Self-adaptive exoskeleton control system and method based on model predictive control and finite element modeling

Abstract

The invention discloses a self-adaptive exoskeleton control system and method based on model predictive control and finite element modeling. According to the system, exoskeleton movement data are collected through a sensor module, and signal conditioning and filtering processing are carried out through a data preprocessing module; the finite element FEM modeling module constructs an exoskeleton dynamic model based on the preprocessed data and outputs dynamic update parameters to the model prediction control MPC module, and the model prediction control MPC module generates a control instruction through a rolling optimization control sequence to drive the control execution module; the motion monitoring feedback unit feeds back execution data to the real-time correction module, and a closed loop is formed through error calculation and parameter estimation optimization model. Through modeling and control cooperation, the precision, stability and adaptability of the exoskeleton in a dynamic environment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wearable exoskeleton technology and mechatronic control, and particularly relates to an adaptive exoskeleton control system and method based on model predictive control and finite element modeling. BACKGROUND

[0002] With the acceleration of population aging and industrial automation, the importance of exoskeletons in medical rehabilitation and industrial handling scenarios is increasing. However, existing technologies face many challenges in dynamic environments: 1. Traditional control models are mostly based on simplified assumptions and cannot match sudden changes in human motion intentions or external load changes in real time, resulting in large deviations between control instructions and actual motion needs; 2. Although some systems introduce finite element modeling, they do not coordinate with control algorithms, and the model accuracy cannot meet real-time control requirements. 3. Lack of multi-modal sensor fusion and dynamic parameter updating capability, which easily leads to exoskeleton oscillation or accidental stop.

[0003] To address the above problems, it is necessary to provide an adaptive exoskeleton control system and method based on MPC and finite element FEM modeling. Through the coordinated operation of the sensor module, data preprocessing module, finite element FEM modeling module, model predictive control MPC module, control execution module and real-time correction module, the real-time correction module feeds back the correction parameters to the finite element FEM modeling module, which optimizes the model based on the correction parameters, forming a closed-loop process from data acquisition, processing, modeling to feedback correction, ultimately improving the accuracy, stability and adaptability of the exoskeleton in dynamic environments. SUMMARY

[0004] To address the problems of insufficient model accuracy, poor control real-time performance and weak adaptability of existing exoskeleton systems, the present application provides an adaptive exoskeleton control system and method based on model predictive control and finite element modeling. Through the deep integration of model predictive control and finite element FEM modeling module, accurate prediction and control of exoskeleton motion are achieved, effectively improving the adaptability, control accuracy and stability of the exoskeleton in complex scenarios.

[0005] To achieve the above purpose, the technical solutions of the present application are as follows:

[0006] An adaptive exoskeleton control system based on model predictive control and finite element modeling,

[0007] The adaptive exoskeleton control system comprises a sensor module, a data preprocessing module, a finite element FEM modeling module, a model predictive control MPC module, a control execution module and a real-time correction module.

[0008] The sensor module includes an inertial measurement unit, a force sensor and an electromyography sensor; the data preprocessing module includes a signal conditioning circuit, a data acquisition card and a communication interface; the finite element modeling module includes a model building component, a solver and a post-processing unit; the model predictive control MPC module includes a prediction model, a model predictive control MPC unit and a control instruction generation unit; the control execution module includes a driver and a motion monitoring feedback unit; the real-time correction module includes an error calculation unit, a parameter estimation unit and a feedback control unit.

[0009] An adaptive exoskeleton control method based on model predictive control and finite element modeling, comprising the following steps:

[0010] Step 1, a multi-modal sensor module acquires human-worn exoskeleton related motion data in real time;

[0011] Step 2, the collected data is transmitted to a data preprocessing module for signal conditioning and filtering processing;

[0012] Step 3, the preprocessed data is input into a finite element FEM modeling module and a dynamic model of the exoskeleton is built;

[0013] Step 4, a model predictive control MPC module generates and outputs control instructions based on the updated finite element model and rolling optimization of the control sequence;

[0014] Step 5, a control execution module receives the control instructions and drives the exoskeleton driver to realize action output;

[0015] Step 6, a real-time correction module receives execution data fed back by the control execution module and identifies the finite element model parameters based on a parameter estimation method.

