Anti-overload protection device based on shoulder joint rehabilitation exoskeleton

Abstract

This invention provides an overload protection device based on a shoulder joint rehabilitation exoskeleton. Utilizing the EKF dynamic overload prediction algorithm, it can identify overload risks 200ms in advance, upgrading "post-event trigger protection" to "pre-event prediction protection," completely avoiding secondary damage to the patient's shoulder joint caused by overload traction. Through a fuzzy PID algorithm, the overload threshold is dynamically adjusted based on the patient's individual differences, rehabilitation stage, training mode, and real-time movement status, replacing the fixed threshold of existing technologies. This improves the accuracy of overload judgment, avoiding protection failure due to excessively high thresholds and solving the problem of excessively low thresholds interfering with normal training. Magnetorheological damping buffers achieve flexible absorption of overload impact energy, eliminating rigid impacts during the protection process and preventing secondary damage, while maximizing the continuity of rehabilitation training without affecting rehabilitation outcomes. It can cover the entire rehabilitation cycle from the acute phase to the sequelae phase, maximizing the effectiveness of rehabilitation training.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation training technology, and in particular to an overload protection device based on a shoulder joint rehabilitation exoskeleton. Background Technology

[0002] Shoulder dysfunction caused by stroke, post-rotator cuff injury surgery, and periarthritis of the shoulder has become a common condition in clinical rehabilitation. Data shows that the incidence of shoulder dysfunction after stroke in my country exceeds 70%. Post-rotator cuff injury patients require continuous and standardized rehabilitation training to restore joint range of motion and muscle strength. Shoulder rehabilitation exoskeleton robots, with their precise motion trajectory control and quantified training parameter management, have become core equipment for shoulder rehabilitation. Mature devices such as the ARAMIS wearable shoulder rehabilitation exoskeleton and the dual-arm collaborative shoulder rehabilitation exoskeleton have achieved three degrees of freedom in shoulder flexion / extension, abduction / adduction, and internal / external rotation, adapting to the rehabilitation needs at different stages of clinical practice.

[0003] However, existing shoulder joint rehabilitation exoskeletons have significant shortcomings in overload protection technology during clinical application, with the following key issues: First, existing overload protection systems mostly use a fixed force threshold triggering mode, which cannot adapt to individual differences in patients' age, muscle strength level, and rehabilitation stage. This can easily lead to problems such as protection failure due to excessively high thresholds or interference with normal rehabilitation training due to excessively low thresholds. Second, the protection mechanisms are all triggered after the event, and the protection action is only executed after the load exceeds the threshold. There is a control lag of more than 100ms, which makes it impossible to predict the overload risk and easily causes secondary traction injury to the patient's shoulder joint. Third, the protection execution method is singular, and the vast majority of protection logic is based on emergency braking without any buffer transition. The rigid impact at the moment of emergency stop can easily aggravate joint damage to patients, while interrupting the rehabilitation training process and affecting the rehabilitation effect. Fourth, the overload monitoring is limited to a single dimension, relying solely on end force sensors to collect load data without combining shoulder joint kinematics, the patient's active movement intentions, and joint dynamic characteristics for comprehensive judgment. The overload misjudgment rate exceeds 15%, resulting in poor clinical applicability. Fifth, the protection mechanism is disconnected from the rehabilitation training process, making it impossible to dynamically adjust the protection strategy according to different modes such as passive training, active assisted training, and resistance training, thus failing to meet the clinical needs of full-cycle rehabilitation. Summary of the Invention

[0004] To address the technical problems existing in the prior art, the present invention provides the following technical solution: An overload protection device based on a shoulder joint rehabilitation exoskeleton, characterized in that it comprises a wearable shoulder joint rehabilitation exoskeleton, an overload monitoring module, an edge computing control system, a multi-level buffer execution module, and a human-computer interaction terminal, wherein: The wearable shoulder joint rehabilitation exoskeleton is a three-degree-of-freedom driven structure, serving as the basic carrier of the device. The overload monitoring module is installed on the wearable shoulder joint rehabilitation exoskeleton, and its signal output end is communicatively connected to the input end of the edge computing control system to collect multi-source load and motion data in real time during the rehabilitation training process. The edge computing control system is bidirectionally connected to the native control system and the multi-level buffer execution module of the wearable shoulder joint rehabilitation exoskeleton, and is used for data processing, overload risk prediction and protection control command generation. The multi-level buffer execution module is coaxially mounted with the drive shaft of the wearable shoulder joint rehabilitation exoskeleton and is used to perform graded compliant overload protection actions. The human-computer interaction terminal is bidirectionally connected to the edge computing control system and is used for parameter configuration, data visualization, and human-computer interaction.

[0005] Preferably, the overload monitoring module includes a six-dimensional force / torque sensing unit, a joint angle sensing unit, a surface electromyography (EMG) signal acquisition unit, and an inertial measurement unit. The six-dimensional force / torque sensing unit is installed at the connection point between the upper arm binding component and the drive chain of the wearable shoulder joint rehabilitation exoskeleton, and is used to collect the interaction force and torque data of each degree of freedom of the shoulder joint in real time. The joint angle sensing unit is coaxially installed with the exoskeleton drive shaft and is used to collect joint angle, angular velocity, and angular acceleration data. The surface EMG signal acquisition unit uses dual-channel wireless EMG electrodes to collect EMG signals of the muscles around the patient's shoulder joint and identify active movement intentions. The inertial measurement unit is installed at the end of the upper arm binding component of the exoskeleton and is used to collect upper arm spatial posture and acceleration data.

