A sagittal plane standing balance dynamics simulation method and system based on centroid time delay feedback and application thereof

Abstract

This invention belongs to the field of dynamic simulation control, and specifically discloses a method, system, and application for human sagittal plane standing balance dynamic simulation based on center-of-mass delay feedback. The method includes: constructing a multi-joint biomechanical model of the human sagittal plane; constructing a joint torque generation model, where joint torques include passive and active torques; employing an active control strategy based on center-of-mass delay feedback, using the displacement and velocity deviations of the human center of mass as feedback state quantities, and introducing neural response delays to calculate the active joint torques; building a physical topology model corresponding to the biomechanical model in a physical simulation environment, connecting the joint torque generation model as a joint torque source to each joint to form a simulation model; inputting external disturbance displacement data into the simulation model, and identifying simulation model parameters based on simulated and real joint torques. This invention overcomes the limitations of traditional single-joint independent control, improving the fitting accuracy of multi-joint torques and the biorealism of collaborative modes.

Description

Technical Field

[0001] This invention belongs to the field of dynamic simulation control, and more specifically, relates to a method, system and application of human sagittal plane standing balance dynamic simulation based on centroid delay feedback. Background Technology

[0002] Standing balance is fundamental for complex movements such as walking and running. Although the human body appears static when standing upright, it is actually a nonlinear unstable system in dynamic adjustment. When subjected to external disturbances (such as sudden ground shifts or external force pushing), the central nervous system (CNS) needs to quickly integrate visual, vestibular, and proprioceptive information. Through spinal reflexes and long latency reflexes, it coordinates the muscle torques of multiple joints in the lower limbs, such as the hip, knee, and ankle, to keep the body's center of mass (CoM) within the support area and prevent falls.

[0003] A deep understanding of the dynamic mechanisms of balance recovery in the human body is crucial for two types of applications: first, clinical rehabilitation assessment, namely, how to quantitatively assess the balance ability of the elderly or patients with neurological injuries; and second, rehabilitation assistive device design, namely, how to design exoskeleton robots that can provide "on-demand assistance" for paraplegic or hemiplegic patients. Currently, forward dynamics simulation is the core method for studying this problem. However, existing simulation methods have significant technical bottlenecks in the following aspects: First, biomechanical models are oversimplified. Existing studies mostly use single inverted pendulum models (considering only the ankle joint) or double inverted pendulum models (ankle-hip joint), generally neglecting the role of the knee joint. In fact, the flexion and extension movements of the knee joint play an irreplaceable role in regulating the height of the center of mass, cushioning ground impact, and coordinating upper and lower limb movements. Ignoring the knee joint leads to kinematically distorted simulation results, failing to accurately reflect the real human joint coordination patterns.

[0004] Second, the control strategies lack physiological realism. Traditional simulation control often employs PD control based on local joint angle feedback. This method severs the dynamic coupling between joints and struggles to handle the inherent time delay (approximately 100ms) in neural signal transmission. Another type of optimal control method based on LQR (linear quadratic regulator), while considering multi-joint coupling, often generates overly smooth and "perfect" motion trajectories, lacking the underdamped oscillation characteristics and complex energy dissipation features exhibited by the real human body under large disturbances.

[0005] Third, there is a lack of realistic interaction with the physical environment. Many simulations only involve numerical integration of pure mathematical equations, simplifying the ground contact model and inertial force input, making it difficult to directly reproduce the complex inertial effects generated by treadmill motion in real experiments.

[0006] Therefore, there is an urgent need for a simulation method that can integrate multi-joint dynamics, consider neural delays, provide feedback based on global centroid variables, and run in high fidelity within a physics engine. Summary of the Invention

[0007] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a method, system and application for human sagittal plane standing balance dynamics simulation based on centroid delay feedback, the purpose of which is to achieve high-fidelity dynamics simulation of human sagittal plane standing balance.

