CN-122005265-A - Control method of finger rehabilitation exoskeleton combining metamorphic principle and rope driving module

CN122005265ACN 122005265 ACN122005265 ACN 122005265ACN-122005265-A

Abstract

The invention discloses a control method of a finger rehabilitation exoskeleton combining a metamorphic principle and a rope driving module, which comprises the following steps of collecting kinematic data of the exoskeleton during flexion and extension movements through a sensor system, carrying out filtering pretreatment on the data, calculating to obtain expected moment of each joint and tension of a cable to obtain final expected tension, inputting the collected data and the calculated data into a PID controller optimized by a TD3 algorithm, obtaining optimized PID parameters through training and updating, realizing cable tension tracking, evaluating the advantages and disadvantages of the PID controller optimized by the TD3 algorithm, measuring the error degree of the expected tension and a true value, and adjusting the PID controller parameters optimized by the TD3 algorithm to control the tension of each cable. According to the invention, the PID controller is optimized by adopting the TD3 algorithm, the PID parameter adjustment quantity is dynamically output, the accuracy of the control signal of the driving motor and the tension tracking precision are obviously improved, the man-machine interaction flexibility is improved, and the safety and the effectiveness of rehabilitation training are ensured.

Inventors

CHEN BING
WU YIXIN
PENG JIANGUO
GAO QIZHI
LIANG HAOXIN

Assignees

合肥工业大学

Dates

Publication Date: 20260512
Application Date: 20260327

Claims (10)