[0006] Preferably, the edge computing control system integrates a data preprocessing module, an overload prediction module, an adaptive threshold tuning module, a protection execution control module, and a data storage module. The data preprocessing module performs filtering, noise reduction, data alignment, and normalization on the acquired raw data. The overload prediction module incorporates a shoulder joint dynamics overload prediction algorithm based on extended Kalman filtering to predict joint load change trends and assess overload risk levels. The adaptive threshold tuning module incorporates an adaptive overload threshold dynamic tuning algorithm based on fuzzy PID to dynamically update overload warning thresholds and protection trigger thresholds. The protection execution control module generates tiered protection control commands to drive the execution module to complete protection actions.

[0007] Preferably, the multi-level buffer execution module includes a magnetorheological damping buffer unit, a servo drive compliance control unit, and an emergency braking safety unit; the magnetorheological damping buffer unit adopts a shear-type magnetorheological damper, which is coaxially connected in series with the drive shafts of each degree of freedom of the exoskeleton, and is used to absorb overload impact energy by dynamically adjusting the damping force; the servo drive compliance control unit communicates bidirectionally with the native servo drive unit of the exoskeleton, and is used to dynamically adjust the drive stiffness and movement speed of the exoskeleton based on an impedance control algorithm; the emergency braking safety unit adopts a power-off electromagnetic brake, which is coaxially installed with the drive shaft of the exoskeleton, and is used for slow and safe braking under extreme overload risks.

[0008] Preferably, the extended Kalman filter algorithm of the overload prediction module constructs state equations and observation equations based on the shoulder joint-exoskeleton coupled dynamic equations. It uses joint angle, angular velocity, and external interaction torque as state variables and real-time sensor data as observation variables to recursively calculate the joint load prediction value within the next 200ms, thereby completing the overload risk assessment.

[0009] Preferably, the fuzzy PID algorithm of the adaptive threshold tuning module takes the patient's muscle exertion deviation, joint motion trajectory deviation, and load change rate as inputs and the PID parameter correction amount as outputs. It completes online self-tuning of PID parameters through fuzzy control rules, calculates the threshold adjustment amount in real time, and dynamically updates the overload protection threshold.

[0010] Preferably, the edge computing control system communicates with the native control system and multi-level buffer execution module of the wearable shoulder joint rehabilitation exoskeleton via a CAN bus with a communication baud rate of 1Mbps and a control cycle of 1ms; the overload monitoring module communicates with the edge computing control system via an RS485 bus; and the edge computing control system communicates with the human-computer interaction terminal via Bluetooth 5.0.

[0011] Preferably, the damping force of the magnetorheological damping buffer unit is continuously adjustable in the range of 0-200 N·m by means of the excitation current; the braking mode of the emergency braking safety unit is 0.5s slow braking, rather than instantaneous emergency stop.

[0012] Preferably, the protection execution control module of the edge computing control system is equipped with a three-level protection mechanism: low risk level maintains normal training state, medium risk level executes deceleration and stiffness reduction warning protection, high risk level executes magnetorheological buffer and compliant reverse unloading protection action, and extreme risk level executes slow emergency braking.

[0013] Preferably, the wearable shoulder joint rehabilitation exoskeleton is the ARAMIS wearable shoulder joint rehabilitation exoskeleton, which supports three rehabilitation modes: passive training, active assisted training, and resistance training. The edge computing control system matches the corresponding protection strategy and basic overload threshold according to different training modes.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. Achieve proactive overload protection and fundamentally solve the problem of delayed protection: This solution uses the EKF dynamic overload prediction algorithm to identify overload risks 200ms in advance, upgrading the existing "post-event trigger protection" to "pre-event prediction protection", completely avoiding secondary damage to the patient's shoulder joint caused by overload traction, and improving the protection response speed by more than 90%.

[0015] 2. Adaptive dynamic threshold tuning for precise and personalized protection: This solution uses a fuzzy PID algorithm to dynamically adjust the overload threshold based on the patient's individual differences, rehabilitation stage, training mode, and real-time exercise status, replacing the fixed threshold of existing technologies. The overload judgment accuracy is improved to over 98%, which avoids protection failure caused by excessively high thresholds and solves the problem of excessively low thresholds interfering with normal training.

[0016] 3. Multi-level buffering and compliant protection mechanism, balancing safety and training continuity: This solution adopts a three-level execution mechanism of magnetorheological damping buffer + servo compliant control + emergency braking to replace the rigid emergency stop of the existing technology. It can achieve flexible absorption of overload impact energy, protect against rigid impact during the protection process, avoid secondary damage, and at the same time maximize the continuity of rehabilitation training without affecting the rehabilitation effect.

[0017] 4. Multi-source sensor fusion monitoring system with comprehensive overload identification dimensions: This solution integrates six-dimensional force / torque, joint kinematics, electromyography signals, and IMU posture data. It not only monitors load data but also identifies the patient's active movement intentions, distinguishes between normal force exertion and abnormal overload, reduces the overload misjudgment rate to below 2%, and significantly improves clinical applicability.