[0008] To achieve the above objectives, according to a first aspect of the present invention, a method for simulating human sagittal plane standing balance dynamics based on centroid delay feedback is proposed, comprising the following steps: A human sagittal multi-joint biomechanical model was constructed, which includes four rigid body segments: foot, lower leg, thigh, and trunk, as well as three rotational joints: ankle, knee, and hip connecting the rigid body segments. A joint torque generation model is constructed to generate joint torques, which include passive torques and active torques. An active control strategy based on center of mass delay feedback is adopted, using the displacement and velocity deviations of the human body's center of mass as feedback state variables, and introducing neural response delays to calculate the active torques of each joint. In a physical simulation environment, a physical topology model corresponding to the biomechanical model is built. In the physical topology model, the active torques of each joint calculated by the joint torque generation model are used as joint torque sources and connected to each joint to form a complete human sagittal plane standing balance control model, i.e., a simulation model. External disturbance displacement data is input into the simulation model, and the parameters in the simulation model are identified based on the simulated joint torque and the actual joint torque output by the simulation model. A simulation of human sagittal plane standing balance dynamics is achieved based on a simulation model with identified parameters.

[0009] As a further preferred embodiment, the formula for calculating the active torque is:

[0010] in, The main dynamic moment vector, The current moment; This refers to the delay time of neural response; It is the offset vector of the centroid position relative to the equilibrium state before the delay time; It is the offset vector of the centroid velocity relative to the equilibrium state before the delay time; , These are the center-of-mass position feedback gain vector and the center-of-mass velocity feedback gain vector, respectively.

[0011] As a further preferred option, the centroid position feedback gain vector and center of mass velocity feedback gain vector All are 3×1 cooperative mapping vectors, representing the coupling driving effect of the centroid state deviation on the torques of the ankle, knee, and hip joints.

[0012] As a further preferred option, neural response delay time Set to 100ms.

[0013] As a further preferred method, external disturbance displacement data is input into the simulation model, and the parameters in the simulation model are identified based on the simulated joint torques and actual joint torques output by the simulation model, including: Kinematic data and ground reaction force data of a real human body under disturbance of a support surface were collected through experiments, and the real joint torques of the human body were calculated by inverse kinematics. ; Simulated joint torque Compared to actual joint torque The sum of squared errors between them is the objective function; The objective function is solved using a nonlinear least squares algorithm to obtain the simulation model parameters corresponding to the minimum objective function, including the centroid position feedback gain vector. and center of mass velocity feedback gain vector .

[0014] As a further preferred approach, the external disturbance displacement data input method is a kinematic drive mode. Specifically, the time series data of the treadmill displacement collected in the experiment is directly used as the position drive signal and connected to the sliding pair between the support platform and the ground in the simulation model, driving the support platform to produce a horizontal translational motion relative to the ground.

[0015] As a further preferred embodiment, the passive torque is determined by the inherent stiffness and damping of the joint, and the passive torque adopts a time-delayed spring-damped model.

[0016] As a further preferred embodiment, the formula for calculating the passive torque is:

[0017] in, The passive moment vector, The current moment; This is the passive stiffness coefficient matrix of the joint. This is the joint passive damping coefficient matrix. , These represent the deviations of the joint angles and angular velocities from their equilibrium positions.

[0018] According to a second aspect of the present invention, a human sagittal plane standing balance dynamics simulation system based on centroid delay feedback is provided, comprising a processor, the processor being used to execute the above-described human sagittal plane standing balance dynamics simulation method based on centroid delay feedback.

[0019] According to a third aspect of the present invention, a collaborative control method for a balance-assisted exoskeleton is provided. The active torque of the human joints is simulated using the above-mentioned human sagittal plane standing balance dynamics simulation method based on the center of mass delay feedback. The human joint torque is used as the target reference value. The exoskeleton actuator provides auxiliary compensation torques corresponding to the target reference value at the ankle, knee and hip joints to realize real-time control of the deviation of the human center of mass.

[0020] In summary, compared with the prior art, the above-described technical solutions conceived by this invention mainly possess the following technical advantages: 1. This invention constructs a three-degree-of-freedom model including the ankle, knee, and hip joints, breaking through the limitations of traditional single-joint independent control. Based on the deviation of the center of mass state and the introduction of neural response delay, it realizes global collaborative regulation of the center of mass, which significantly improves the fitting accuracy of multi-joint torque and the bio-realism of the collaborative mode. Combined with physical simulation, it can accurately reproduce the main torque generation mechanism and center of mass movement trajectory of the human body under external disturbance.

