Closed-loop feedback functional electrical stimulation control method and device, medium and equipment

Abstract

The invention discloses a closed-loop feedback functional electrical stimulation control method and device, a medium and equipment, and relates to the technical field of computers. According to the invention, the current target motion angle of the healthy side limb of the patient and the actual motion angle of the affected side limb of the patient under the current functional electrical stimulation control signal are obtained, and the error between the target motion angle and the actual motion angle and the change rate of the error are determined; then, an adaptive weight fuzzy iterative learning control closed-loop feedback control algorithm is adopted to accurately adjust control parameters through learning of historical control experience, so that the affected-side limb of the patient can quickly approach a target movement track of the unaffected-side limb of the patient, and the iterative learning frequency of the iterative learning control algorithm is reduced; the limb movement reconstruction real-time performance is improved, the target trajectory tracking precision is improved, and the accomplishment degree of the limb movement on the affected side of the patient is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly relates to a closed-loop feedback functional electrical stimulation control method, device, medium, and equipment. Background Art

[0002] Currently, when stroke hemiplegia patients use existing open-loop or closed-loop functional electrical stimulation (FES) systems for functional rehabilitation training, there are generally problems such as low action completion rate, unstable and incoherent limb movement states, which severely restrict the limb function recovery effect of patients during the golden period of rehabilitation training.

[0003] Open-loop controlled FES has significant limitations in promoting the movement of paralyzed limbs, mainly manifested as the inability to correct errors and disturbances during movement, resulting in a relatively low action completion rate. In addition, open-loop FES also faces challenges such as electrode placement, muscle fatigue, body posture changes, time-varying characteristics of muscle dynamics, and external disturbances. These factors jointly affect the ability of paralyzed limbs to perform stable and smooth functional training actions, thereby limiting its rehabilitation effect.

[0004] To overcome the limitations of open-loop control, closed-loop controlled FES came into being. The closed-loop system can more effectively suppress position offset, muscle response changes, and fatigue phenomena by introducing a feedback controller to monitor muscle responses or limb movement states in real time. Based on this feedback information, the system can dynamically adjust the electrical stimulation parameters to ensure that the charge amount is accurately applied to the target muscle. This personalized and precise stimulation strategy enables paralyzed limbs to perform functional actions that match the expected movement trajectory stably and smoothly, significantly improving the rehabilitation effect. However, the design and implementation of closed-loop FES systems also face many challenges, especially the tracking control of joint angles. Due to the nonlinearity of the musculoskeletal system and the delay effect of electrical stimulation responses, it is particularly difficult to achieve precise control.

[0005] In the design of closed-loop feedback FES control strategies, the controller and the model are two core elements. According to whether a muscle electrical stimulation response model is used to evaluate the effectiveness of the controller, closed-loop feedback FES control strategies are divided into model-free control and model-based control categories. Model-free control relies on real-time muscle signals or predicted responses to adjust parameters, with strong adaptability and the ability to handle complex nonlinear relationships, but it has high data dependence, large computational resource requirements, and insufficient model interpretability. Model-based control can provide accurate prediction and control when the model is accurate, with strong adaptability and diverse controller selection, but the model construction is complex, vulnerable to interference, and computationally intensive.

[0006] Finally, both model-free closed-loop feedback and model-based closed-loop feedback control algorithms will be deployed in the embedded rehabilitation training system to assist paralyzed patients in performing rehabilitation training tasks and promoting the recovery of paralyzed limb functions. Iterative learning control (ILC) is a closed-loop feedback control algorithm that is more suitable for combination with FES technology. It can utilize the repeatability of rehabilitation training tasks to update control signals, reduce tracking errors, and improve tracking accuracy. However, the ILC control algorithm also faces challenges such as dealing with nonlinear systems and non-repetitive disturbances, learning rate selection, and real-time performance. For this reason, researchers have proposed methods such as estimation-based multi-model ILC and proportional-integral-derivative (PID)-type ILC combined with adaptive network fuzzy inference systems to address these challenges. However, the current closed-loop feedback control algorithms based on ILC still have problems such as a long number of iterations, poor real-time performance, and unstable control performance. Summary of the Invention

[0007] Based on this, it is necessary to provide a closed-loop feedback functional electrical stimulation control method, device, medium, and equipment for the above technical problems.

[0008] The present invention adopts the following technical solutions:

[0009] The present invention provides a closed-loop feedback functional electrical stimulation control method, including:

[0010] Obtain the target motion angle of the patient's healthy limb and the actual motion angle of the patient's affected limb under the current functional electrical stimulation control signal, and determine the error between the target motion angle and the actual motion angle and the change rate of the error;

[0011] Fuzzify the current error and the change rate of the error, and perform fuzzy inference according to the preset fuzzy control rules to determine the next round of PID control parameters corresponding to the current error and the change rate of the error;

[0012] Determine the adaptive weight of the next round of PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold, and the adaptive weight is negatively correlated with the ratio;

[0013] Determine an exponent according to the absolute value of the current error to determine an exponentially increasing penalty factor, and the penalty factor is positively correlated with the absolute value of the error;

[0014] Weight the PID control parameters according to the adaptive weight, and determine the control gain based on the weighted PID control parameters; weight the control gain by a penalty factor, and update the next-round functional electrical stimulation control signal according to the weighted control gain.

[0015] Optionally, the fuzzification of the current error and the rate of change of the error specifically includes:

[0016] Map the current error and the rate of change of the error to a preset universe of discourse;

[0017] According to the membership functions of the fuzzy sets corresponding to different error degrees in the universe of discourse, determine the membership degrees of the current error and the rate of change of the error belonging to different fuzzy sets.

[0018] Optionally, the fuzzy inference according to the preset fuzzy control rules to determine the next-round PID control parameters corresponding to the current error and the rate of change of the error specifically includes:

[0019] According to the preset fuzzy control rules and the membership degrees of the current error and the rate of change of the error belonging to different fuzzy sets, determine the membership degrees of the current error and the rate of change of the error to each activated fuzzy control rule;

[0020] Defuzzify according to the membership degrees of the current error and the rate of change of the error to each activated fuzzy control rule, and obtain the next-round PID control parameters corresponding to the current error and the rate of change of the error.

[0021] Optionally, the determination of the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold specifically includes:

[0022] Determine the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold through the following formula:

[0023]

[0024] where α k (n) is the adaptive weight of the PID control parameters in the nth round of iteration, β and γ are custom parameters based on the controlled model, e(n) is the current error, and e max is the maximum error threshold.

[0025] Optionally, the determination of the exponential to determine the exponentially growing penalty factor according to the absolute value of the current error specifically includes:

[0026] Determine the exponential to determine the exponentially growing penalty factor according to the absolute value of the current error through the following formula:

[0027] ρ = 1 + e (θ·|e(n)|) ;

[0028] Where ρ is the penalty factor, θ is a custom parameter based on the controlled model, and e(n) is the current error.

[0029] The present invention provides a closed-loop feedback functional electrical stimulation control device, comprising:

[0030] An acquisition module, configured to acquire the target motion angle of the patient's healthy limb and the actual motion angle of the patient's affected limb under the current functional electrical stimulation control signal, and determine the error between the target motion angle and the actual motion angle and the change rate of the error;

[0031] A fuzzy inference module, configured to fuzzify the current error and the change rate of the error, and perform fuzzy inference according to a preset fuzzy control rule to determine the next-round PID control parameters corresponding to the current error and the change rate of the error;

[0032] A weight determination module, configured to determine the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and a preset maximum error threshold, and the adaptive weight is negatively correlated with the ratio;

[0033] A penalty factor determination module, configured to determine an exponent according to the absolute value of the current error to determine an exponentially increasing penalty factor, and the penalty factor is positively correlated with the absolute value of the error;

[0034] A weighted control module, configured to weight the PID control parameters according to the adaptive weight, and determine a control gain according to the weighted PID control parameters; weight the control gain by the penalty factor, and update the next-round functional electrical stimulation control signal according to the weighted control gain.