[0018] 5. Deep integration of protection mechanisms and rehabilitation training, adapting to clinical needs throughout the entire cycle: This program targets three core rehabilitation modes—passive training, active assisted training, and resistance training—and matches them with differentiated protection strategies. It can cover the entire rehabilitation cycle of patients from the acute phase to the sequelae phase, solving the problem of the disconnect between the protection mechanisms and rehabilitation training in existing technologies. While ensuring safety, it maximizes the effectiveness of rehabilitation training. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the shoulder joint rehabilitation exoskeleton application system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the functional architecture of the overload protection device provided in this embodiment of the invention; Figure 3 This is a diagram illustrating the operating mechanism of the shoulder joint dynamic overload prediction algorithm provided in this embodiment of the invention; Figure 4 This is a schematic diagram of the operating mechanism of the adaptive overload threshold dynamic tuning algorithm provided in the embodiment of the present invention. Detailed Implementation

[0021] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0022] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0023] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0024] In this embodiment of the invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0025] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0026] This solution proposes an overload protection device based on a shoulder joint rehabilitation exoskeleton. Through multi-source sensor fusion overload monitoring, algorithm-based overload prediction and adaptive threshold tuning, and a multi-level buffer protection execution mechanism, it achieves precise, forward-looking, and compliant overload protection throughout the entire shoulder joint rehabilitation training process, taking into account both the safety and continuity of rehabilitation training.

[0027] like Figure 1 As shown, this device uses a shoulder joint rehabilitation exoskeleton (such as the ARAMIS wearable exoskeleton) as the basic carrier (basic carrier layer). The core consists of five major components: an overload monitoring module (sensing and acquisition layer), an edge computing control system (core edge control layer), a multi-level buffer execution module (execution drive layer), a human-computer interaction terminal (human-computer interaction and application layer), and a power management unit (power management support unit). The whole device forms a full-link overload protection system of "multi-source sensing - algorithm prediction - hierarchical control - compliant execution - closed-loop feedback".

[0028] The specific functional architecture of this device will be described in detail below.

[0029] Example 1: Overload protection device based on shoulder joint rehabilitation exoskeleton like Figure 2 As shown, the device includes the following parts: (I) Basic Carrier: Wearable Shoulder Joint Rehabilitation Exoskeleton This solution utilizes a clinically mature shoulder joint rehabilitation exoskeleton (such as the ARAMIS wearable device), a publicly available device. Its core structure includes: a back fixation base that is bound to the human torso; a three-degree-of-freedom motion chain for the shoulder joint (shoulder flexion / extension, abduction / adduction, and internal / external rotation); an upper arm binding assembly; a servo drive unit (servo motor + harmonic reducer); and a position encoder. This exoskeleton enables three core rehabilitation modes: passive training, active assisted training, and resistance training. Its maximum range of motion matches the normal physiological range of motion of the human shoulder joint. The device's overload protection module communicates bidirectionally with the exoskeleton's native control system, without altering the exoskeleton's basic rehabilitation training functions.

[0030] The exoskeleton used here is optional and its structure and control system will not be described in detail here.

[0031] (ii) Overload monitoring module The overload monitoring module is the core sensing component of the device, responsible for collecting force, kinematic, and bioelectrical signals throughout the rehabilitation training process, providing a data foundation for overload detection. Each unit communicates with the edge computing control system in real time via an RS485 bus, with a data acquisition cycle of 1ms, meeting real-time control requirements. Its core components and functions are as follows: 1. Multi-dimensional force / torque sensing unit: It adopts a six-dimensional force sensor, which is installed at the connection position between the upper arm binding component of the exoskeleton and the drive chain. The sampling frequency is 1000Hz. It is responsible for collecting the interaction force and torque data of the three degrees of freedom of the shoulder joint in real time, including the traction force, compression force and torsional torque between the patient and the exoskeleton. It is the core data source for load monitoring. 2. Joint Angle Sensing Unit: Composed of the exoskeleton's native 17-bit high-precision absolute encoder, it is coaxially mounted with the drive axis of each degree of freedom, and collects real-time angle, angular velocity, and angular acceleration data of each degree of freedom of the shoulder joint in real time for shoulder joint kinematics and dynamics modeling. 3. Surface electromyography (sEMG) acquisition unit: It adopts dual-channel wireless electromyography electrodes, which are attached to the surface of the deltoid and supraspinatus muscles of the patient. The sampling frequency is 2000Hz. It is responsible for acquiring the electromyography signals of the muscles around the patient's shoulder joint, identifying the patient's active movement intention and muscle exertion state, distinguishing between "load increase caused by the patient's active exertion" and "overload caused by abnormal entrapment", and reducing the overload misjudgment rate. 4. Inertial Measurement Unit (IMU): Installed at the end of the upper arm binding component of the exoskeleton, with a sampling frequency of 500Hz, it collects the spatial attitude and acceleration data of the upper arm in real time, and fuses it with the data of the joint angle sensing unit to correct the shoulder joint kinematic model and improve the accuracy of load calculation.