[0021] 2. The control strategy proposed in this invention overcomes the limitations of traditional independent joint control, which severs inter-joint coupling. The human lower limb is a multi-redundant degree-of-freedom system. If each joint is directly adjusted, the controller needs to handle complex kinematic couplings and is prone to inter-joint conflicts. The center of mass is a core indicator for measuring overall balance. By focusing on the state of the center of mass, this invention simplifies the complex multi-joint coordination problem into a single-objective control, making the calculation simpler and more consistent with the coordination theory of human neuromuscular control. Specifically, by introducing a coordination mapping vector, the coupling driving effect of the center of mass state deviation on the torques of the ankle, knee, and hip joints is clarified. Simulation results show that this mechanism can spontaneously generate "ankle-knee in-phase coordination" and "hip joint anti-phase adjustment" movement patterns that are highly consistent with the real human body.

[0022] 3. This invention combines a physical simulation environment, abandoning the simplified assumptions about disturbance signals found in pure mathematical models. Employing a "kinematic drive mode," it directly uses experimentally acquired displacement data to drive the simulation support platform, enabling extremely accurate reproduction of inertial effects, nonlinear ground contact mechanics characteristics, and attitude response under large disturbances during the experimental process.

[0023] 4. The joint torques generated by the method of this invention show a high degree of fit with real human experimental data; at the same time, the simulated centroid phase plane trajectory can accurately reproduce the spiral convergence characteristics and energy dissipation rate during the human body's balance recovery process. This is suitable for the development and evaluation of assisted control strategies for exoskeleton robots. Attached Figure Description

[0024] Figure 1 A flowchart of the human sagittal plane standing balance dynamics simulation method based on centroid delay feedback provided in an embodiment of the present invention; Figure 2 A schematic diagram of a three-degree-of-freedom four-bar biomechanical model of the human body provided in an embodiment of the present invention; Figure 3 The above is the platform disturbance acceleration curve used in the simulation of this invention embodiment; Figure 4 This is a time-series comparison chart of ankle, knee, and hip joint torques generated by simulation and experimental data in an embodiment of the present invention; Figure 5 This is a comparison diagram of the experimental and simulated centroid phase plane trajectories in the embodiments of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0026] This invention provides a human sagittal plane standing balance dynamics simulation method based on centroid delay feedback. This method is mainly implemented on the MATLAB / Simulink and Simscape Multibody simulation platforms. Figure 1 As shown, it includes the following steps: Step 1: Obtain the anthropometric parameters of the subjects and construct a human sagittal multi-joint biomechanical model based on them. The biomechanical model includes four rigid body segments: foot, lower leg, thigh and trunk, as well as three rotational joints: ankle, knee and hip, which connect the segments, forming a three-degree-of-freedom underactuated mechanical system.

[0027] Specifically, this step aims to establish a basic dynamic model that accurately reflects the subject's body inertial characteristics. For example... Figure 2As shown, the biomechanical model is a three-degree-of-freedom four-bar rigid body system in the sagittal plane, consisting of four rigid body segments from bottom to top: foot, shank, thigh, and trunk (HAT, i.e., the combination of head, arms, and trunk). The segments are connected by revolute joints and are defined as the ankle, knee, and hip joints, respectively. The angles of the ankle, knee, and hip joints are defined as the angles between the links. The model parameters include the length, mass, center of mass position, and moment of inertia of each segment, which are determined based on the subject's height, weight, and human inertial parameter standards.

[0028] Furthermore, to improve the individualized accuracy of the simulation, the model parameters are not based on universal standard values, but rather on the subjects' actual height. (Unit: m) and body weight (Unit: kg) Calculations are performed. This invention uses standard human segment data to determine the length of each segment. (Obtained based on linear regression of height), mass (As given directly in the standard).

[0029] Since human body segments are not homogeneous rods, and given the difficulty in measuring the true center of mass distribution of each segment, the relative distances of the center of mass of the lower leg, thigh, and upper body to the ankle, knee, and hip joints are determined by referring to the national standard for adult human inertial parameters. Moment of inertia of each segment It is also difficult to measure directly, referring to the rotational inertia model parameters commonly used in existing research for various segments of the human body.