[0035] The present invention provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the above-mentioned closed-loop feedback functional electrical stimulation control method is implemented.

[0036] The present invention provides a computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, the above-mentioned closed-loop feedback functional electrical stimulation control method is implemented.

[0037] The above at least one technical solution adopted by the present invention can achieve the following beneficial effects:

[0038] The present invention obtains the target motion angle of the healthy limb of the current patient and the actual motion angle of the affected limb of the patient under the current functional electrical stimulation control signal, and determines the error between the target motion angle and the actual motion angle and the change rate of the error; then determines the next round of PID control parameters corresponding to the current error and the change rate of the error based on fuzzy control; and determines the adaptive weight of the next round of PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold, and determines an exponent according to the absolute value of the current error to determine a penalty factor with exponential growth, so as to weight the PID control parameters according to the adaptive weight, and determine the control gain according to the weighted PID control parameters; weight the control gain by the penalty factor, and update the next round of functional electrical stimulation control signal according to the weighted control gain.

[0039] The present invention adopts an adaptive weight fuzzy ILC closed-loop feedback control algorithm to accurately adjust control parameters through learning historical control experience, enabling the affected limb of the patient to quickly approach the target motion trajectory of the healthy limb of the patient, reducing the number of iterative learning times of the ILC control algorithm, improving the real-time performance of limb movement reconstruction, improving the target trajectory tracking accuracy, and improving the completion degree of the movement of the affected limb of the patient. Brief Description of the Drawings

[0040] The drawings described herein are used to provide a further understanding of the present invention, form a part of the present invention, and the schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0041] Figure 1 It is a schematic flowchart of a closed-loop feedback functional electrical stimulation control method provided by the present invention;

[0042] Figure 2 It is a schematic block diagram of the application of an adaptive weight fuzzy AW-FILC closed-loop feedback control algorithm provided by the present invention;

[0043] Figure 3 It is a schematic block diagram of a closed-loop feedback control algorithm based on adaptive weight fuzzy ILC provided by the present invention;

[0044] Figure 4 It is a schematic diagram of a triangular membership function provided by the present invention;

[0045] Figure 5 It is a schematic flowchart of an adaptive weight fuzzy ILC provided by the present invention;

[0046] Figure 6 It is a schematic diagram of the tracking effect of an adaptive weight fuzzy iterative learning control algorithm provided by the present invention under a step function motion curve;

[0047] Figure 7 Schematic diagram of the tracking effect of an adaptive weight fuzzy iterative learning control algorithm provided by the present invention under a triangular wave function motion curve;

[0048] Figure 8 Schematic diagram of the tracking effect of an adaptive weight fuzzy ILC algorithm provided by the present invention under a pronation wrist flexion motion curve;

[0049] Figure 9 Schematic diagram of a closed-loop feedback functional electrical stimulation control device provided by the present invention;

[0050] Figure 10 Schematic diagram of a computer device for implementing a closed-loop feedback functional electrical stimulation control method provided by the present invention. Detailed implementation manners

[0051] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0052] At present, model-free closed-loop feedback control strategies have shown high rehabilitation training quality in the application of functional electrical stimulation (FES), mainly by dynamically adjusting electrical stimulation parameters through real-time acquired surface electromyogram signals (sEMG) or predicted muscle responses. Specific methods include using long short-term memory (LSTM) neural networks to predict surface electromyogram signals and adjust stimulation intensity, optimizing stimulation using surface electromyogram signals to promote walking coordination, using inertial measurement units (IMU) to accurately test gait parameters to adjust surface electromyogram signals to optimize the FES effect, and intelligently adjusting parameters through predictive control and rolling optimization algorithms to match the motion patterns of healthy limbs.

[0053] In the model-based closed-loop feedback control strategy, the constructed muscle electrical stimulation response relationship model is used to predict the actual motion trajectories, including torque and angle, etc. The closed-loop feedback controller modulates the electrical stimulation parameters according to the deviation between the actual and target motion trajectories to ensure that the model output accurately tracks the target trajectory. Commonly used muscle response models include the Hill model, Hammerstein model, Neuro-fuzzy model, and empirical models, etc. Among them, the Hammerstein model is widely popular due to its simple structure and high computational efficiency, and is particularly suitable for predicting the isometric contraction muscle dynamics induced by FES. Researchers have also proposed various improvement methods, such as the adaptive Wiener-Hammerstein model, online parameter adjustment using the Kalman filter, and the adaptive Hammerstein model based on the variable forgetting factor recursive least squares method, to improve the accuracy and adaptability of the model. In addition, considering the inter-subject variability and the complexity of the musculoskeletal system, the influence of electrode position on motion is also taken into account in the model, such as through the Neuro-fuzzy modeling method using the Recurrent fuzzy neural network (RFNN).

[0054] The empirical model is also a commonly used tool by researchers. It does not consider the physiological structure of the muscle, but focuses on the accuracy and practicality of the prediction output. Scholars have adopted various empirical models, such as linear models, NARX-RNN models, and nonlinear least squares fitting models, etc., to characterize the relationship between electrical stimulation and muscle response in the FES system.

[0055] The electrical stimulation muscle response model is crucial for designing high-performance closed-loop feedback control algorithms. These algorithms will ultimately be deployed in an embedded rehabilitation training system to assist the rehabilitation training of paralyzed patients. An accurate model helps improve the completion of rehabilitation training actions and the functional recovery of paralyzed limbs. The feedback controllers used in the research include PID controllers, Gain Scheduling Controllers (GSCs), fuzzy logic controllers, Sliding Mode Controllers (SMCs), Adaptive Neuro-Fuzzy Inference Systems (ANFISs), and iterative learning control, etc. The traditional type-1 PID controller performs poorly in dealing with nonlinear effects in FES applications, while SMC control can provide fast adjustment and optimal tracking performance but is vulnerable to chattering. It is challenging to select appropriate scheduling variables for GSC, which needs to be repeated daily to adapt to muscle response changes. The Fuzzy Logic Controller (FLC) can effectively handle nonlinear systems and provide solutions for time-varying response engineering systems. The improved FLC and parameter estimation methods can determine personalized FLC parameter values using information on an individual's input-output relationship, effectively addressing the challenges posed by individual differences and tracking the target motion trajectory with an acceptable tracking error.

[0056] ILC is a control strategy that utilizes task repeatability to enhance system performance, and is particularly suitable for the combination in the field of rehabilitation engineering and FES technology. ILC updates the control signal by using the data from the previous control cycle to reduce the tracking error and improve the tracking accuracy of the motion trajectory. This strategy can effectively compensate for external disturbances and maintain high-precision trajectory tracking when there are variations in the controlled object. Research shows that ILC is effective in complex muscle activation models, can reject disturbances and time-varying uncertainties in arm dynamics, and achieve precise control. In addition, ILC can also adjust the stimulation parameters, avoid over-stimulation of muscles, and learn the previous motion trajectory to control the affected limb to achieve the desired motion. However, ILC faces challenges in dealing with non-linear systems and non-repetitive disturbances. To address these problems, researchers have explored the combination of non-linear model predictive control (NMPC) and biomechanical state space models to achieve more precise trajectory tracking and muscle drive. In addition, to improve the applicability and efficiency of ILC, researchers have proposed various improvement methods, such as the estimation-based multiple model ILC (EMMILC) adaptive control scheme, and the application of EMMILC to electrode array control, which have significantly improved the usability of the system. There are also researchers who use the adaptive network-based fuzzy inference system (ANFIS) to establish a joint muscle model and combine it with a PID-type ILC controller to adjust the FES electrical stimulation parameters to improve the control accuracy of trajectory tracking in rehabilitation training.