[0016] In step 1, the multi-modal sensor module acquires human-worn exoskeleton joint angle, force feedback and electromyography signal information.

[0017] In step 3, the finite element FEM modeling module generates a real-time dynamic model based on the preprocessed mechanical data through a model building component, performs mesh division and mechanical property calculation using an implicit iterative solver, and outputs updated parameters to the model predictive control MPC module after optimization processing by a post-processing unit.

[0018] In step 4, the model predictive control MPC module generates and outputs control instructions based on the updated finite element model and rolling optimization of the control sequence, specifically including the following steps:

[0019] Step 4.1, the model predictive control MPC module receives the updated finite element model and predicts the future 20-step motion state of the exoskeleton based on a prediction model;

[0020] Step 4.2, dynamically adjust the control input parameters in the rolling time domain based on the MPC algorithm to form a rolling optimization control sequence;

[0021] Step 4.3, the optimized parameters are converted into control instructions by the control instruction generation unit to drive the exoskeleton to perform corresponding actions.

[0022] In step 6, the real-time correction module receives the execution data fed back by the control execution module, and identifies the finite element model parameters based on the parameter estimation method, which specifically includes the following steps:

[0023] Step 6.1, the real-time correction module calculates the root mean square error of the control execution data and the finite element model prediction data through the error calculation unit, and extracts the error information;

[0024] Step 6.2, the parameter estimation unit dynamically adjusts the finite element parameters based on the extended Kalman filter algorithm;

[0025] Step 6.3, the feedback control unit feeds back the estimated parameters to the finite element FEM modeling module, and synchronously updates the weight matrix of the model predictive control MPC module.

[0026] The motion monitoring feedback unit real-time collects the joint torque and contact stress of the exoskeleton driver, and feeds back the data to the real-time correction module through the dynamic weight adjustment mechanism to update the material parameters of the finite element model.

[0027] The FEM modeling module updates the finite element model by using the extended Kalman filter combined with the implicit iteration method, and the update period is ≤5ms; the state equation and the observation equation of the extended Kalman filter are:

[0028] x i+1 =x i +w i ;

[0029] z i =H i x i +v i ;

[0030] In the formula, x i is the system state vector at time i, w i is the process noise, z i is the observation value at time i, v i is the observation noise, and H is the observation matrix.

[0031] The formula of the implicit iteration method is:

[0032]

[0033] In the formula, u tis the displacement at time t, v is the velocity at time t, t is the velocity at time t, a is the acceleration at time t, t is the acceleration at time t, β is a function adjustment parameter, used for adjusting the weight of the acceleration term, β∈[0,1, Δt is a time interval.

[0034] The MPC algorithm is based on a quadratic programming solver, and a rolling optimization objective function is as follows:

[0035]

[0036] In the formula, y is a reference vector, y i is the actual vector of the i-th step, and Q is a weight matrix.

[0037] Compared with the prior art, the present application has the following beneficial effects:

[0038] The model predictive control MPC module and the finite element FEM modeling are deeply integrated in the present application, and a closed-loop control system of multi-module collaborative control is constructed. Specifically, the real-time correction module analyzes and processes the execution data fed back by the control execution module based on an extended Kalman filtering algorithm, transmits the correction parameters to the finite element FEM modeling module, and the finite element FEM modeling module updates the stiffness matrix through an implicit iteration method to form a closed-loop link from execution feedback, parameter correction, model updating to control optimization. The finite element FEM modeling module provides high-precision dynamic parameters for the MPC module at a period of 5 ms, and the model predictive control MPC module generates control instructions based on the dynamic parameters to drive the exoskeleton to accurately perform actions.