[0030] Methods for determining parameters in biomechanical models include: based on the subject's height. The length of each segment was calculated using a linear regression equation, including the lower leg length. Thigh length Length of upper limbs and trunk Based on human inertial parameter standards, the mass and moment of inertia of each segment were calculated according to the subject's weight. The mass of each segment was distributed proportionally to body weight: lower leg mass... Thigh quality trunk mass The center of mass is distributed proportionally according to segment length: distance from the center of mass of the lower leg to the ankle joint. Distance from the center of mass of the thigh to the knee joint Distance from the center of mass of the upper limb to the hip joint The rotational inertia of each segment's center of mass about the coronal axis was used as a reference for commonly used rotational inertia models for human segments in existing research: Lower leg rotational inertia. Thigh rotational inertia upper limb rotational inertia Through the above parameterization process, the constructed model can specifically reflect the biomechanical characteristics of different subjects, providing an accurate inertial basis for subsequent dynamic calculations.

[0031] Furthermore, based on the principles of Lagrange dynamics, the nonlinear dynamic equations of the biomechanical model can be derived, and linearized at the upright equilibrium position to establish a system state-space model. This state-space model can then be used to analyze the inherent dynamic characteristics of the human mechanical system, verifying the instability of the system in the open-loop state, and revealing the dynamic characteristics of the system from a mathematical theoretical perspective. Specifically, this includes: (1) Selecting generalized coordinates , respectively, represent the absolute angles of the ankle, knee, and hip joints relative to the vertical line of gravity; (2) Establish the nonlinear dynamic equations based on the Lagrange equations:

[0032] in, The inertia matrix, For the Coriolis and centrifugal force terms, This is the potential energy term (including gravitational potential energy and elastic potential energy). It is a generalized force (including active torque and damping torque); (3) To facilitate control theory analysis, the above nonlinear dynamic equations are expanded using Taylor expansion near the upright equilibrium point, and the first-order terms are retained to obtain the state equations, i.e., the system state-space model:

[0033] in, Let be the system state vector. This is the torque input vector. and These are the system matrix and the input matrix, respectively. (4) System matrix based on state equation Eigenvalue analysis was used to verify the instability of the human body under uncontrolled conditions. Calculation results show that... The existence of eigenvalues ​​with positive real parts corresponds physically to the instability of an inverted pendulum system, thus proving that the human multi-joint system cannot maintain uprightness solely through passive stiffness and must incorporate active neural control. Simultaneously, based on the input matrix... Dimensions ( This established that the feedback gain matrix in the subsequent active control strategy must have the structural characteristic of mapping 6-dimensional state variables to 3-dimensional torque vectors.

[0034] Step 2: Establish a joint torque generation model, joint torque By passive torque and driving torque It is composed of two superimposed parts, that is .

[0035] Furthermore, the passive torque is determined by the inherent stiffness and damping of the joint, and is used to simulate the inherent viscoelastic resistance generated by the ligaments, tendons, and muscle-tendon complexes around the joint during tension. The passive torque model is a time-delayed spring-damped model, and its expression is:

[0036] in, The passive moment vector; This is the joint passive stiffness coefficient matrix, which characterizes the joint's ability to resist deformation. This parameter is obtained through the parameter identification process. This is the joint passive damping coefficient matrix, which characterizes the energy dissipation capacity. In this embodiment, to simplify the calculation, the damping coefficients of each joint are set to constant values. To improve the speed and accuracy of parameter identification; and These represent the deviations of each joint angle and angular velocity from the equilibrium position.

[0037] Furthermore, regarding the active torque: an active control strategy based on center-of-mass delay feedback is constructed. The displacement and velocity deviations of the human body's total center of mass (CoM) are calculated in real time as global feedback state variables. Neural transmission delay is introduced to calculate the active torques that collaboratively drive the ankle, knee, and hip joints. This aims to simulate the collaborative regulation mechanism of the central nervous system based on task-level variables.

[0038] The core of the active control strategy lies in using the center of mass as a task-level variable to coordinate the control of multiple joints. The active control strategy does not rely on the local angular deviation of a single joint, but rather on the kinematic state of the human body's total center of mass (CoM). Its calculation process includes: 1) State calculation: using the forward kinematics formula, the horizontal position of the human body's total center of mass is synthesized in real time. and horizontal speed 2) Delay introduction: Introduce a fixed time delay. This simulates the total physiological lag from sensory organs (vestibular, visual, proprioceptive) perceiving disturbances, the transmission of neural signals to the central nervous system, the central processing and decision-making, to the transmission of motor commands to muscles, and finally to the generation of joint torques. 3) Cooperative mapping: The active control torque of each joint is calculated using the following feedback control law. :

[0039] in, The main dynamic moment vector; The neural response delay time is preferably set to 100ms. This value is a typical physiological delay time in the process of human balance regulation and is used to simulate the physiological lag from the human body's perception of disturbance to the generation of active joint torque. Let be the offset vector of the centroid position relative to the equilibrium state before the delay time. It is the offset vector of the centroid velocity relative to the equilibrium state before the delay time. and These are the center-of-mass position feedback gain vector and the center-of-mass velocity feedback gain vector, respectively.