[0057] The FES control strategy based on closed-loop feedback can significantly improve the accuracy and fluency of paralyzed patients in performing rehabilitation training actions, thereby promoting the recovery of their lost limb functions. For this reason, scholars have conducted extensive and in-depth research on control algorithms and the relationship model between electrical stimulation and muscle response. The performance of the controller algorithm is crucial for ensuring that the patient's limb can execute a motion trajectory as smooth, coherent, and stable as the target limb under the action of electrical stimulation. Given that the rehabilitation training of the paralyzed limb of stroke patients is essentially a repetitive re-learning functional task, the ILC control algorithm may be a more suitable control paradigm. However, despite the many advantages of the ILC algorithm, challenges such as the real-time problem in the iterative learning process and the selection of the learning rate still need to be concerned about.

[0058] The following are some current functional electrical stimulation regulation methods:

[0059] A multi-modal information visualization functional electrical stimulation closed-loop regulation system and method, publication number (CN117563135A). The multi-modal information visualization functional electrical stimulation closed-loop regulation method includes the following steps: 1. Start the closed-loop visualization multi-target functional electrical stimulation rehabilitation program in the upper computer and log in to the user account; 2. Select the rehabilitation action, paste the stimulation patch and the myoelectric-inertial information acquisition device according to the visualization interface, and then enter the system adaptation step after completion; 3. The electrical stimulation system confirms smooth communication with the upper computer, performs device initial configuration, conducts preliminary tests and adjustments, and determines the initial stimulation parameters for training; 4. After initialization, the system will operate according to the selected rehabilitation action and stage. If it is the first use in this rehabilitation stage, it will enter the model training stage. For the recorded rehabilitation training process in this rehabilitation stage, the system will obtain the reinforcement learning model constructed from previous rehabilitation training and enter the rehabilitation training process; 5. During the rehabilitation process, the system will prompt the user through pictures, sounds, etc., guide the user to perform rehabilitation actions and correct themselves. At the same time, functional electrical stimulation is applied to make specific muscles contract according to the stimulation time sequence to assist the user in performing rehabilitation actions; 6. During the rehabilitation training process, the system will process and extract features from the collected myoelectric-inertial signals in real time, and quantify the evaluation results of the movement actions and stimulation effects, and input them into the functional electrical stimulation control module; 7. Each rehabilitation training process is refined into several rehabilitation sub-processes, with a ten-minute relaxation time left between sub-processes. At the same time, the electrical stimulation parameters will also be adjusted according to the output of the reinforcement learning model during the sub-process interval. The next sub-process will continue the basic closed-loop process of stimulation-evaluation-feedback-model update and output-modify stimulation parameters-re-stimulation; 8. Each rehabilitation training process will automatically terminate this training after reaching the preset time and number of training times, or after the system detects that the user's fatigue level is higher than the threshold.

[0060] Variable-length iterative learning control method for foot-drop functional electrical stimulation rehabilitation system, publication number (CN113786556B), which includes the following steps: 1. Convert the foot-drop functional electrical stimulation rehabilitation system operating repeatedly under variable batch lengths into a time series input-output lifting matrix model with equal batch lengths. 2. Aiming at the problem of variable batch lengths, an optimization iterative learning control design framework based on multi-set successive projection is proposed, and a causal feedforward plus feedback optimization iterative learning control algorithm suitable for the case of variable batch lengths is designed based on this framework. 3. Based on the multi-set successive projection framework, the convergence of the designed optimization iterative learning control algorithm is proved. 4. This method can solve the tracking control problem of the foot-drop functional electrical stimulation rehabilitation system under the condition of variable batch lengths, so as to achieve high-precision tracking of the desired trajectory.

[0061] Closed-loop control method and device for electromyogram signals based on functional electrical stimulation, publication number (CN116474264A), the method includes: obtaining a first electromyogram signal of the muscle tissue of the subject in a resting state; applying electrical stimulation with initial parameters to the muscle tissue, and obtaining a second electromyogram signal generated by the muscle tissue during the electrical stimulation, where the initial parameters are preset; performing feature analysis on the second electromyogram signal to obtain an electromyogram feature value; judging whether the muscle tissue reaches a fatigue state according to the electromyogram feature value and based on a pre-constructed muscle fatigue model; in response to the muscle tissue not reaching a fatigue state, adjusting the second electromyogram signal based on a feedback regulation strategy to perform electrical stimulation on the muscle tissue. This method adopts a more intuitive and vivid fatigue determination method, can more accurately achieve closed-loop feedback control stimulation of muscle tissue, complete the electrical stimulation control of muscle tissue, and provide technical support for subsequent rehabilitation training.

[0062] Closed-loop regulation method of functional electrical stimulation based on muscle activation degree and LSTM, publication number (CN115177864A), discloses a closed-loop regulation method of functional electrical stimulation combining muscle activation degree and deep learning, combines muscle activation degree analysis and the LSTM model in deep learning, designs and develops a closed-loop regulation method of functional electrical stimulation based on muscle activation degree and LSTM. This method can obtain the muscle state according to the real-time analysis of electromyogram signals and automatically learn appropriate functional electrical stimulation parameters, so that when the patient makes a fist movement on the healthy side, the functional electrical stimulation parameters can be automatically adjusted according to the change of muscle activation degree, making the grip strength of the affected side and the healthy side under functional electrical stimulation tend to be the same; and the LSTM model will continuously learn and optimize the output electrical stimulation parameters as the input data set increases, solving the problems that in the clinical treatment of functional electrical stimulation, the parameters cannot be adjusted in real time according to the user's muscle state, the parameter adjustment completely depends on experience, the patient participation is not high, and the patient cannot actively rehabilitate.

[0063] Condition feedback control method of an upper limb rehabilitation system for functional electrical stimulation, publication number (CN112546440B), this method is based on an upper limb muscle model established by an isometric recruitment curve and a linear activation dynamics, calculates the input signal of a non-linear compensator through a designed feed-forward controller containing an expected dynamic equation and a feed-forward compensator and a feedback controller composed of a proportional-integral-derivative controller; calculates the control quantity of the upper limb muscle model through the obtained input signal of the non-linear compensator and the designed non-linear compensator, and adjusts the change of the actuator in real time through the obtained control quantity of the upper limb muscle model. This method has the advantages of strong robustness, fast tracking performance, strong anti-interference performance, etc., and has a very practical application prospect.

[0064] A dual-closed-loop regulation system for hand function rehabilitation integrating a manipulator and functional electrical stimulation, with the publication number (CN113952614A), includes an electromyogram-inertial signal synchronous acquisition and data preprocessing module, a fine hand movement feature extraction and intention parsing module, a hand movement function monitoring and intelligent evaluation module, a wearable manipulator adaptive control module, a functional electrical stimulation dynamic regulation module, an information transmission and database module, and a wearable rehabilitation hand hardware. This method fully considers the patient's subjective intention and motor ability, integrates the passive rehabilitation of rehabilitation robots and the active rehabilitation advantages of neuromodulation, solves problems such as the simple training mode of existing hand rehabilitation strategies and systems, poor human-computer interaction and individual adaptability, and unsatisfactory rehabilitation training effects, and helps to accelerate the recovery of neural plasticity function and the rehabilitation process.

[0065] A feedback-type functional electrical stimulation system with multi-signal fusion, with the publication number (CN113058157B), includes: an information acquisition module for real-time acquisition of the patient's central nerve information and peripheral movement information during the patient's limb movement training process; a display operation module for selecting rehabilitation training methods, storing rehabilitation data information, and displaying evaluation results; an information processing and fusion module for processing and multi-modal synchronous fusion analysis of the central nerve information and peripheral movement information obtained by the information acquisition module; a multi-signal fusion FES control module for simultaneously fusing the central nerve information and peripheral movement evaluation information transmitted from the information processing and fusion module according to the movement mode selected in the display operation module, establishing a time-effective control model, and outputting an electrical stimulation control command according to the control model based on the change value; a multi-channel FES output module for controlling the stimulation parameters of each channel according to the electrical stimulation control command and outputting an electrical stimulation current.