[0032] (III) Edge Computing Control System The edge computing control system is the core of the device's computation and control. It uses the STM32H743 high-performance microcontroller as the main control chip, runs an embedded real-time operating system, and has a control cycle of 1ms. It is responsible for data preprocessing, algorithm calculation, overload risk assessment, and control command generation. It achieves bidirectional real-time communication with the exoskeleton's native control system via a CAN bus. Its core functional modules and interaction logic are as follows: 1. Data Preprocessing Module: Receives raw data collected by the overload monitoring module and sequentially performs filtering and noise reduction, data alignment, outlier removal, and normalization. Among these processes, Kalman filtering is used to remove motion interference for the six-dimensional force signal, wavelet transform is used to remove power frequency interference for the electromyography signal, and moving average filtering is used to smooth the kinematic data, providing high-quality data for subsequent algorithm calculations. 2. Overload prediction module: It has a built-in shoulder joint dynamics overload prediction algorithm based on extended Kalman filter (EKF) (see the description below). Based on preprocessed multi-source data, it establishes a shoulder joint-exoskeleton coupled dynamic model, predicts the joint load change trend within the next 200ms, identifies overload risks in advance, and realizes "predictive protection" to replace the "post-event triggering" of the existing technology. Explanation of the application mechanism of the shoulder joint dynamics overload prediction algorithm based on extended Kalman filter (EKF): This algorithm is used to establish a coupled dynamic model of the shoulder joint and exoskeleton. Based on real-time acquired multi-source sensor data, it predicts the trend of joint load changes within the next 200ms, identifies overload risks in advance, and solves the problem of protection lag in existing technologies. It is the core of realizing "predictive protection". The algorithm's operation cycle is 1ms, which is perfectly matched with the control cycle of the device, ensuring the real-time performance of the prediction.

[0033] First, the dynamic equations of the shoulder joint-exoskeleton coupling are established using the Lagrange method, as shown in equation (1):

[0034] Definitions of the symbols in the formula: , , These are the joint angle, angular velocity, and angular acceleration vectors of the shoulder joint, representing three degrees of freedom. The vectors are 3×1 in dimension and are acquired in real time by the joint angle sensing unit and the IMU. The inertia matrix of the joint space, with dimensions of 3×3, is obtained by calibrating the inertia parameters of the shoulder joint and exoskeleton. The Coriolis force and centrifugal force matrix has a dimension of 3×3 and is related to the joint motion state in real time. : Gravity vector, with dimensions of 3×1, calculated from the gravity terms of the shoulder joint and exoskeleton; Frictional torque vector, with dimensions of 3×1, including mechanical friction of the exoskeleton and friction of human joints; The output driving torque vector of the exoskeleton servo motor has a dimension of 3×1 and is fed back in real time by the exoskeleton control system. The torque vector generated by the patient's active muscle exertion has a dimension of 3×1 and is calculated from the electromyographic signals of the sEMG acquisition unit. The external interaction torque vector of the joint has a dimension of 3×1, which is the core load data for overload monitoring. It is collected in real time by a six-dimensional force / torque sensing unit.

[0035] Based on the above dynamic equations, the state equations and observation equations of the extended Kalman filter are constructed to achieve load prediction.

[0036] The state equation is as shown in equation (2): ; The observation equation is as shown in equation (3): ; Definitions of symbols and operators in the formula: State variables , represents the system state at time k, including joint angles, angular velocities, and external interaction torques, with a dimension of 9×1; The state transition matrix, with dimensions of 9×9, is obtained by discretizing the dynamic equations and describes the recursive relationship of the system state. The control input matrix, with dimensions of 9×3, consists of the motor output torque and the human body's active torque. ; The system process noise vector has a dimension of 9×1 and follows a Gaussian distribution with a mean of 0 and a covariance of Q. It is used to describe the system modeling error. The observation vector at time k has a dimension of 9×1 and consists of real-time data collected by the joint angle sensing unit and the six-dimensional force sensor. The observation matrix, with dimensions 9×9, describes the mapping relationship between state variables and observed values; The observation noise vector, with dimensions of 9×1, follows a Gaussian distribution with a mean of 0 and a covariance of R, and is used to describe the measurement error of the sensor. Subscript , : These represent the discrete time series of the current time and the previous time, respectively, with a time step of 1ms; superscript : Matrix transpose operator, used to convert a row vector into a column vector.

[0037] like Figure 3 As shown, the algorithm operates as follows: Initialization: During the patient parameter calibration phase, complete the parameter identification of the shoulder joint-exoskeleton dynamic model, and determine the initial values ​​of the state transition matrix A, control input matrix B, observation matrix H, as well as the process noise covariance Q and observation noise covariance R. State prior estimation: Based on the optimal state estimate at time k-1, the state prior estimate at time k is calculated using equation (2). At the same time, update the prior estimation error covariance matrix; Observation Update: Receive real-time observation data from the overload monitoring module at time k. The Kalman gain is calculated using equation (3), and the prior estimate is corrected to obtain the optimal state estimate at time k. ; Load trend prediction: Based on the state transition matrix and the optimal state estimate, the external interaction torque within the next 200ms is recursively calculated. The changing trend is used to obtain the predicted load value. ; Overload risk assessment: predicting load values By comparing the current overload threshold with the established overload risk level, the overload risk level is classified, providing a basis for decision-making for the graded protection mechanism.

[0038] The algorithm's output is directly input into the protection execution control module, achieving an upgrade from "real-time load monitoring" to "overload risk prediction," advancing the triggering time of protection actions by 200ms, and fundamentally solving the problem of protection lag in existing technologies.