[0040] Feedback gain vector and for The co-mapping vector implies a single centroid state deviation. It will simultaneously convert the force commands of the ankle, knee, and hip joints through different weighted elements in the matrix; the structure of the active torque calculation formula can be expressed as:

[0041] in, The centroid position feedback gain vector; This is the centroid velocity feedback gain vector; the subscripts a, k, and h represent the ankle, knee, and hip joints, respectively.

[0042] Feedback gain vector and It characterizes the coupled driving effect of the center of mass deviation on the torques of the ankle, knee and hip joints, thus realizing the coordinated regulation among multiple joints in mathematical structure and reproducing the coordinated characteristics of the human body in balance recovery.

[0043] Step 3: Build a physical topology model corresponding to the biomechanical model in the physical simulation environment, connect the joint torque generation model as a torque source to each joint, and form a complete human sagittal plane standing balance control system model, i.e., simulation model.

[0044] Furthermore, the physical simulation environment is Simscape Multibody. This embodiment builds a physical topology model using the Simscape Multibody toolbox, mainly including the following modules: Multibody Dynamics Module: Use the Brick Solid module to construct the geometric solids of each limb and input the mass, center of mass and inertia matrix calculated in step S1; use the Revolute Joint module to connect adjacent limbs and configure the passive stiffness and damping parameters of step S2 in the "Internal Mechanics" property of the joint module.

[0045] Disturbance Input Interface Module: Construct a platform model and connect it to the World Frame using a Prismatic Joint. This joint is configured to move only in the horizontal direction, serving as the physical interface for external disturbance input.

[0046] Sensing and signal processing module: Transform Sensors are configured on each limb to output the pose matrix relative to the world coordinate system in real time for centroid calculation; a Transport Delay module is introduced to send the calculated centroid state signal to the controller after a 100ms delay.

[0047] Execution and Drive Module: Converts the calculated total torque signal into a physical signal using the Simulink-PS Converter and connects it to the Actuation port of each joint.

[0048] Step 4: Input the external disturbance displacement data into the support platform in the simulation model, perform forward dynamic simulation analysis, and identify and iteratively optimize the undetermined parameters in the simulation model based on real experimental data, so that the simulation output approximates the real human body response.

[0049] Furthermore, the external disturbance input is applied in a "kinematic drive" mode. Specifically, the actual displacement time series data of the treadmill, which is collected by the optical motion capture system in the real experiment, is directly connected to the "Position Target" port of the support platform sliding joint in the Simscape model. That is, the treadmill displacement time series data collected in the experiment is directly used as the position drive signal and connected to the sliding joint between the support platform and the ground in the simulation model to drive the support platform to produce a horizontal translational motion relative to the ground.

[0050] Compared to traditional force-driven or acceleration-driven methods, the above method has significant advantages: the Simscape physics engine automatically calculates the nonlinear inertial force generated by the movement of the support platform based on the input displacement curve, and automatically handles the contact and friction between the human foot and the support platform, thereby avoiding numerical noise caused by manually differentiating the displacement data and ensuring that the physical field in the simulation environment is highly consistent with the real experiment.

[0051] Furthermore, in step 4, the parameter identification and optimization process specifically includes: (1) Through real experiments, kinematic data and ground reaction force data of real human body when it is disturbed by the supporting surface are collected, and the real joint torque of each joint of the human body during the experiment is calculated by inverse kinematics. ; (2) Construct the objective function The objective function is the simulated joint torque generated by the simulation. Compared to actual joint torque The sum of squared errors between them is expressed mathematically as follows:

[0052] in, For the number of joints, This represents the number of sampling points; (3) Using the joint passive stiffness coefficient and active feedback gain vector as optimization variables, iteratively run the forward dynamics simulation; specifically, use the nonlinear least squares algorithm to iteratively optimize the feedback gain vector. , and joint passive stiffness coefficient matrix This process continues until the objective function converges. This makes the response characteristics of the simulation model approximate real human data. The optimized parameters represent the characteristics of the subject's neural control strategy under specific perturbations.