[0066] A functional electrical stimulation instrument capable of adaptively adjusting the output intensity and its control method, with the publication number (CN113332597B), characterized in that it includes: a processor, and respectively connected to the processor: a Bluetooth communication module, an angular velocity acquisition module, a power supply module, a boost module, and a pulse output module; the power supply module is connected to the pulse output module through the boost module. This device is powered by a lithium battery, is small in size, light in weight, and is convenient to wear. It communicates with a smartphone through Bluetooth, so as to use the smartphone as a host computer to control the device and transmit data between the two on the smartphone, which is convenient to use. A mathematical model of the calf angular velocity signal and the electromyogram signal is established offline using a BP neural network with a simple network structure, and the real-time prediction of the tibialis anterior electromyogram signal during walking is realized on this device, and the output intensity modulation is performed according to the predicted electromyogram signal, and an electrical stimulation output intensity that meets the requirements of natural gait can be provided.

[0067] Closed-loop brain-controlled functional electrical stimulation system, publication number (CN113332597B), characterized in that the system comprises: a signal acquisition module for the cerebral motor cortex area, an information control module, a stimulator, a limb movement acquisition module, and a cerebral frontal lobe signal acquisition module; wherein, the information control module comprises: a signal preprocessing sub-module, a pattern recognition sub-module, and a command control sub-module. The technical solution provided by this system has the advantages of ensuring the accuracy of limb movement recognition and improving the rehabilitation treatment effect of paralyzed patients.

[0068] Programmable functional electrical stimulator with real-time feedback function, publication number (CN106345054A), comprising a main controller, a power supply module, and an analog constant current output module; the main controller, according to the instructions of the host computer, outputs a control signal to the analog constant current output module, and is also used to correct the control signal output to the analog constant current output module according to the current output by the analog constant current output module; the analog constant current output module, according to the control signal, outputs a current of a corresponding waveform; the power supply module provides a working voltage for the main controller and the analog constant current output module. This functional electrical stimulator can receive control instructions sent by the host computer in real time to change the electrical stimulation waveform, and can receive feedback signals; and provides a variety of stimulation modes and stimulation pulse selections, which can meet the diverse requirements.

[0069] An adjustable functional electrical stimulation control method based on multi-modal fusion feedback, publication number (CN110420383A), specifically includes: Step 1, set the working mode of the electrode, and set the working mode of the electrode to one of the following modes: Mode A, only perform signal acquisition; Mode B, both perform electrical stimulation output and signal acquisition; Mode C, only perform electrical stimulation output; Step 2, set the electrical stimulation output mode; Step 3, set the signal acquisition mode. This method provides multi-modal sensing information and combines high-density electrode electrical stimulation. The electrical stimulation output can feedback the sensing information to the human body, and the multi-modal sensing information can also be used as the feedback of the electrical stimulation output, forming a closed-loop signal acquisition - electrical stimulation process to stabilize the electrical stimulation effect.

[0070] A functional electrical stimulation fuzzy control method, publication number (CN102319482A), relates to the technical field of rehabilitation medical devices for the disabled. It multi-channel collects surface electromyography signals, filters them through a high-pass filter to remove low-frequency noise caused by electrode sliding, filters out power frequency interference through fast independent component analysis, performs time-domain analysis on the surface electromyography signals after the second filtering to obtain the root mean square, takes the error between the output root mean square matrix value and the expected root mean square matrix value and the error change rate as the inputs of the fuzzy controller, and takes the precise value of the stimulation current intensity as the output. It converts the error and the error change rate into numerical values in the universe of discourse according to an appropriate ratio to determine the membership function, uses the control rules and the membership function to perform inference processing on the error and the error change rate to obtain the fuzzy quantity of the stimulation current intensity, and performs defuzzification processing to obtain the precise value of the stimulation current intensity. It can effectively improve the accuracy and stability of the FES system.

[0071] A functional electrical stimulation closed-loop fuzzy PID control method, publication number (CN102488964A), obtains the expression of the knee joint torque, establishes a muscle model, adjusts the magnitude of the stimulation current through a fuzzy PID controller, and obtains the actually output knee joint torque value through the muscle model, obtains the error and the error change rate, inputs the error, the error change rate, and the stimulation current into a fuzzy inference system, and the fuzzy inference system processes and converts them into corresponding fuzzy quantities, obtains the control rules, tunes the three parameters of the fuzzy PID controller according to the control rules, adjusts the magnitude of the stimulation current according to the tuned parameters, and obtains the newly actually output knee joint torque value through the muscle model until the error between the newly actually output knee joint torque value and the expected knee joint torque value is less than the threshold, and the process ends. This method can effectively improve the stability, accuracy, and stability of the functional electrical stimulator and obtain considerable social and economic benefits.

[0072] An adaptive neuro-fuzzy muscle modeling method under functional electrical stimulation, publication number (CN102521508B), collects the knee joint angle parameters and acceleration parameters during calf movement, obtains the expression of the knee joint torque through inverse dynamics derivation, inputs the real knee joint torque value into an adaptive neuro-fuzzy inference system to obtain the actually output knee joint torque value, inputs the error, the error change rate, and the stimulation current into the adaptive neuro-fuzzy inference system to be converted into corresponding fuzzy quantities, obtains the control rules through the corresponding fuzzy quantities to synthesize the corresponding stimulation current, simultaneously trains the neural network through the error and the error change rate to obtain the membership function parameters and the membership function structure, and adjusts the adaptive neuro-fuzzy inference system until the error is less than the threshold, and the process ends. The method provided by this method makes the error and the error rate between the actually output knee joint torque value and the real value small, and accurately measures the knee joint torque value.

[0073] Functional electrical stimulation joint angle genetic fuzzy control method, publication number (CN101846977A). To achieve accurate, stable, and real-time control of the current mode of the FES system and effectively improve the accuracy and stability of the FES system, the technical solution adopted by this method is as follows: First, determine the quantization factor, proportional factor, and membership function parameters of fuzzy control; Second, select the appropriate final evolution generation G, crossover probability Pc, and mutation probability Pm of the genetic algorithm; Through genetic algorithm optimization, reach the optimal state and obtain the decision variable kfuzzi of fuzzy control; After calculating the system output and its deviation from the muscle model under the new fuzzy control parameters, then enter the next step of the genetic algorithm to adaptively adjust the parameters of the fuzzy controller. Repeat this process until the parameters of the fuzzy controller are adaptively tuned online and used in the FES system. This method is mainly applied to functional electrical stimulation joint angle genetic fuzzy control.

[0074] An upper limb rehabilitation functional electrical stimulation closed-loop control method based on voluntary will, publication number (CN106693178A), includes the following two parts: 1. Judgment of the start of electrical stimulation: Determine the start and stop moments of electrical stimulation by collecting the voluntary electromyographic signals of the biceps brachii and triceps brachii; 2. Closed-loop control of the electrical stimulation intensity: Measure the actual elbow joint movement angular velocity as the feedback quantity and compare it with the desired angular velocity, and use a control method combining PID and inverse dynamics model to adjust the magnitude of the stimulation current S in real time. This method has a very strong initiative for the rehabilitation training of patients and is beneficial to the rehabilitation of hemiplegic or paralyzed patients.