[0039] 3. Adaptive Threshold Tuning Module: It has a built-in adaptive overload threshold dynamic tuning algorithm based on fuzzy PID (see the description below). According to the patient's personalized parameters, rehabilitation training mode, and real-time motion status, it dynamically adjusts the overload warning threshold and protection trigger threshold of each degree of freedom, replacing the fixed threshold of the existing technology and adapting to the needs of different patients and rehabilitation stages.

[0040] Adaptive Overload Threshold Dynamic Tuning Algorithm Based on Fuzzy PID This algorithm dynamically adjusts the overload warning threshold and protection trigger threshold based on the patient's personalized parameters, rehabilitation training mode, and real-time exercise status, solving the problem of poor adaptability of fixed thresholds in existing technologies and achieving precise protection with "one policy per person and one policy per stage". The algorithm is based on PID control and combines fuzzy control rules to achieve adaptive tuning of the threshold. The core formula of PID control is shown in equation (4):

[0041] Definitions of symbols and operators in the formula: The threshold adjustment amount at time t is used to correct the basic overload threshold and obtain the protection threshold after real-time tuning. The proportional coefficient determines the response speed of threshold adjustment; Integral coefficient, used to eliminate steady-state error in threshold adjustment; : Differential coefficients, used to predict the trend of threshold adjustment and suppress overshoot; The deviation input at time t is obtained by normalizing parameters from three dimensions: patient muscle exertion deviation, joint motion trajectory deviation, and load change rate. : The integral operator for deviation, which accumulates the deviation over time t;

[0042] The differential operation symbol for the deviation with respect to time calculates the rate of change of the deviation.

[0043] Fuzzy control links with deviation and rate of change of deviation

[0044] As input, the correction values ​​of the three PID coefficients , , As output, a fuzzy control rule base is established to realize online self-tuning of PID parameters, as shown in equation (5):

[0045] In the formula, , , The initial values ​​for the PID parameters are obtained from the basic parameters calibrated during the patient's recovery phase.

[0046] The final real-time set overload protection threshold is shown in equation (6):

[0047] In the formula, The basic overload threshold for this rehabilitation model is set by clinicians based on the patient's rehabilitation assessment results. The overload protection trigger threshold is set to the real-time adjusted threshold at time t, and the warning threshold is also set to... This enables two-level threshold control.

[0048] like Figure 4 As shown, the algorithm's execution logic is as follows: Initial parameter settings: Set the baseline overload threshold based on the patient's age, weight, muscle strength level, rehabilitation stage, and training mode. With the initial values ​​of the PID parameters; Input quantity calculation: Real-time acquisition of the patient's electromyography signals, joint movement trajectory, and load change rate to calculate the deviation input quantity. and the rate of change of deviation ; Fuzzy inference and PID parameter tuning: After fuzzifying the input quantity, inference is performed through the fuzzy control rule base to obtain the correction quantity of the PID parameter, thus completing the online self-tuning of the PID parameter; Threshold dynamic adjustment: The threshold adjustment amount is calculated using equation (4). Then, the real-time overload warning threshold and protection trigger threshold are obtained through equation (6); Closed-loop correction: Based on the execution of protective actions and patient feedback, continuously optimize fuzzy control rules to improve the accuracy of threshold tuning.

[0049] The algorithm's output is directly input into the overload prediction module as the judgment benchmark for overload risk assessment, realizing dynamic adaptive adjustment of the overload threshold and solving the core pain point of poor adaptability of fixed threshold.

[0050] 4. Protection Execution Control Module: Based on the overload risk level, it generates corresponding hierarchical protection control commands and sends them to the multi-level buffer execution module and exoskeleton drive unit. At the same time, it receives feedback signals from the execution unit to form a closed-loop control. 5. Data storage and interaction module: Responsible for storing load data, protective action records, and patient rehabilitation parameters during the training process. It also communicates bidirectionally with the human-computer interaction terminal via Bluetooth 5.0, receiving parameter configuration commands and uploading real-time training data.

[0051] (iv) Multi-level buffer execution module The multi-level buffer execution module is the core of the device's protective execution. It is coaxially mounted with the exoskeleton drive unit, receives control commands from the edge computing control system, and achieves graded and compliant overload protection, replacing the rigid emergency stop of existing technologies. Its core components and functions are as follows: 1. Magnetorheological damping buffer unit: It adopts a shear-type magnetorheological damper, which is coaxially connected in series with the drive shaft of each degree of freedom of the exoskeleton. Its damping force can be continuously adjusted in the range of 0-200 N·m by the excitation current, and the response time is ≤10ms. Its core function is to achieve flexible buffering of the load by dynamically adjusting the damping force when overload occurs, absorb the overload impact energy, and avoid secondary damage from rigid braking. 2. Servo Drive Compliant Control Unit: It communicates bidirectionally with the exoskeleton's native servo drive unit. Based on the impedance control algorithm, it dynamically adjusts the exoskeleton's drive stiffness and movement speed during the overload warning stage to achieve compliant deceleration. At the same time, it adjusts the assist force according to the patient's active movement intention to avoid the exoskeleton's forced movement from exacerbating the overload. 3. Emergency Braking Safety Unit: It adopts a power-off electromagnetic brake, which is installed coaxially with the exoskeleton drive shaft. Under normal circumstances, it is in the released state. It only performs slow braking when an extreme overload risk is anticipated, the patient experiences severe pain, or the equipment malfunctions. At the same time, it cuts off the power output of the servo drive unit to achieve safety protection under extreme conditions. The braking process is a slow stop of 0.5 seconds, rather than an instantaneous emergency stop.