[0053] This invention provides a human sagittal plane standing balance dynamics simulation system based on centroid delay feedback, including a processor, which is used to execute the above-described human sagittal plane standing balance dynamics simulation method based on centroid delay feedback.

[0054] This invention provides a collaborative control method for a balance-assisted exoskeleton. The method uses the above-mentioned human sagittal plane standing balance dynamics simulation method based on the center of mass delay feedback to simulate the active torque of the human joints and use it as the target reference value. The exoskeleton actuators provide corresponding auxiliary compensation torques at the ankle, knee and hip joints to realize real-time control of the human center of mass deviation and enhance the user's balance recovery ability when subjected to external disturbances.

[0055] Specifically, the application of the simulation method of this invention in lower limb exoskeleton robot systems includes the following four aspects: (1) Exoskeleton hardware design and selection based on “digital twin” (offline stage).

[0056] During the prototype development phase of the exoskeleton, the simulation model of this invention was used as a "digital twin" platform: Extreme operating condition simulation: In the Simscape physics environment, input a treadmill displacement disturbance of extreme intensity (e.g., acceleration exceeding...). or displacement exceeds (This simulates extreme working conditions where the human body faces the risk of falling.)

[0057] The selection of the actuator is based on: running forward dynamics simulations to record the peak torque and power requirements of the ankle, knee, and hip joints needed to maintain balance. Based on this simulation data, the R&D team accurately determines the rated torque, reduction ratio, and battery power parameters of the motors for each joint of the exoskeleton, avoiding cumbersome equipment due to excessive design margins or safety hazards due to insufficient margins.

[0058] (2) The real-time control strategy of “Assist-as-Needed” (online stage).

[0059] During the exoskeleton's wear and operation phase, the centroid delay feedback algorithm of this invention is embedded into the exoskeleton's onboard controller to achieve human-machine collaborative control: Real-time status awareness. Utilizing the IMU (Inertial Measurement Unit) integrated into the exoskeleton, the wearer's current center of mass (CoM) position is calculated in real time. and speed .

[0060] Ideal bio-torque calculation. The perceived center of mass state is input into the control model of this invention, and a preset neural delay is introduced. (e.g., 100ms) Using co-mapping vectors, the theoretically required "ideal bio-torque" for the human body to restore balance is calculated in real time. :

[0061] Auxiliary torque generation. Based on the principle of "on-demand assistance," the exoskeleton motor's output torque... Set as part of the ideal torque:

[0062] in, ( () represents the auxiliary gain coefficient. For patients in the early stages of rehabilitation with weak muscle strength, a larger value can be set. (e.g., 0.6); gradually decrease as rehabilitation progresses. This forces patients to use their own muscle strength more.

[0063] The assist torque provided by this strategy perfectly matches the natural "ankle-knee-hip" coordination pattern of the human body, avoiding the antagonistic feeling between the machine and the human body in traditional PID control, and significantly improving the comfort of the wearer.

[0064] (3) Patient-specific parameter identification and prescription configuration (calibration stage).

[0065] This simulation system allows for personalized parameter configuration for different patients (e.g., different heights and weights, different degrees of motor dysfunction): Digital modeling: Before the patient's first use, their height and weight are measured, and a unique biomechanical model is generated in the simulation system of this invention.

[0066] Virtual parameter tuning: In a simulation environment, virtual perturbation tests are performed on the patient model to adjust the feedback gain. And passive stiffness parameters, until the simulated recovery trajectory is most stable and the energy consumption is lowest.

[0067] Parameter Download: The optimized control parameters, prepared in the virtual environment, are downloaded to the exoskeleton controller. This is equivalent to prescribing a personalized "exercise prescription" for each patient, achieving precise rehabilitation.

[0068] (4) Active security protection (security monitoring) based on phase plane stability domain.

[0069] Using the centroid phase plane data generated in the simulation of this invention, an active safety protection mechanism for the exoskeleton is constructed: Stability region construction: Through a large number of simulation experiments, the safe convergence region of the human body's center of mass state (displacement-velocity) under different disturbance intensities is determined.