[0075] Functional electrical stimulation PID parameter dual-source feature fusion microparticle swarm tuning method, publication number (CN101816822B). A dual-source feature fusion chaotic microparticle swarm tuning method for PID parameters in functional electrical stimulation is provided, which can accurately, stably, and real-time control the current intensity of the FES system and effectively improve the accuracy and stability of the FES system. The technical solution adopted by this method is as follows: First, use the handle reaction vector HRV during the walking process to predict the knee joint angle; Second, use the chaotic particle swarm algorithm to tune the proportional integral derivative PID parameters and real-time control the FES current level intensity. Finally, achieve the adaptive online tuning of the proportional integral derivative PID control parameters and use them in the functional electrical stimulation FES system. This method is mainly applied to tuning the PID parameters in functional electrical stimulation.

[0076] A functional electrical stimulation closed-loop control system and method with electromyogram signal feedback, publication number (CN105031812A). First, after the functional electrical stimulator completes initialization and signal synchronization, electrical stimulation is started on the controlled object; the electromyogram signal collector collects the original electromyogram signals generated when the muscle is electrically stimulated and performs preprocessing; then a muscle contraction and relaxation model is established using a Hammerstein model with a time-delay term, and parameter identification is performed using the Kalman filtering method; finally, online real-time prediction is performed on the muscle contraction and relaxation model, the control quantity of the optimal electrical pulse width is calculated, and it is fed back to the functional electrical stimulator to update its parameters, realizing real-time adaptive control. This method completes the closed-loop control adjustment of the number of electrical stimulation pulses by calculating the absolute average amplitude of the electromyogram signals of the controlled object; improves the control accuracy of the closed-loop functional electrical stimulation system, and realizes the adaptive control of the functional electrical stimulation with electromyogram signal feedback.

[0077] A functional electrical stimulation exercise rehabilitation system and method integrating manual and autonomous control, publication number (CN104971433B), mainly consists of a closed-loop control system composed of three modules: a feedforward controller, a PID feedback controller, and an impedance controller. Among them, the feedforward controller generates a functional electrical stimulation intensity in a timely manner according to the set joint angle to ensure the response speed of the control system; the PID feedback controller ensures the anti-interference ability of the system; the impedance controller obtains a joint angle based on the subject's autonomous force and adds it to the originally set joint angle to realize the collaborative work of the autonomous force and the functional electrical stimulation. Through this control mode, this system can integrate the subject's autonomous movement awareness into the rehabilitation training and realize precise movement training. This system can improve the rehabilitation effect of the existing functional electrical stimulation.

[0078] An adaptive electrical stimulation balance rehabilitation training system, publication number (CN117899358B), realizes disturbance-oriented adaptive electrical stimulation by collecting the joint angles of the subject during balance maintenance on a multi-degree-of-freedom platform and controlling the functional electrical stimulation (FES) parameters based on the real-time joint angles. An adaptive electrical stimulation strategy is realized based on the patient's joint angles, which can help the patient maintain balance and provide simple and effective assistance for the rehabilitation treatment of stroke patients. Due to the clear technical effect, it provides a technical solution for the personalized and active rehabilitation treatment of patients.

[0079] An electrostimulation adaptive control method for an electrostimulation rehabilitation instrument, with the publication number (CN118203762B), comprises the following steps: 1. Improve the optimization mathematical model and optimization strategy of the Great Wall construction optimization algorithm; 2. Establish a functional electrostimulation control system model for the electrostimulation rehabilitation instrument; 3. Convert the adaptive control problem of the functional electrostimulation signal into an optimization mathematical model to be used as the optimization objective function of the improved Great Wall construction optimization algorithm; 4. Encode the parameter values of the incremental PID algorithm as the individual positions of the improved Great Wall construction optimization algorithm; 5. Update the individual positions through the objective function and the improved Great Wall construction optimization algorithm, and decode the individual position with the minimum objective function value into the optimal Kp, Ki, and Kd parameter values of the PID algorithm; 6. Input the optimal PID parameter values into the incremental PID algorithm module, and repeat steps 3 to 5 to achieve the adaptive control of the functional electrostimulation signal of the electrostimulation rehabilitation instrument.

[0080] The FES control strategy based on closed-loop feedback has been proven to be superior to the traditional open-loop controlled FES. According to whether an electrostimulation muscle response model is used during the design or application of the control algorithm, it can be divided into model-free closed-loop feedback control and model-based closed-loop feedback control.

[0081] In summary, the existing technologies in this field have the following defects:

[0082] The existing model-free closed-loop feedback FES control algorithms do not rely on pre-built models, so they may have limitations in dealing with physiological characteristics such as the nonlinearity and fatigue of muscle responses. This may lead to inaccurate adjustment of electrostimulation parameters and affect the rehabilitation effect. In addition, the performance of these algorithms highly depends on the quality and stability of the feedback signal, and any noise or interference in the signal may affect the control accuracy. Finally, due to the lack of in-depth understanding of muscle physiological characteristics, these algorithms may not be able to achieve the optimal stimulation strategy, thus limiting the further improvement of the rehabilitation effect.

[0083] Aiming at the limitations of the ILC algorithm, based on previous work, the present invention proposes an iterative learning control (AW-FILC) algorithm based on adaptive weight fuzzy logic. This algorithm combines the flexibility of fuzzy logic control and the self-learning ability of iterative learning control (ILC). By introducing a dynamic adjustment mechanism, it can adjust the proportional, integral, and differential (PID) gains in real time, thereby enhancing the system's adaptability to muscle nonlinear characteristics and complex dynamic environments. In addition, a weight factor based on the error is introduced, which dynamically adjusts the control input according to the magnitude of the current error. Specifically, when the error is large, the penalty factor increases the learning gain to accelerate the system convergence; when the error is small, the learning gain is reduced to reduce over-regulation and maintain the control stability. This mechanism effectively optimizes the control effect and especially improves the robustness and accuracy of the system under non-ideal conditions.

[0084] Combined with adaptive weight fuzzy logic inference and ILC, it can effectively compensate for the interference and muscle time-varying uncertainty during the rehabilitation training process, and enhance the control accuracy of the controlled object by iteratively learning historical experience. The algorithm aims to reduce the number of iterations of the control algorithm while ensuring that the controlled limb can execute a motion trajectory as similar as possible to that of the healthy limb. Adaptive weight fuzzy control is selected because it can enhance the flexibility and robustness of the control system, while PID control provides a solid foundation for the system with its excellent stability and dynamic response performance. In the ILC algorithm, the choice of learning rate has a decisive impact on the system performance. Therefore, in view of the nonlinear characteristics of muscle contraction under electrical stimulation, this invention selects an ILC algorithm based on the PID-type learning rate and combines it with adaptive weight fuzzy logic inference to achieve the adaptive adjustment of the controller learning rate gain, thereby enhancing the robustness of the algorithm. Considering that directly adopting existing muscle response models such as Hammerstein may face challenges such as individual differences, model applicability, model accuracy, and parameter adjustment, this invention uses the electrical stimulation parameters and torque data obtained from preliminary experiments to construct a more practical muscle response relationship model. This model provides strong support for the performance testing of the control algorithm.

[0085] Compared with the control algorithms used in the above patents, the adaptive weight fuzzy used in this invention has the advantages of stronger modeling ability, being able to handle uncertainties within and between individuals simultaneously; faster ILC convergence speed, adopting an aggressive control algorithm when the error is large and a stable control algorithm when the error is small; better control effect, especially the control surface near the steady state is smoother, which helps to improve the robustness and suppress oscillations; wider adaptability, being able to adapt to complex nonlinear systems and automatically adjust control parameters to cope with system changes and external disturbances; and combining the advantages of fuzzy control and PID control, being able to maintain good control performance in the face of system parameter changes and external disturbances.

