of rehabilitation robots, Indicates user quality. and Represents the coefficient matrix. Indicating that rehabilitation robots are in , The trajectory of motion in three directions: rotation angle, and direction. The control force is input to the four wheels of the rehabilitation robot. This represents the moment of inertia of the rehabilitation robot. This represents the user's moment of inertia. This represents the angle between the horizontal axis and the line connecting the robot's center and the center of the first wheel. As can be seen from the structure of the rehabilitation robot, , , , This represents the distance from the system's center of gravity to the center of each wheel. Indicates the distance from the center to the centroid. express Axle and each wheel corresponding The angle between them; During active training, the patient needs to reduce stress on the robot, stand at the center of the human-machine system, and increase walking speed. The coefficient matrix is ​​then used to determine the patient's influence on the robot's movement. and Represented as: , ; During passive training, the patient relies on a robotic support board to reduce body weight and performs walking training with the robot's assistance. Typically, this causes a shift in the center of gravity within the human-machine system, affecting the coefficient matrix. and Represented as: , ; Thus, model (1) can be transformed into the following switching expression: （2） in, and This represents the system's uncertain parameter matrix, and ， ， in This represents the distance from the robot's center to each wheel. It is a steady physical quantity; When the rehabilitation robot switches between active and passive training, model (2) takes the following form: （3） in , Represents the training task set. This indicates a switch to the training signal. Meaning the first Individual training tasks are activated. Indicates the first Individual training tasks; obtained based on the robot's active and passive training modes. , This indicates that the human-machine system is undergoing active training. This indicates that the human-machine system is undergoing passive training; As shown in model (3), the switching dynamics model for active and passive training of the rehabilitation robot is as follows: （4） in This represents the interference environment of the human-machine system during the active-passive training cycle switching mode, which can be understood from a physical perspective. Bounded; Next, the SSCN method is used to analyze the interference environment. Perform switching estimation based on robot motion trajectory and speed Input to the network, and input weights. and threshold Connected to the hidden layer, the hidden layer outputs... Gaussian function get （5） in For the hidden layer The output of each node, and in For the first The input layer connection is the first The input weights of each hidden layer node, For the hidden layer The threshold of each node; the SSCN hidden layer outputs weights. Connected to the output layer, the network output of the system's interference environment is obtained. as follows: （6） in ; When the number of hidden layer nodes is At that time, let the network output of the model be The obtained human-machine system interference environment estimation error Random configuration of the first The parameters of each hidden layer node are set to satisfy the following inequality: （7） in, It is a sequence of non-negative real numbers and , , ; Next, we will explain how the estimation error of the disturbance environment tends to zero under the constraint of inequality (7); make （8） From equations (7) and (8), we can see that the following inequalities hold. （9） From equation (9), we can see that This allows for an increase in the number of hidden layer nodes. Thus, the switching estimation of the interference environment in the active and passive training modes of the rehabilitation robot was realized. .

2. The tracking control method for cyclically switching active and passive training modes of a rehabilitation robot according to claim 1, characterized in that: Step 2 includes: make This represents the robot's actual movement trajectory. To represent the robot's actual speed, model (4) is simplified to the following form: （10） During the active training phase, based on the actual movement trajectory of the rehabilitation robot... and the active training trajectory specified by the doctor The trajectory tracking error is obtained. and speed tracking error They are respectively: （11） From equations (10) and (11), the active training tracking error system can be defined as follows: （12） During the passive training phase, based on the actual movement trajectory of the rehabilitation robot... and the passive training trajectory specified by the doctor The trajectory tracking error is obtained. and speed tracking error They are respectively: （13） From equations (10) and (13), the passive training tracking error system can be defined as follows: （14） To achieve the switching between active and passive training cycles, the tracking controller is designed as follows: （15） in and They are and The generalized inverse matrix, and It is a positive definite matrix. It is the identity matrix; To implement the tracking controller (15), the output weights for the interference environment estimation are designed. ;make The optimal value is Then the interfering environment can be represented as ,and Therefore, the output weight estimation error can be obtained. The adaptive law for design weights is as follows: （16） in It is a positive number; Consider the following multi-Lyapunov function: （17） Then the first The Lyapunov function of the system under individual training tasks is as follows: （18） Differentiating equation (18) and substituting equations (15) and (16) into it, we get （19） According to Young's inequality, we know that... （20） Substituting equation (20) into equation (19) yields （21） in , ; To further illustrate that, under the condition of satisfying the shortest average training time, the tracking error of the human-machine system and the output weight estimation error of the SSCN are bounded, a constructor is constructed. From the above formula, we can obtain （22） in ; Integrating equation (22), and when the active and passive training tasks switch, i.e. , can be obtained （23） in Indicates the robot has occurred for the first time. The moment of the next switch, and ; when Time training task Switch to sub-training task At that time, according to equation (18), it can be known that （24） in It's about adjusting parameters. Left limit value , and That is, when hour ,when hour ; From equations (23) and (24), we can see that （25） from arrive Repeating the above calculation process, we can obtain: （26） in Indicates the total training time. Indicates the time interval Number of switching times within; Design to switch training signals Average training time satisfy Then for any It can be known that （27） Based on average training time The relationship with the number of switching can be seen （28） in It represents the minimum number of switching operations designed for. Combining equations (27) and (28), we can obtain: （29） According to the above formula, from We can obtain: （30） Depend on , can be obtained （31） Combining equations (26), (30), and (31), we can obtain （32） From the above formula, we can obtain: （33） From equation (18), it can be seen that there exists and , making （34） Combining equations (33) and (34), we can obtain （35） From the above formula, we can obtain （36） Average training time Under the given conditions, it can be seen from the above formula that （37） Therefore, human-machine systems meet the average training time requirement. Under certain conditions, both the tracking error and the SSCN output weight estimation error are bounded, and the rehabilitation robot can achieve active and passive cyclic switching training.

3. The tracking control method for cyclically switching active and passive training modes of a rehabilitation robot according to claim 2, characterized in that: It also includes step 3, which uses an STM32F411 series microcontroller to output a PWM signal to the motor drive module, enabling the robot to help the patient undergo active-passive cycle switching training as specified by the doctor.

4. The tracking control method for cyclically switching active and passive training modes of a rehabilitation robot according to claim 3, characterized in that: The STM32F411 series microcontroller is used as the main controller. The input of the main controller is connected to the motor speed measurement module, and the output is connected to the motor drive module. The motor drive module is connected to the DC motor. The power supply system supplies power to various electrical devices. The main controller control method involves reading the feedback signal from the motor encoder and the command signal given by the main controller, and then calculating the error signal. Based on the error signal, the main controller calculates the control quantity of the motor according to the predetermined control algorithm and sends it to the motor drive module. The motor rotates to drive the wheels to maintain their own balance and move in the specified manner.

Images