1. The control method of the finger rehabilitation exoskeleton combining the metamorphic principle and the rope driving module is characterized by comprising the following steps of: S10, acquiring kinematic data of the exoskeleton during flexion and extension movements through a sensor system, wherein the kinematic data comprise cable tension data acquired by a tension sensor and rotation angle data of each joint acquired by an angle sensor; S11, respectively carrying out filtering pretreatment on the collected angle data and the cable tension data; S12, according to the motion angle track of each joint, calculating to obtain expected moment of each joint through a dynamic model of a human finger, carrying out stress analysis calculation on each cable through a finger exoskeleton dynamic model, calculating to obtain pulling force of each cable required by finger bending/stretching through combining the obtained expected moment of each joint, and finally judging to obtain friction compensation models under different configurations through metacarpophalangeal joint angles to obtain final expected pulling force; s13, inputting the calculated expected value of the tension of each cable and the value of each joint angle under each configuration, and the tension and each joint angle acquired by an actual sensor into a PID controller optimized by a TD3 algorithm, training through a reward function, a strategy network and a value network of the TD3 algorithm, and continuously carrying out corresponding network updating to obtain optimized PID parameters, thereby realizing cable tension tracking of the motor in a speed mode; S14, evaluating the merits of the PID controller optimized by the TD3 algorithm by utilizing root mean square error, and measuring the error degree of the expected tension and the true value; and S15, comparing the calculated expected tension curve with the actual tension and angle value acquired by the sensor by the built-in data processing software of the upper computer, automatically adjusting parameters of the PID controller optimized by the TD3 algorithm, and controlling the driving mechanism to work through the lower computer to adjust the tensioning degree of each cable, so as to control the tension of each cable.
2. The method for controlling the finger rehabilitation exoskeleton combining the metamorphic principle and the rope driving module according to claim 1, wherein in the step S11, the specific method for preprocessing the angle data is as follows: Performing linear interpolation on joint angle data acquired by an angle sensor, eliminating the point loss phenomenon in the data acquisition process, then removing power frequency interference by adopting a 50Hz notch filter, and filtering high-frequency noise by adopting a Butterworth band-pass filter with the bandwidth of 20Hz-500Hz to obtain a smooth joint angle curve; The specific method for preprocessing the cable tension data comprises the following steps: The tension data collected by the tension sensor is firstly removed by a hardware filter circuit, then a 50Hz notch filter and a 20Hz-500Hz Butterworth band-pass filter which are the same as the angle data are adopted for software filtering, the preprocessed tension data are used as windows with a preset number of sampling points, the root mean square RMS of the time domain characteristics is extracted for characteristic analysis, and the calculation formula of the RMS is as follows: ; Wherein, the Is a cable tension signal; Is the window size.
3. The method for controlling a finger rehabilitation exoskeleton combining a metamorphic principle and a rope driving module according to claim 1, wherein in step S12, a formula for calculating a desired moment of any joint of the finger is: ; Wherein, the For the mass matrix of the finger in which the joint is located, , , The angular acceleration vector, the velocity vector and the angle vector of the joint are respectively, In the form of a coriolis Li Juzhen, In order to create a moment of the joint due to gravity, In the form of a matrix of viscous damping coefficients, Is a joint stiffness coefficient matrix.
4. The method for controlling a finger rehabilitation exoskeleton combining a metamorphic principle and a rope driving module according to claim 3, wherein the friction compensation model comprises a rope-rope sleeve friction model and a rope-pulley friction model; The rope-rope sleeve friction model adopts a static transmission model of coulomb friction, and the calculation formula is as follows: ; Wherein, the In order to input the rope tension, As a coefficient of the direction of movement of the rope, Is the friction coefficient of the rope sleeve, The bending angle of the rope sleeve is set; the rope-pulley friction model takes coulomb friction as a core, and the calculation formula is as follows: ; Wherein, the For the number of fixed pulleys, 、、 The constants of the experimental identification are respectively set, The wrap angle of the rope on the pulley is the sum of wrap angles of the movable pulley and the corresponding fixed pulley in the buckling process, and the sum of wrap angles of the movable pulley and the corresponding fixed pulley in the stretching process.
5. The method for controlling a finger rehabilitation exoskeleton combining a metamorphic principle and a rope driving module according to claim 4, wherein the calculation formula for calculating the basic tension of the cable by combining the expected moment of the joint is as follows: ; Wherein, the For a desired torque when the cable is flexed and extended for the corresponding joint of the corresponding finger, For the conversion between the rope tension of the rope and the tension required by the corresponding joint flexion and extension of the corresponding finger, An equivalent arm of the bending moment is generated on the metacarpophalangeal joints, the proximal interphalangeal joints and the distal interphalangeal joints for the rope tension of the rope, Is the angle parameter of metacarpophalangeal joints in the buckling process, Is the angle parameter of the metacarpophalangeal joint in the stretching process; the final desired tension For superposition of basic tension and friction compensation value, the calculation formula is: ; ; Wherein, the first three configurations are respectively the bending and stretching states of the metacarpophalangeal joints, and the second four configurations are respectively the bending and stretching states of the rest joints.
6. The method for controlling finger rehabilitation exoskeleton combining metamorphic principle and rope driving module according to any one of claims 1 to 5, wherein in step S13, in the PID controller optimized by the TD3 algorithm, the specific process of training through the reward function, the policy network and the value network of the TD3 algorithm and continuously performing corresponding network update to obtain the optimized PID parameters is as follows: The tensile force error, the error differential, the error integral, the joint angle and the actual tensile force are used as state space parameters of the TD3 algorithm, the state space parameters are used as input parameters of a reward function of the TD3 algorithm, the input parameters are processed by the reward function and then output to a reinforcement learning module of the TD3 algorithm, the reinforcement learning module is trained by a strategy network and a value network of the reinforcement learning module to continuously perform corresponding updating processing to obtain optimized PID parameter gain, then the PID parameter gain is adjusted by an action space of the TD3 algorithm, and finally a control signal of a PID controller is output.
7. The method for controlling a finger rehabilitation exoskeleton combining a metamorphic principle and a rope drive module according to claim 6, wherein the state space of the TD3 algorithm is characterized in that The definition is as follows: ; Wherein, the To account for the error in the desired tension from the actual tension, As a differential of the error, As an integral of the error, For the actual detected metacarpophalangeal joint angle of the thumb, A desired tension for the cable; action space of TD3 algorithm The definition is as follows: ; Wherein, the 、、 Incremental adjustment amounts of PID proportion, integral and differential parameters respectively; Reward function of TD3 algorithm The design is as follows: ; Wherein, the 、、 Respectively weight coefficients for balancing tension errors, thumb metacarpophalangeal joint angle deviation and controlling energy consumption, For the deviation of the actual angle of the metacarpophalangeal joint of the thumb from the desired angle, Is the module length of the control signal; the formula for updating the value network is: ; In the formula, A loss function value for the value network; Is the batch size, i.e., the number of samples sampled at each update; In-state for the current value network Take the following steps During the action Estimating a value; is the first Instant rewards for individual samples; for discount factors, weights for balancing current rewards and future rewards are in the range of ; The system is a target value network and is used for stabilizing training; The target strategy network is used for outputting a target action of the next state; kou update delay update policy network:  ; In the formula, Gradient as a policy network objective function; Is a parameter of the policy network; acting on for value network Is a gradient of (2); outputting a gradient to the own parameters for the policy network; In-state for policy network A lower output action; The soft update target network is: ; In the formula, The target network comprises a target value network and a target strategy network as parameters of the target network; Is a parameter of the current online network; for soft update coefficient, controlling update speed of target network; Control signal of PID controller The calculation formula is as follows: ; Wherein, the 、、 The initial proportional, integral, derivative gains of the PID, In order to account for the tension error, Is the differential of the error.
8. The method for controlling the finger rehabilitation exoskeleton combining the metamorphic principle and the rope driving module according to claim 1, wherein the strategy network of the TD3 algorithm is a 4-layer deep neural network, and comprises a state input layer, 2 hidden layers and an action output layer, the number of neurons is 5, 64, 48 and 3 in sequence, an activation function is LeakyReLU, and an output layer is a Tanh function; The value network is a 5-layer deep neural network and comprises an input layer, 3 hidden layers and a Q value output layer, the number of neurons is 8, 48, 24 and 1 in sequence, and the activation function adopts LeakyReLU, leakyReLU super-parameters lambda to take 0.001.
9. The method for controlling a finger rehabilitation exoskeleton combining a metamorphic principle and a rope drive module according to any one of claims 1 to 5 or 7 or 8, wherein in step S14, a root mean square error is used The formula of (2) is: ; Wherein, the Is the pulling force of the actual cable and, Is the expected tension of the cable obtained by a kinematic and dynamic model and a friction model, Is the data length of the test sample sequence.
10. A finger rehabilitation exoskeleton combining a metamorphic principle and a rope drive module, wherein the finger rehabilitation exoskeleton employs the control method of any one of claims 1 to 9.