[0086] The main purpose of this invention is to develop and deploy a closed-loop feedback-based FES control algorithm to the FES prototype system designed in the research to assist paralyzed patients in reconstructing smooth, stable, and coherent rehabilitation training movements. The system calculates the motion trajectory error between the controlled limb and the healthy limb in real time, and uses the controller algorithm to dynamically adjust the output electrical stimulation intensity, thereby achieving the stability and smoothness of the motion trajectory of the controlled limb and making it as similar as possible to the healthy limb.

[0087] The following will detail the technical solutions provided by each embodiment of this invention in conjunction with the accompanying drawings.

[0088] Figure 1 It is a schematic flowchart of a closed-loop feedback functional electrical stimulation control method in this invention, specifically including the following steps:

[0089] S101: Obtain the target motion angle of the healthy limb of the current patient and the actual motion angle of the affected limb of the patient under the current functional electrical stimulation control signal, and determine the error between the target motion angle and the actual motion angle and the change rate of the error.

[0090] S102: Fuzzify the current error and the change rate of the error, and perform fuzzy inference according to the preset fuzzy control rules to determine the next-round PID control parameters corresponding to the current error and the change rate of the error.

[0091] S103: Determine the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold, and the adaptive weight is negatively correlated with the ratio.

[0092] S104: Determine an exponent according to the absolute value of the current error to determine a penalty factor with exponential growth, and the penalty factor is positively correlated with the absolute value of the error.

[0093] S105: Weight the PID control parameters according to the adaptive weight, and determine the control gain according to the weighted PID control parameters; weight the control gain by the penalty factor, and update the next-round functional electrical stimulation control signal according to the weighted control gain.

[0094] The server mentioned in the present invention may be a server set up on a service platform, or a device such as a desktop computer or a laptop computer that can execute the solution of the present invention. For the convenience of description, only the server is used as the execution subject for description below.

[0095] The purpose of implementing the present invention is to provide a closed-loop feedback functional electrical stimulation (FES) control strategy adaptive weight fuzzy AW-FILC for reconstructing the motor function of hemiplegic limbs after stroke. As Figure 2 shown, Figure 2 is a schematic diagram of the application block diagram of an adaptive weight fuzzy AW-FILC closed-loop feedback control algorithm in the present invention. This control strategy can be applied to the FES mode regulated by electromyogram signals (i.e., electromyogram-controlled FES, as shown by the dotted line box in the figure), and can also be applied to FES applications that only rely on kinematic sensor feedback.

[0096] First, an electrical stimulation torque muscle response relationship model can be obtained by fitting based on electrical stimulation torque human experimental data using the nonlinear least squares method as the controlled object for the performance simulation test of the control algorithm. The controlled object for the simulation test can also be other existing electrical stimulation muscle response relationship models.

[0097] The present invention proposes an adaptive weight fuzzy inference, combined with ILC having a PID-type learning rate gain, enabling the control system to automatically adjust control parameters according to the current system state and performance, improving the robustness and convergence speed of the algorithm.

[0098] The present invention uses a mathematical relationship model of the controlled object constructed from the electrical stimulation parameters (including pulse width and stimulation frequency) and corresponding muscle response data sourced from human experiments for testing the control performance of the algorithm. During the modeling process, first, normalization processing is performed on the original experimental data to eliminate the dimension difference. Subsequently, the non-linear least squares method is used to accurately fit the data to construct a mathematical model that can accurately capture and describe the non-linear relationship between the electrical stimulation parameters and muscle response: T = aW b + cF d .

[0099] Among them, T, W, and F respectively represent the normalized wrist torque, pulse width, and stimulation frequency. The model parameters are a = 0.7503, b = 1, c = 0.2154, d = 1. Based on the in-depth analysis and experimental verification of the wrist extension movement, it is ensured that the transfer function model can accurately reflect the response behavior of the wrist muscles during the execution of the extension movement.

[0100] For the adaptive weight fuzzy AW-FILC closed-loop feedback FES control strategy, as Figure 3 , Figure 3 is a schematic diagram of a closed-loop feedback control algorithm block based on adaptive weight fuzzy ILC in the present invention.

[0101] The target movement angle y of the healthy limb of the patient 目标 is sourced from the inertial measurement unit IMU1 worn on the healthy limb, while the actual movement angle y of the affected limb of the patient under the current functional electrical stimulation control signal 实际 is provided by the inertial measurement unit IMU2 on the affected limb. These two sets of data are respectively input into the adaptive weight fuzzy logic inference system and the ILC controller for processing and calculation to generate the control output u(t). The output u(t) dynamically adjusts the stimulation intensity through the electrical stimulation parameters (modulating pulse width), thereby achieving precise adjustment of the movement angle of the affected limb.

[0102] In the present invention, the adaptive weight fuzzy logic inference plays a key role. It dynamically adjusts the learning rate gain of the ILC controller according to the error e between y 目标 and y 实际 and the change rate de / dt of the error. In particular, since the ILC controller uses a PID-type learning rate, the output of the adaptive weight fuzzy logic inference covers three control parameters kp, ki, and kd, thus realizing the adaptive tuning of the controller parameters and enhancing the flexibility and performance of the control system.

[0103] Adaptive weight fuzzy logic inference includes five parts: fuzzification, fuzzy inference, defuzzification, dynamic adjustment of adaptive weights, and adjustment of the penalty factor to the learning rate. The specific process is as follows:

[0104] (1) Fuzzification

[0105] To implement fuzzification processing, the present invention first determines the universe of discourse and the membership function to map the current error and the rate of change of the error onto a preset universe of discourse; according to the membership function of the fuzzy set corresponding to different error degrees on the universe of discourse, the membership degrees of the current error and the rate of change of the error belonging to different fuzzy sets are determined.

[0106] Specifically, in one or more embodiments of the present invention, in view of the different numerical ranges of the error e and its rate of change de / dt, as well as the output variables kp, ki, and kd, the present invention maps the actual values onto a unified universe of discourse for subsequent processing and analysis. For example, the universe of discourse can be set to [-3, 3], where the endpoint -3 represents NB (negative large), -2 represents NM (negative medium), -1 represents NS (negative small), 0 represents ZO (zero), 1 represents PS (positive small), 2 represents PM (positive medium), and 3 represents PB (positive large). For example, when the range of the input e is [-80, 80] and the value of e is 40, the value obtained through mapping is 1.5, which is in the interval [1, 2], indicating that this point is between positive small (PS) and positive medium (PM). This mapping method helps to convert specific numerical values into fuzzy semantics that can be processed by the fuzzy logic system, thereby realizing the adaptive adjustment of the control system parameters. The mapping is carried out by the following formula:

[0107] where, E max is the maximum value 3 of the universe of discourse, E min is the minimum value -3 of the universe of discourse, M is the maximum value 80 of the actual value of e, and N is the minimum value -80 of the actual value of e.

[0108] In one or more embodiments of the present invention, a triangular membership function is used to describe the fuzzy characteristics of the input variables, as Figure 4 shown, Figure 4 is a schematic diagram of a triangular membership function in the present invention. When the value of the input variable e after mapping is 1.5, this value belongs to both the fuzzy sets PM (positive medium) and PS (positive small) at the same time. In this case, the membership degree u(1.5) corresponding to the value of e is 0.5.

[0109] After that, according to the preset fuzzy control rules and the membership degrees of the current error and the change rate of the error belonging to different fuzzy sets, the membership degrees of the current error and the change rate of the error to each activated fuzzy control rule can be determined; thus, defuzzification is performed on the membership degrees of the current error and the change rate of the error to each activated fuzzy control rule to obtain the next-round PID control parameters corresponding to the current error and the change rate of the error.