[0039] The present application is equipped with a multi-modal sensor module, integrates an inertial measurement unit, a force sensor and an electromyography sensor, realizes multi-source information fusion, and effectively reduces the deviation between the control instructions and the actual movement. The finite element FEM modeling module updates the model by combining the extended Kalman filtering with the implicit iteration method, enhances the adaptability of the exoskeleton to complex environments, avoids the problem of exoskeleton oscillation or accidental emergency stop, and provides more reliable and stable support for medical rehabilitation and industrial carrying scenes. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a structural framework diagram of the adaptive exoskeleton control system based on model predictive control and finite element modeling of the present application.

[0041] Figure 2 is a flowchart of the adaptive exoskeleton control algorithm based on model predictive control and finite element modeling of the present application.

[0042] Figure 3 is a finite element FEM modeling flowchart of the present application.

[0043] Figure 4Flow chart of MPC algorithm of the present application.

[0044] Figure 5 Performance verification chart of adaptive joint control of the present application. DETAILED DESCRIPTION

[0045] The present application is described in more detail below with reference to the accompanying drawings of embodiments of the present application. Figures 1 to 5 The present application is described in more detail below with reference to the accompanying drawings of embodiments of the present application.

[0046] The present application discloses an adaptive exoskeleton control system based on model predictive control and finite element modeling, referring to Figure 1 , comprising the following modules: a sensor module, a data preprocessing module, a finite element FEM modeling module, a model predictive control MPC module, a control execution module, and a real-time correction module.

[0047] The sensor module includes an inertial measurement unit, a force sensor, and an electromyography sensor, arranged at the joint parts, force points, and muscle surfaces of the human body in contact with the exoskeleton; the data preprocessing module includes a signal conditioning circuit, a data acquisition card, and a communication interface, integrated in the exoskeleton control box; the finite element FEM modeling module includes a model construction component, a solver, and a post-processing unit; the model predictive control MPC module includes a prediction model, an MPC unit, and a control instruction generation unit; the control execution module includes a driver and a motion monitoring feedback unit, arranged close to the motor of the exoskeleton; the real-time correction module includes an error calculation unit, a parameter estimation unit, and a feedback control unit, deployed together with the finite element FEM modeling module and the model predictive control MPC module on the data processing unit installed in the exoskeleton control box.

[0048] The present application also discloses an adaptive exoskeleton control method based on model predictive control and finite element modeling, referring to Figure 2 , comprising the following steps:

[0049] Step 1, a multi-modal sensor module acquires human motion data related to wearing an exoskeleton in real time;

[0050] Step 2, the acquired data is transmitted to a data preprocessing module for signal conditioning and filtering processing;

[0051] Step 3, the preprocessed data is input into a finite element FEM modeling module and a dynamic model of the exoskeleton is constructed;

[0052] Step 4, the MPC module generates and outputs control instructions based on the updated finite element model and rolling optimization of the control sequence;

[0053] Step 5, the control execution module receives the control instruction, drives the exoskeleton driver to realize the action output;

[0054] Step 6, the real-time correction module receives the execution data fed back by the control execution module, and identifies the finite element model parameters based on a parameter estimation method.

[0055] In this embodiment, in step 1, the multi-modal sensor module collects the exoskeleton joint angle, force feedback and electromyographic signal information of the human body wearing the exoskeleton at a sampling frequency of 100 Hz.

[0056] In this embodiment, in step 2, the signal conditioning circuit board filters and amplifies the collected data, and then converts the data into digital signals through a data acquisition card and transmits them through a communication interface.

[0057] Reference Figure 3 The finite element FEM modeling module generates a real-time dynamic model based on the preprocessed mechanical data through a model construction component, performs grid division and mechanical property calculation using an implicit iterative solver, and determines whether to converge. If it does not converge, the parameters are adjusted and the solution is recalculated. If it converges, the dynamic update parameters are output to the model predictive control MPC module after optimization processing by a post-processing unit.