[0052] (v) Human-computer interaction terminal and power management unit 1. Human-computer interaction terminal: It adopts a 7-inch touch screen and is equipped with an Android embedded system. It is responsible for patient information input, personalized parameter calibration, rehabilitation training mode selection, overload protection parameter configuration, real-time training data and protection action visualization display. It also supports clinicians to set protection strategies and view historical training data, realizing full-process management of human-computer interaction. 2. Power Management Unit: Powered by a 24V lithium battery pack, equipped with overcurrent, overvoltage, and overheat protection circuits, providing stable power supply to all modules of the device. It also supports sharing power with the exoskeleton, and has power monitoring and low battery warning functions. The battery life is ≥8 hours, meeting the needs of clinical use throughout the day.

[0053] The device adopts a layered communication architecture, using different communication protocols according to the real-time requirements of the data to ensure both the real-time performance of control and the stability of data transmission, as detailed below: 1. Real-time control layer: CAN 2.0B bus communication is adopted, with a baud rate of 1Mbps. It is responsible for the data transmission of control commands and status feedback between the edge computing control system and the exoskeleton native control system and the multi-level buffer execution module. The maximum transmission delay is ≤1ms, which meets the control requirements of real-time protection. 2. Sensor data acquisition layer: It adopts RS485 bus communication with a baud rate of 57600bps. It is responsible for the raw data transmission between each sensing unit of the overload monitoring module and the edge computing control system. It supports multi-node parallel acquisition and ensures data synchronization. 3. Human-Computer Interaction and Data Upload Layer: Adopts dual communication modes of Bluetooth 5.0 and WiFi. Bluetooth 5.0 is responsible for real-time data interaction between the edge computing control system and the human-computer interaction terminal, with a transmission latency of ≤10ms; WiFi is responsible for cloud upload and remote management of training data, supporting clinicians to remotely view patient rehabilitation data.

[0054] The device as a whole forms a closed-loop control logic of "sensing-prediction-decision-execution-feedback", and the complete process is as follows: 1. The overload monitoring module collects multi-source sensor data in real time and sends it to the preprocessing module of the edge computing control system to complete data filtering and standardization. 2. The overload prediction module is based on the EKF algorithm to predict the trend of joint load and assess overload risk; the adaptive threshold tuning module is based on the fuzzy PID algorithm to dynamically update the overload threshold. 3. The protection execution control module classifies the overload risk level based on the comparison between the predicted load and the real-time threshold, and generates corresponding hierarchical protection control instructions; 4. Control commands are sent via the CAN bus to the multi-level buffer execution module and the exoskeleton drive unit to execute corresponding protection actions; 5. The action status of the execution unit, the motion status of the exoskeleton, and the real-time load data are fed back to the edge computing control system to complete the state correction of the closed-loop control and update the parameters of the algorithm model.

[0055] Example 2: Overload Protection Method Based on Shoulder Joint Rehabilitation Exoskeleton This method is implemented based on the above-mentioned apparatus, and the steps are as follows: Step 1: Patient-specific parameter calibration and system initialization Clinicians input basic patient information (age, gender, weight, disease duration), rehabilitation assessment results (shoulder joint muscle strength level, joint range of motion, pain threshold), and rehabilitation stage (acute phase, recovery phase, sequelae phase) through a human-computer interaction terminal. The exoskeleton is then fitted to the patient, and coaxiality calibration between the shoulder joint and the exoskeleton is completed. The system is then started to initialize parameters. Through passive slow-motion training, parameter identification of the shoulder joint-exoskeleton coupled dynamic model is completed, determining the initial parameters for the EKF algorithm and the fuzzy PID algorithm. Simultaneously, basic overload thresholds, protection strategies, and training modes are set. This completes the personalized adaptation of the system to the patient, providing accurate basic parameters for subsequent algorithm calculations and protective controls, ensuring the accuracy of overload judgment.

[0056] Step 2: Matching Rehabilitation Training Modes and Configuring Protection Strategies Based on the patient's rehabilitation stage, the system selects the corresponding rehabilitation training mode and automatically matches the appropriate protection strategy: ① Passive training mode: The patient does not actively exert force; the exoskeleton drives the shoulder joint to complete the trajectory movement. The protection strategy focuses on "position limiting + torque overload protection," with a lower threshold setting to prioritize safety. ② Active assisted training mode: The patient actively exerts force, and the exoskeleton provides assistance. The protection strategy focuses on "active intent recognition + compliant overload buffering," with the threshold dynamically adjusted according to the patient's exertion state. ③ Resistance training mode: The exoskeleton provides resistance to train the patient's muscle strength. The protection strategy focuses on "resistance overload limiting + extreme safety braking," with a higher threshold setting to balance training effectiveness and safety. This achieves deep adaptation between the protection strategy and the rehabilitation training mode, preventing the protection mechanism from interfering with normal rehabilitation training while ensuring safety protection under different training modes.