[0070] Fall prediction and intervention: During exoskeleton operation, the wearer's center of mass is mapped onto the phase plane in real time. If the state point is within the safe zone, the exoskeleton executes the "on-demand assistance" mode described above, allowing for some postural fluctuations to train the patient's balance. If the state point crosses the safe boundary (indicating an impending fall), the exoskeleton immediately switches to "high-stiffness safety mode" or "instantaneous high-torque support mode," using knee joint locking or reverse torque output to forcibly pull the body posture back to the safe zone, preventing fall accidents.

[0071] This invention provides a real-time fall risk prediction method based on the above simulation method, comprising: (1) Batch simulation: In the Simscape simulation environment, instead of applying a single treadmill disturbance, a series of different initial centroid states are set.

[0072] State traversal: On the centroid phase plane, select initial points in a meshing manner. For example, location Range Coverage ,speed Range Coverage .

[0073] Simulation execution: For each initial state point, forward dynamics simulation is performed based on the simulation model.

[0074] (2) Convergence determination and stability region calibration: A binary classification determination is performed on each simulation result: Stability: If the trajectory of the center of mass eventually converges to the origin and does not exceed the limits of human anatomy (e.g., the knee joint does not fold in the opposite direction), the initial point is marked as a "safe point".

[0075] Fall: If the trajectory of the center of mass diverges, or the joint angle exceeds the physiological limit, mark the initial point as the "failure point".

[0076] Boundary fitting: All "safe points" are fitted to the phase plane, which can usually be fitted into a closed elliptical region or an irregular polygonal region. This region is the "dynamic stability domain" of the subject under specific neural control ability.

[0077] (3) Fall prediction application: The above-calibrated "dynamic stability domain" is stored in the controller of the exoskeleton or rehabilitation robot as a safety criterion: Real-time monitoring: The device collects the wearer's current center of mass status in real time.

[0078] Calculate the safety margin: Calculate the current state point The shortest Euclidean distance to the boundary of the stable domain is defined as the "Dynamic Stability Margin" (DSM).

[0079] When the DSM (Displacement Strength Scale) is greater than the threshold, the system determines it to be safe. Otherwise, a fall will be inevitable, and the skeletal system will immediately trigger emergency protection strategies, thus significantly advancing the fall prevention window.

[0080] This invention demonstrates that the simulation method can not only be used to "reproduce" experiments, but also to "predict" extreme working conditions, providing a quantitative theoretical tool for the active safety design of rehabilitation aids.

[0081] The effectiveness of the present invention is verified and illustrated below with a specific embodiment: Based on the method of this invention, simulation verification was performed on data from a healthy subject subjected to a backward translational perturbation. Figure 3 The acceleration curves corresponding to platform disturbances are shown below, and the simulation results are as follows: (1) Joint torque fitting analysis as follows Figure 4 As shown in the figure, the solid line represents the torque generated by the simulation, and the dashed line represents the torque measured in the experiment.

[0082] Ankle: The simulation curve accurately reproduces the rapid rise of plantar flexion torque after the start of the disturbance, which is used to resist the forward lean of the body.

[0083] Knee: The simulation curve successfully captured the peak flexion moment at approximately 0.88s, which corresponds to the human body's strategy of lowering its center of gravity by "bending the knee." This is a feature that cannot be reproduced by the traditional single inverted pendulum model.

[0084] Hip joint: The simulated hip joint torque curve and the experimental measurement data show consistency in both amplitude and phase. This verifies that the control strategy proposed in this invention can accurately calculate the posture adjustment commands required to maintain an upright upper body.

[0085] Statistical analysis shows that the determination coefficients of the torques of the three joints ( The average value reached 0.93, proving that the centroid feedback strategy proposed in this invention can accurately predict the timing and amplitude of human force exertion.

[0086] (2) Dynamic stability analysis of the centroid, such as Figure 5 The diagram shown is a phase plane diagram of the center of mass, with the horizontal axis representing the displacement of the center of mass and the vertical axis representing the velocity of the center of mass.

[0087] Experimental data (blue dashed line) shows that after the human body is disturbed, the center of mass undergoes a spiral trajectory that first diverges and then converges, eventually returning to the origin (equilibrium position). The simulation trajectory (solid red line) shows that the constructed simulation model can smoothly return to the stable region and has a spiral convergence rate similar to that of the human body. This indicates that the simulation model has energy dissipation characteristics and dynamic stability consistent with the real human body, verifying the physiological authenticity of the model.