[0110] (2) Fuzzy inference

[0111] Fuzzy inference is to determine the fuzzy state of the output variable by looking up a table according to the membership degree values of the input variables and their change rates. When the input variable is 40 and the change rate of the error is -10, the fuzzified value of the error is 1.5, with a membership degree of 0.5, belonging to the fuzzy sets PS (Positive Small) and PM (Positive Medium). The fuzzified value of the change rate of the error is -1, with a membership degree of 1, belonging to the fuzzy set NS (Negative Small). According to these membership degree values, the corresponding control gain adjustment values are looked up from the preset fuzzy rule table. Through fuzzy inference, the system can generate the fuzzy output value of the controller gain adjustment according to the input error and the change rate of the error, providing a basis for the subsequent defuzzification process.

[0112] (3) Defuzzification

[0113] In the fuzzification process, the values of ZO (Zero), NS (Negative Small), and PS (Positive Small) are set to 0, -1, and 1 respectively. Through membership degree calculation, the expected values of kp, ki, and kd are obtained as E(kp) = -1.5, E(ki) = 1.5, and E(kd) = -1.5 respectively. Considering that the actual value ranges of kp and ki are [-0.5, 0.5], while the actual value range of kd is [-0.01, 0.01], the interval mapping formula can be used to convert the expected values into actual values. Accordingly, the actual value of kp is approximately -0.25, the actual value of ki is approximately 0.25, and the actual value of kd is approximately -0.005. This conversion process ensures that the results of fuzzy logic inference can be effectively applied to the parameter adjustment of the actual control system.

[0114] (4) Dynamic adjustment of adaptive weights

[0115] In AW-FILC, the design of the adaptive weight aims to automatically adjust the controller gain according to the magnitude of the error, so as to achieve more efficient learning and stable control response. When the error is large, the system reduces the learning gain to avoid over-adjustment and ensure the stability of the system; when the error is small, the learning gain is increased to accelerate convergence.

[0116] The adaptive weight of the next-round PID control parameters can be determined by the following formula according to the ratio between the absolute value of the current error and the preset maximum error threshold:

[0117]

[0118] where α k (n) is the adaptive weight of the PID control parameter in the nth iteration, β and γ are custom parameters based on the controlled model, and e max is the maximum threshold of the error.

[0119] When the error is large, the weight α k (n) will decrease, thus avoiding overreaction of the system. When the error is small, the weight α k (n) will increase, thus accelerating the convergence speed. The final control output is the result calculated according to the adjusted controller gain and is used to update the control signal of the system. Finally, α k (n) is multiplied by kp[n], ki[n], and kd[n] respectively to obtain the corresponding coefficients.

[0120] (5) Penalty factor adjusts the learning rate

[0121] To further improve the convergence speed of the control algorithm, the present invention introduces a penalty factor ρ. This penalty factor is dynamically updated according to the absolute value of the current error and increases the learning rate adjustment when the error is large through exponential growth, thereby accelerating the convergence process. When the convergence is stable in the later stage, the influence of the penalty factor is reduced to ensure the stability of system convergence. The exponential growth penalty factor can be determined according to the absolute value of the current error by the following formula:

[0122] u[i + 1, n] = u[i, n] + ρ·δ n

[0123] ρ = 1 + e (θ·|erro(n)|)

[0124] where ρ is the penalty factor, θ is a custom parameter based on the controlled model, and e(n) is the current error. δ n is the iterative learning compensation amount obtained according to the adaptive weight fuzzy mechanism before, and ρ is dynamically calculated according to the absolute value of the current error. Thus, when the error is large, ρ rises rapidly, thereby amplifying the correction amplitude of δ n , and accelerating the convergence speed; when the error decreases, ρ approaches 1, avoiding oscillations caused by excessive corrections, thereby ensuring the stability of later convergence.

[0125] In the adaptive weight fuzzy ILC control algorithm, the ILC controller adopts a PID-type learning rate:

[0126] u[i + 1, n] = u[i, n] + ρ(α k (n)kp[n]{e[n] - e[n - 1]} + αk (n)ki[n]e[n]+α k (n)kd[n]{e[n] - 2e[n - 1] + e[n - 2]})

[0127] Wherein, u[i,n] is the controller output of the nth node of the ith iteration trajectory; kp[n], ki[n], kd[n] are the learning rate gains of the nth node on the trajectory.

[0128] The adaptive weight fuzzy ILC algorithm adopted by the present invention controls the controlled object. The simulation process of this algorithm is as Figure 5 shown Figure 5 It is a schematic diagram of an adaptive weight fuzzy ILC process in the present invention.

[0129] Among them, the AWFuzzyILCcontroller function is responsible for performing adaptive weight fuzzy logic inference and ILC calculation. This function receives the error e(n) between the actual output and the target value of node n as an input parameter. First, according to the current error e(n) and the error e(n - 1) at the previous moment, the rate of change of the error is calculated Subsequently, using e(n) and perform adaptive weight fuzzy inference to determine the learning rate gains kp[n], ki[n], kd[n] of node n in the ith iteration. Finally, calculate the controller output u[i + 1,n] of the (i + 1)th iteration.

[0130] In the Matlab simulation environment, the present invention sets the trigonometric function curve, step function curve, and pronation wrist flexion motion curve as the desired target trajectory. To achieve precise control, the target trajectory is decomposed into 1000 nodes for iterative learning. During the simulation process, the present invention sets the value ranges of the proportional gain kp and the integral gain ki to [-0.5, 0.5], while the value range of the differential gain kd is [-0.01, 0.01]. The setting of these parameters aims to ensure the stability and performance optimization of the control system.

[0131] Under the step function target motion trajectory with 10 steps, each step amplitude of 7, and the duration of one step of 10 s, AW - FILC converges at the sixth iteration, the RMSE at convergence is 0.0005, and the time required for the algorithm to converge is 30 ms. The tracking effect of the adaptive weight fuzzy iterative learning control algorithm under the step function motion curve is as Figure 6As shown in the figure. Under the trigonometric function target motion trajectory with a frequency of 10 Hz and an amplitude of 10, the Adaptive Weighted Fuzzy Iterative Learning Control (AW-FILC) converges at the seventh iteration by adaptively adjusting the learning gain under this motion curve. The RMSE at convergence is 0.0007, and the time required for the algorithm to converge is 30 ms. The tracking effect of the Adaptive Weighted Fuzzy Iterative Learning Control algorithm under the triangular wave function motion curve is as Figure 7 shown. The actual pronation wrist flexion motion curve data is collected through an IMU, and the electrical stimulation parameters and the muscle response transfer function are used as the simulation model. The control algorithm uses the Adaptive Weighted Fuzzy Iterative Learning Control. The algorithm converges at the fifth iteration. The RMSE at convergence is 0.033, and the time required for convergence is 30 ms. The tracking effect of the Adaptive Weighted Fuzzy ILC algorithm under the pronation wrist flexion motion curve is as Figure 8 shown.

[0132] Based on Figure 1 the closed-loop feedback functional electrical stimulation control method shown in the figure, the present invention obtains the target motion angle of the healthy limb of the patient and the actual motion angle of the affected limb of the patient under the current functional electrical stimulation control signal, and determines the error between the target motion angle and the actual motion angle and the change rate of the error; then determines the next round of PID control parameters corresponding to the current error and the change rate of the error based on fuzzy control; and determines the adaptive weight of the next round of PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold, and determines the exponent according to the absolute value of the current error to determine the exponentially increasing penalty factor, so as to weight the PID control parameters according to the adaptive weight, and determine the control gain according to the weighted PID control parameters; weight the control gain by the penalty factor, and update the next round of functional electrical stimulation control signal according to the weighted control gain.

[0133] The present invention adopts the Adaptive Weighted Fuzzy ILC closed-loop feedback control algorithm to accurately adjust the control parameters through the learning of historical control experience, enabling the affected limb of the patient to quickly approach the target motion trajectory of the healthy limb of the patient, reducing the number of iterative learning times of the ILC control algorithm, improving the real-time performance of limb movement reconstruction, improving the target trajectory tracking accuracy, and improving the completion degree of the limb movement of the affected limb of the patient.