[0058] Reference Figure 4 In step 4, the model predictive control MPC module generates and outputs control instructions based on the updated finite element model, including the following steps:

[0059] Step 4.1: The MPC module receives the updated finite element model and predicts the future 20-step motion state of the exoskeleton based on the prediction model.

[0060] Step 4.2: Based on the MPC algorithm, dynamically adjust the control input parameters in the rolling time domain to form a rolling optimization control sequence.

[0061] Step 4.3: Through the control instruction generation unit, the optimized parameters are converted into control instructions to drive the exoskeleton to perform corresponding actions.

[0062] Step 6.1: The real-time correction module calculates the root mean square error of the control execution data and the finite element model prediction data through an error calculation unit, and extracts the error information.

[0063] Step 6.2: When the residual error exceeds 1×10 -6 The parameter estimation unit dynamically adjusts the finite element parameters based on the Kalman filter algorithm.

[0064] Step 6.3: The feedback control unit feeds back the estimated parameters to the finite element FEM modeling module, and synchronously updates the weight matrix of the model predictive control MPC module.

[0065] In the embodiment, the motion monitoring feedback unit collects the joint torque and contact stress of the exoskeleton driver in real time, and feeds the data to the real-time correction module through a dynamic weight adjustment mechanism to update the material parameters of the finite element model.

[0066] In the embodiment, the finite element FEM modeling module updates the finite element model by using the extended Kalman filter combined with the implicit iteration method, and the update period is ≤5 ms.

[0067] The state equation and observation equation of the extended Kalman filter are:

[0068] x i+1 =x i +w i ;

[0069] z i =H i x i +v i ;

[0070] In the formula, x i is the system state vector at time i, w i is the process noise, z i is the observation value at time i, v i is the observation noise, and H is the observation matrix.

[0071] The formula of the implicit iteration method is:

[0072]

[0073] In the formula, u t is the displacement at time t, v t is the velocity at time t, a t is the acceleration at time t, β is a function adjustment parameter for adjusting the weight of the acceleration term, β ∈ [0, 1], and △t is the time interval.

[0074] The MPC algorithm is based on a quadratic programming solver, and the rolling optimization objective function is:

[0075]

[0076] In the formula, y is the reference vector, y i is the actual vector at the i-th step, and Q is the weight matrix.

[0077] Reference Figure 5Is the adaptive joint control performance verification chart based on MPC control and finite element FEM modeling module, including joint angle tracking effect comparison, elastic modulus online estimation result and MPC control torque output, control period is 5ms, the error obtained meets the actual requirement, and the synergy advantage of FEM control and MPC modeling rolling optimization is verified.

[0078] The adaptive exoskeleton control system and method based on model predictive control and finite element modeling combine MPC and FEM modeling, form a closed-loop process from data acquisition, processing, modeling to feedback correction, and optimize the accuracy, stability and adaptability of the exoskeleton in a dynamic environment.

[0079] It should be clear that the above examples are only part of the preferred embodiments of the technical solutions of the present application, which are intended to help understand the technical concept of the present application and do not constitute a limiting specification of the technical solutions. Equivalent substitutions and reasonable improvements made on the basis of the basic purpose of the patent all belong to the legal protection scope defined in the patent claim book.

Claims

1. An adaptive exoskeleton control system based on model predictive control and finite element modeling, characterized by: The adaptive exoskeleton control system includes a sensor module, a data preprocessing module, a finite element (FEM) modeling module, a model predictive control (MPC) module, a control execution module, and a real-time correction module. The sensor module includes an inertial measurement unit, a force sensor and an electromyographic sensor; the data preprocessing module includes a signal conditioning circuit, a data acquisition card and a communication interface; the finite FEM modeling module includes a model building component, a solver and a post-processing unit; the model predictive control MPC module includes a prediction model, a model predictive control MPC unit and a control instruction generation unit; the control execution module includes a driver and a motion monitoring feedback unit; and the real-time correction module includes an error calculation unit, a parameter estimation unit and a feedback control unit.