[0057] Step 3: Real-time acquisition and preprocessing of multi-source sensor data After rehabilitation training begins, the overload monitoring module collects six-dimensional force / torque data, joint angle / angular velocity data, electromyography signals, and IMU posture data in real time at 1ms intervals, and sends them to the edge computing control system via RS485 bus. The data preprocessing module sequentially performs filtering and noise reduction, timestamp alignment, outlier removal, and dimension normalization on the raw data to remove motion interference, power frequency interference, and measurement noise, obtaining high-quality standardized data, which is then input into the algorithm module. This provides an accurate and stable data source for algorithm operation, reduces the impact of sensor measurement errors on overload judgment, and improves the accuracy of overload identification.

[0058] Step 4: Load prediction and overload risk assessment based on EKF algorithm The overload prediction module receives preprocessed standardized data, performs state prior estimation and observation update based on the EKF algorithm, obtains the optimal estimate of the joint state and load at the current moment, and recursively calculates the load change trend within the next 200ms to obtain the predicted load value. The predicted load value is compared with the real-time tuned overload threshold to classify three overload risk levels: ① Low risk: predicted load < 0.8Fth (warning threshold), the system maintains normal training state; ② Medium risk: 0.8Fth ≤ predicted load < Fth (protection trigger threshold), triggering warning-level protection; ③ High risk: predicted load ≥ Fth, triggering protection-level protection; Extreme risk: predicted load ≥ 2Fth, triggering emergency safety braking. This achieves early prediction of overload risk, replacing the post-event triggering of existing technologies, advancing the response of protective actions by 200ms, fundamentally avoiding damage to the patient's shoulder joint caused by overload traction.

[0059] Step 5: Dynamic tuning of overload threshold based on fuzzy PID algorithm Executed synchronously with step 4, the adaptive threshold tuning module receives real-time data on the patient's electromyography (EMG) signals, joint motion trajectory deviation, and load change rate. It calculates the deviation input and deviation change rate, performs online self-tuning of the PID parameters through fuzzy inference, calculates the threshold adjustment amount, and updates the overload warning threshold and protection trigger threshold in real time. Simultaneously, based on the patient's real-time pain feedback, it supports manual adjustment of the threshold range via a human-computer interaction terminal, ensuring the threshold matches the patient's tolerance. This achieves dynamic adaptive adjustment of the overload threshold, adapting to the protection needs of patients under different movement states and force levels, solving the problem of poor adaptability of fixed thresholds, and reducing the overload misjudgment rate.

[0060] Step 6: Overload level determination and triggering of graded protection mechanism The protection execution control module triggers corresponding graded protection mechanisms based on the overload risk level: ① Low risk level: The system maintains normal training status, the servo drive unit runs according to the preset trajectory, the magnetorheological damper maintains a low damping state, and no protection action is performed; ② Medium risk level: Warning-level protection is triggered, the servo drive compliant control unit reduces the movement speed and drive stiffness of the exoskeleton based on the impedance control algorithm, the magnetorheological damper slightly increases the damping force, and the human-machine interface terminal issues an audible and visual warning to remind the patient and doctor; ③ High risk level: Protection-level protection is triggered, the magnetorheological damper rapidly increases the damping force to absorb the overload impact energy, the servo drive unit performs compliant deceleration, and at the same time adjusts the movement trajectory in the opposite direction to release the overload tension. Normal training resumes after the load drops to a safe range; ④ Extreme risk level: Emergency braking protection is triggered, the magnetorheological damper reaches the maximum damping force, the electromagnetic brake performs 0.5s slow braking, the power output of the servo drive unit is cut off, and the human-machine interface terminal issues an emergency alarm. It achieves graded and compliant overload protection, replacing the rigid emergency stop technology of the existing technology, avoiding secondary injuries caused by the protective action itself, while maximizing the continuity of rehabilitation training.

[0061] Step 7: Closed-loop adjustment of protective action execution and training status During the execution of the protective action, the overload monitoring module collects load data and joint motion status in real time and feeds it back to the edge computing control system to form a closed-loop control. The protective execution control module dynamically adjusts the damping force of the magnetorheological damper and the motion parameters of the servo drive unit according to the real-time load changes to ensure that the load is smoothly reduced to a safe range. If the high-risk protection is triggered three times consecutively, the system automatically pauses training and reminds the doctor to check the patient's joint status and exoskeleton wearing condition, and adjust the training parameters and protection threshold. This achieves closed-loop precise control of the protective action, ensuring the compliance and stability of the protective process, while adapting to the patient's real-time status through dynamic adjustment of training parameters.

[0062] Step 8: Training Completion and Data Archiving and Analysis After rehabilitation training concludes, the system automatically stops data collection, storing complete training data (training duration, joint range of motion, load change curve, protective action trigger records, and overload risk events) locally, while simultaneously uploading it to the human-computer interaction terminal and the cloud. It automatically generates a rehabilitation training report, statistically analyzing the patient's training completion rate, number of overload events, and execution of protective actions, providing data support for clinicians to adjust rehabilitation plans. This achieves end-to-end data-driven management of patient rehabilitation training, providing quantitative evidence for rehabilitation effectiveness evaluation, and continuously optimizing the algorithm model and protective strategies.