Description

Control method of finger rehabilitation exoskeleton combining metamorphic principle and rope driving module Technical Field The invention belongs to the technical field of finger rehabilitation exoskeleton robots, and particularly relates to a control method of a finger rehabilitation exoskeleton by combining a metamorphic principle with a rope driving module. Background The finger rehabilitation exoskeleton is a wearable human-machine collaboration system capable of producing controllable assistance torque at the joints of the wearer's fingers. Compared with the traditional manual rehabilitation, the finger rehabilitation exoskeleton has the characteristics of high rehabilitation pertinence and accurate and controllable training intensity, has a similar kinematic structure as the human hand, is generally designed with a driver, a sensor, a controller, a power supply and the like, can provide controllable auxiliary force/moment for the finger joints of a patient, and helps the patient to regain finger muscle strength and motion control functions. After wearing the exoskeleton, the finger dysfunction patient can realize the actions of finger bending and stretching, grasping, releasing and the like as normal people under the supporting and power driving actions of the exoskeleton, and can greatly improve the self-care ability of the daily life and the confidence of the life of the patient. For the finger rehabilitation exoskeleton, due to obvious individual variability, the rope driving structure is easy to be interfered by friction, so that the prior art has a plurality of defects. Generally, the existing device collects data through a single sensor or calculates expected tension of a cable through a simplified model, however, because a rope-rope sleeve, rope-pulley friction and other nonlinear interference exist in a rope driving system, tension data collected by the single tension sensor is easy to be influenced by noise, the error is large, the calculation of the expected tension is inaccurate, if the expected tension is output through only establishing a simple dynamics model, individual differences of all joint movement angles of fingers cannot be adapted, friction change under different movement configurations is not considered, calculation complexity is high, system response is slow, and tension tracking accuracy is insufficient. In addition, the conventional PID control is adopted in the existing device to realize tension closed loop tracking, parameters of the existing device need manual adjustment through manual experience, time and labor are wasted, the existing device cannot be dynamically adapted to parameter changes of time sequence movements of multiple joints of fingers, and the control effect is poor. Meanwhile, the existing exoskeleton does not combine with metamorphic principle to design a configuration recognition mechanism, so that orderly collaborative switching of metacarpophalangeal joints and proximal and distal interphalangeal joints is difficult to realize, joint movement configuration recognition is inaccurate, tension changes are discontinuous during configuration conversion, problems of transmission jamming and poor man-machine interaction flexibility are easy to occur, and even secondary damage to fingers of a patient is likely to occur. Disclosure of Invention In order to quickly and accurately acquire the accurate expected tension of the cable, the invention establishes a collaborative calculation mechanism of a human finger dynamics model, a finger exoskeleton kinematics model and a friction compensation model (comprising a rope-rope sleeve and a rope-pulley friction model), and judges the movement configuration by combining the angle data of each joint, and the friction compensation model is matched pertinently, so that the accurate expected tension of the cable under different configurations is obtained. In addition, a TD3 algorithm is adopted to optimize the PID controller, PID parameter adjustment quantity is dynamically output, and accuracy of control signals of the driving motor and tension tracking accuracy are improved. In order to solve the problems of coordinated control of multi-joint time sequence of fingers and inaccurate configuration identification, joint angle data are acquired in real time through a multi-angle sensor, finger movement configurations are identified, automatic switching is triggered when metacarpophalangeal joints reach buckling or stretching limit, ordered movement of metacarpophalangeal joints and near-end and far-end interphalangeal joints is achieved, meanwhile, the configurations of rope pull-up force and motor angles are established, closed-loop control of a driving motor is achieved, man-machine interaction compliance is improved, the problem of transmission jamming is avoided, and safety and effectiveness of rehabilitation training are guaranteed. In order to solve the technical problems, the invention adopts a technical scheme that: a