[0134] The present invention proposes an adaptive weight fuzzy ILC application to reduce the number of iterative learning times of the ILC control algorithm and improve the real-time performance of limb movement reconstruction. The proposed adaptive weight fuzzy AW-FILC closed-loop feedback control algorithm of the present invention will not cause a decrease in the correlation coefficient of limb movement trajectory reconstruction due to an increase in the number of repetitions of the rehabilitation training actions of the paralyzed limbs of stroke patients. Compared with the existing technical solutions, the present invention proposes a model-based closed-loop feedback control algorithm, which combines adaptive weight fuzzy logic inference PID and ILC, can effectively compensate for the interference and muscle time-varying uncertainties during the rehabilitation training process, and enhance the control accuracy of the controlled object through iterative learning historical experience.

[0135] The adaptive weight fuzzy PID used in the present invention has stronger modeling ability compared to the PID used in the above patents, and can handle both intra-individual and inter-individual uncertainties simultaneously; has better control effects, especially the control surface near the steady state is smoother, which helps to improve robustness and suppress oscillations; has a wider adaptability, can adapt to complex non-linear systems and automatically adjust control parameters to cope with system changes and external disturbances; and combines the advantages of fuzzy control and PID control, and can maintain good control performance in the face of system parameter changes and external disturbances.

[0136] When applying the closed-loop feedback functional electrical stimulation control method provided by the present invention, it is not necessary to execute according to Figure 1 the order of the steps shown. The specific execution order of each step can be determined according to needs, and the present invention does not limit this.

[0137] The above is the closed-loop feedback functional electrical stimulation control method provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding closed-loop feedback functional electrical stimulation control device, as Figure 9 shown.

[0138] Figure 9 The figure shows a schematic diagram of a closed-loop feedback functional electrical stimulation control device provided by the present invention, including:

[0139] An acquisition module 201, configured to acquire the target movement angle of the healthy limb of the current patient and the actual movement angle of the affected limb of the patient under the current functional electrical stimulation control signal, and determine the error between the target movement angle and the actual movement angle and the change rate of the error;

[0140] A fuzzy inference module 202, configured to fuzzify the current error and the change rate of the error, and perform fuzzy inference according to the preset fuzzy control rules to determine the next-round PID control parameters corresponding to the current error and the change rate of the error;

[0141] A weight determination module 203 is configured to determine an adaptive weight of PID control parameters for the next round according to a ratio between an absolute value of a current error and a preset maximum error threshold, and the adaptive weight is negatively correlated with the ratio.

[0142] A penalty factor determination module 204 is configured to determine an exponent according to an absolute value of a current error to determine a penalty factor with exponential growth, and the penalty factor is positively correlated with the absolute value of the error.

[0143] A weighted control module 205 is configured to weight the PID control parameters according to the adaptive weight, and determine a control gain according to the weighted PID control parameters; weight the control gain by the penalty factor, and update a functional electrical stimulation control signal for the next round according to the weighted control gain.

[0144] For specific limitations of the closed-loop feedback functional electrical stimulation control device, reference may be made to the limitations of the closed-loop feedback functional electrical stimulation control method in the foregoing text, which will not be elaborated herein. Each module in the above-mentioned closed-loop feedback functional electrical stimulation control device may be implemented in whole or in part by software, hardware, and their combination. The above-mentioned modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above-mentioned modules.

[0145] The present invention further provides a computer-readable storage medium storing a computer program, and the computer program can be used to execute the above Figure 1 provided closed-loop feedback functional electrical stimulation control method.

[0146] The present invention further provides Figure 10 a schematic structural diagram of the computer device shown in Figure 10 As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory. Of course, other hardware required for other services may also be included. The processor reads a corresponding computer program from the non-volatile memory into the memory and then runs it to implement the above Figure 1 provided closed-loop feedback functional electrical stimulation control method.

[0147] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above various methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided by the present invention can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0148] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope recorded by the present invention.

Claims

1. A closed-loop feedback functional electrical stimulation control method, characterized in that, Including: Obtain the target motion angle of the healthy limb of the current patient and the actual motion angle of the affected limb of the patient under the current functional electrical stimulation control signal, and determine the error between the target motion angle and the actual motion angle and the change rate of the error; Fuzzify the current error and the change rate of the error, and perform fuzzy inference according to the preset fuzzy control rules to determine the next-round PID control parameters corresponding to the current error and the change rate of the error; Determine the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold, and the adaptive weight is negatively correlated with the ratio; Determine an exponent based on the absolute value of the current error to determine an exponentially increasing penalty factor, and the penalty factor is positively correlated with the absolute value of the error; Weight the PID control parameters according to the adaptive weight, and determine the control gain according to the weighted PID control parameters; weight the control gain by the penalty factor, and update the next-round functional electrical stimulation control signal according to the weighted control gain.

2. The closed-loop feedback functional electrical stimulation control method according to claim 1, wherein, The fuzzifying the current error and the change rate of the error specifically includes: Map the current error and the change rate of the error to the preset universe of discourse; Determine the membership degrees of the current error and the change rate of the error belonging to different fuzzy sets according to the membership functions of the fuzzy sets corresponding to different error degrees on the universe of discourse.

3. The closed-loop feedback functional electrical stimulation control method according to claim 2, wherein The performing fuzzy inference according to the preset fuzzy control rules to determine the next-round PID control parameters corresponding to the current error and the change rate of the error specifically includes: Determine the membership degrees of the current error and the change rate of the error to each activated fuzzy control rule according to the preset fuzzy control rules and the membership degrees of the current error and the change rate of the error belonging to different fuzzy sets; Defuzzify the membership degrees of the current error and the change rate of the error to each activated fuzzy control rule to obtain the next-round PID control parameters corresponding to the current error and the change rate of the error.

4. The closed-loop feedback functional electrical stimulation control method according to claim 1, wherein, The determining the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold specifically includes: Determine the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and the preset maximum error threshold through the following formula: Where, α k (n) is the adaptive weight of the PID control parameter in the n-th iteration, β and γ are custom parameters based on the controlled model, e(n) is the current error, and e max is the maximum threshold of the error.

5. The closed-loop feedback functional electrical stimulation control method according to claim 1, wherein The determining an exponent based on the absolute value of the current error to determine an exponentially increasing penalty factor specifically includes: Determine an exponent based on the absolute value of the current error to determine an exponentially increasing penalty factor through the following formula: ρ = 1 + e (θ·|e(n)|) ; Where ρ is the penalty factor, θ is a custom parameter based on the controlled model, and e(n) is the current error.

6. A closed-loop feedback functional electrical stimulation control device, characterized in that, Including: An obtaining module, configured to obtain the target motion angle of the healthy limb of the current patient and the actual motion angle of the affected limb of the patient under the current functional electrical stimulation control signal, and determine the error between the target motion angle and the actual motion angle and the change rate of the error; A fuzzy inference module, which is used to fuzzify the current error and the change rate of the error, and perform fuzzy inference according to preset fuzzy control rules to determine the next-round PID control parameters corresponding to the current error and the change rate of the error; A weight determination module, which is used to determine the adaptive weight of the next-round PID control parameters according to the ratio between the absolute value of the current error and a preset maximum error threshold, and the adaptive weight is negatively correlated with the ratio; A penalty factor determination module, which is used to determine an exponent according to the absolute value of the current error to determine a penalty factor with exponential growth, and the penalty factor is positively correlated with the absolute value of the error; A weighted control module, which is used to weight the PID control parameters according to the adaptive weight, and determine a control gain according to the weighted PID control parameters; weight the control gain by the penalty factor, and update the next-round functional electrical stimulation control signal according to the weighted control gain.

7. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

8. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

Images