[0063] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0064] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0065] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0066] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0067] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0068] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0069] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0070] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0071] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0072] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An overload protection device based on a shoulder joint rehabilitation exoskeleton, characterized in that, This includes a wearable shoulder joint rehabilitation exoskeleton, an overload monitoring module, an edge computing control system, a multi-level buffered execution module, and a human-computer interaction terminal, among which: The wearable shoulder joint rehabilitation exoskeleton is a three-degree-of-freedom driven structure, serving as the basic carrier of the device. The overload monitoring module is installed on the wearable shoulder joint rehabilitation exoskeleton, and its signal output end is communicatively connected to the input end of the edge computing control system to collect multi-source load and motion data in real time during the rehabilitation training process. The edge computing control system is bidirectionally connected to the native control system and the multi-level buffer execution module of the wearable shoulder joint rehabilitation exoskeleton, and is used for data processing, overload risk prediction and protection control command generation. The multi-level buffer execution module is coaxially mounted with the drive shaft of the wearable shoulder joint rehabilitation exoskeleton and is used to perform graded compliant overload protection actions. The human-computer interaction terminal is bidirectionally connected to the edge computing control system and is used for parameter configuration, data visualization, and human-computer interaction.

2. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 1, characterized in that, The overload monitoring module includes a six-dimensional force / torque sensing unit, a joint angle sensing unit, a surface electromyography signal acquisition unit, and an inertial measurement unit. The six-dimensional force / torque sensing unit is installed at the connection position between the upper arm binding component and the drive chain of the wearable shoulder joint rehabilitation exoskeleton, and is used to collect the interaction force and torque data of each degree of freedom of the shoulder joint in real time. The joint angle sensing unit is coaxially mounted with the exoskeleton drive shaft and is used to collect joint angle, angular velocity and angular acceleration data. The surface electromyography signal acquisition unit uses a dual-channel wireless electromyography electrode to acquire electromyography signals of the muscles around the patient's shoulder joint and identify the intention to move actively. The inertial measurement unit is installed at the end of the upper arm binding component of the exoskeleton and is used to collect upper arm spatial attitude and acceleration data.

3. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 1, characterized in that, The edge computing control system incorporates a data preprocessing module, an overload prediction module, an adaptive threshold tuning module, a protection execution control module, and a data storage module. The data preprocessing module performs filtering, noise reduction, data alignment, and normalization on the acquired raw data. The overload prediction module incorporates a shoulder joint dynamics overload prediction algorithm based on extended Kalman filtering to predict joint load change trends and assess overload risk levels. The adaptive threshold tuning module incorporates an adaptive overload threshold dynamic tuning algorithm based on fuzzy PID, which is used to dynamically update the overload warning threshold and the protection trigger threshold. The protection execution control module is used to generate hierarchical protection control commands and drive the execution module to complete protection actions.

4. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 1, characterized in that, The multi-level buffer execution module includes a magnetorheological damping buffer unit, a servo drive compliance control unit, and an emergency braking safety unit. The magnetorheological damping buffer unit adopts a shear-type magnetorheological damper, which is coaxially connected in series with the drive axes of each degree of freedom of the exoskeleton, and is used to absorb overload impact energy by dynamically adjusting the damping force. The servo drive compliance control unit communicates bidirectionally with the exoskeleton's native servo drive unit, and is used to dynamically adjust the drive stiffness and movement speed of the exoskeleton based on an impedance control algorithm. The emergency braking safety unit uses a power-off electromagnetic brake, which is coaxially mounted with the exoskeleton drive shaft, and is used for slow and safe braking under extreme overload risks.

5. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 3, characterized in that, The extended Kalman filter algorithm of the overload prediction module constructs state equations and observation equations based on the shoulder joint-exoskeleton coupled dynamic equations. It uses joint angle, angular velocity, and external interaction torque as state variables and real-time sensor data as observation variables to recursively calculate the joint load prediction value within the next 200ms, thus completing the overload risk assessment.

6. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 3, characterized in that, The fuzzy PID algorithm of the adaptive threshold tuning module takes the patient's muscle force deviation, joint motion trajectory deviation, and load change rate as inputs and the PID parameter correction amount as output. It completes online self-tuning of PID parameters through fuzzy control rules, calculates the threshold adjustment amount in real time, and dynamically updates the overload protection threshold.

7. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 1, characterized in that, The edge computing control system communicates with the native control system and multi-level buffer execution module of the wearable shoulder joint rehabilitation exoskeleton via a CAN bus with a communication baud rate of 1Mbps and a control cycle of 1ms; the overload monitoring module communicates with the edge computing control system via an RS485 bus; and the edge computing control system communicates with the human-computer interaction terminal via Bluetooth 5.

0.

8. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 4, characterized in that, The damping force of the magnetorheological damping buffer unit is continuously adjustable in the range of 0-200 N·m by the excitation current; the braking mode of the emergency braking safety unit is 0.5s slow braking, rather than instantaneous emergency stop.

9. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 1, characterized in that, The protection execution control module of the edge computing control system is equipped with a three-level protection mechanism: low risk level maintains normal training state, medium risk level executes deceleration and stiffness reduction warning protection, high risk level executes magnetorheological buffer and compliant reverse unloading protection action, and extreme risk level executes slow emergency braking.

10. The overload protection device based on a shoulder joint rehabilitation exoskeleton according to claim 1, characterized in that, The wearable shoulder joint rehabilitation exoskeleton is the ARAMIS wearable shoulder joint rehabilitation exoskeleton, which supports three rehabilitation modes: passive training, active assisted training, and resistance training. The edge computing control system matches the corresponding protection strategy and basic overload threshold according to different training modes.

Images