[0088] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for simulating sagittal standing balance dynamics based on centroid delay feedback, characterized by, Includes the following steps: A human sagittal multi-joint biomechanical model was constructed, which includes four rigid body segments: foot, lower leg, thigh, and trunk, as well as three rotational joints: ankle, knee, and hip connecting the rigid body segments. A joint torque generation model is constructed to generate joint torques, which include passive torques and active torques. An active control strategy based on center of mass delay feedback is adopted, using the displacement and velocity deviations of the human body's center of mass as feedback state variables, and introducing neural response delays to calculate the active torques of each joint. In a physical simulation environment, a physical topology model corresponding to the biomechanical model is built. In the physical topology model, the active torques of each joint calculated by the joint torque generation model are used as joint torque sources and connected to each joint to form a complete human sagittal plane standing balance control model, i.e., a simulation model. External disturbance displacement data is input into the simulation model, and the parameters in the simulation model are identified based on the simulated joint torque and the actual joint torque output by the simulation model. A simulation of human sagittal plane standing balance dynamics is achieved based on a simulation model with identified parameters.

2. The center of mass based time-delayed feedback human sagittal plane standing balance dynamics simulation method of claim 1, wherein, The formula for calculating the active torque is: wherein, is the active torque vector, is the current time instant; is the neural reaction delay time; is the offset vector of the center of mass position with respect to the equilibrium state before the delay time instant; is the offset vector of the center of mass velocity with respect to the equilibrium state before the delay time instant; , are the center of mass position feedback gain vector, center of mass velocity feedback gain vector, respectively.

3. The human sagittal plane standing balance dynamics simulation method based on centroid delay feedback as described in claim 2, characterized in that, centroid position feedback gain vector and a centroid velocity feedback gain vector are 3 x 1 coplanar mapping vectors representing the coupled driving effects of centroid state deviation on the ankle, knee, and hip joint torques.

4. The human sagittal plane standing balance dynamics simulation method based on centroid delay feedback as described in claim 2, characterized in that, Neural response latency was set to 100 ms.

5. The center of mass based time-delayed feedback human sagittal plane standing balance dynamics simulation method of claim 2, wherein, External disturbance displacement data is input into the simulation model. Based on the simulated joint torques and actual joint torques output by the simulation model, the parameters in the simulation model are identified, including: The kinematics data and ground reaction force data of a real human body disturbed by a supporting surface are collected through experiments, and real joint torque of the human body is calculated through inverse kinematics ; to minimize the sum of squared errors between simulated joint torques and real joint torques as the objective function. The nonlinear least square algorithm is used to solve the objective function, and the simulation model parameters corresponding to the minimum of the objective function are obtained, including the centroid position feedback gain vector and the centroid velocity feedback gain vector .

6. The center of mass based time-delayed feedback human sagittal plane standing balance dynamics simulation method of claim 5, wherein, The external disturbance displacement data input method is kinematic driving mode. Specifically, the time series data of the treadmill displacement collected in the experiment is directly used as the position driving signal and connected to the sliding pair between the support platform and the ground in the simulation model, driving the support platform to produce horizontal translational motion relative to the ground.

7. The center of mass based time-delayed feedback human sagittal plane standing balance dynamics simulation method according to any one of claims 1-6, wherein, The passive torque is determined by the inherent stiffness and damping of the joint, and the passive torque adopts a time-delayed spring-damped model.

8. The center of mass based time-delayed feedback human sagittal plane standing balance dynamics simulation method of claim 7, wherein, The formula for calculating the passive torque is: wherein, is the passive torque vector, is the current time instant; is the joint passive stiffness coefficient matrix, is the joint passive damping coefficient matrix, , are the deviations of the joint angles, angular velocities, respectively, with respect to the equilibrium position.

9. A center of mass delay feedback based human sagittal plane standing balance dynamics simulation system, characterized in that, Includes a processor, the processor being configured to execute the human sagittal plane standing balance dynamics simulation method based on centroid delay feedback as described in any one of claims 1-8.

10. A collaborative control method for a balance-assisted exoskeleton, characterized in that, The active torque of human joints is simulated using the human sagittal plane standing balance dynamics simulation method based on centroid delay feedback as described in any one of claims 1-8. The human joint torque is used as the target reference value, and the exoskeleton actuator provides auxiliary compensation torques corresponding to the target reference value at the ankle, knee, and hip joints to achieve real-time control of the human centroid deviation.

Images