2. An adaptive exoskeleton control method based on model predictive control and finite element modeling, characterized in that: The following steps are involved: Step 1: The multimodal sensor module collects relevant motion data of the human body wearing the exoskeleton in real time; Step 2: The collected data is transmitted to the data preprocessing module for signal conditioning and filtering; Step 3: The pre-processed data is input into the finite element (FEM) modeling module to construct the exoskeleton dynamic model; Step 4: The model predictive control (MPC) module performs rolling optimization of the control sequence based on the updated finite element model, and generates and outputs control instructions. Step 5: The control execution module receives the control instruction and drives the exoskeleton driver to achieve action output; Step 6: The real-time correction module receives the execution data fed back by the control execution module and identifies the finite element model parameters based on the parameter estimation method.

3. The adaptive exoskeleton control method based on model predictive control and finite element modeling according to claim 2, characterized in that: In step 1, the multimodal sensor module collects the joint angle, force feedback and electromyographic signal information of the exoskeleton worn by the human body.

4. The adaptive exoskeleton control method based on model predictive control and finite element modeling according to claim 2, characterized in that: In step 3, the finite element (FEM) modeling module generates a real-time dynamic model based on the pre-processed mechanical data through the model building component, uses an implicit iterative solver to perform meshing and mechanical property calculations, and outputs updated parameters to the model predictive control (MPC) module after optimization processing by the post-processing unit.

5. The adaptive exoskeleton control method based on model predictive control and finite element modeling according to claim 2, characterized in that: In step 4, the model predictive control (MPC) module performs rolling optimization of the control sequence based on the updated finite element model, generates and outputs control instructions, and specifically includes the following steps: Step 4.1: The model predictive control (MPC) module receives the updated finite element model and predicts the exoskeleton's motion state for the next 20 steps based on the prediction model. Step 4.2: Dynamically adjust the control input parameters in the rolling time domain based on the MPC algorithm to form a rolling optimization control sequence; Step 4.3: The control instruction generation unit converts the optimized parameters into control instructions to drive the exoskeleton to perform corresponding actions.

6. The adaptive exoskeleton control method based on model predictive control and finite element modeling according to claim 2, characterized in that: In step 6, the real-time correction module receives the execution data fed back by the control execution module and identifies the finite element model parameters based on the parameter estimation method, which specifically includes the following steps: Step 6.1, the real-time correction module calculates the root mean square error between the control execution data and the finite element model prediction data through the error calculation unit, and extracts the error information; Step 6.2: The parameter estimation unit dynamically adjusts the finite element parameters based on the extended Kalman filter algorithm; Step 6.3: The feedback control unit feeds back the estimated parameters to the finite element (FEM) modeling module and simultaneously updates the weight matrix of the model predictive control (MPC) module.

7. The adaptive exoskeleton control method based on model predictive control and finite element modeling according to claim 2, characterized in that: The motion monitoring feedback unit collects the joint torque and contact stress of the exoskeleton actuator in real time, and feeds the data back to the real-time correction module through a dynamic weight adjustment mechanism to update the material parameters of the finite element model.

8. The adaptive exoskeleton control method based on model predictive control and finite element modeling according to claim 4, characterized in that: The FEM modeling module uses an extended Kalman filter combined with an implicit iterative method to update the finite element model, and the update period is ≤5ms; the state equation and observation equation of the extended Kalman filter are: x i+1 =x i +w i ; z i =H i x i +v i ; Where x i is the system state vector at time i, w i is the process noise, z i is the observation value at time i, v i is the observation noise, H is the observation matrix; The formula for the implicit iteration method is: Where u t is the displacement at time t, v t is the speed at time t, a t is the acceleration at time t, β is the function adjustment parameter used to adjust the weight of the acceleration term, β∈[0,1], and Δt is the time interval.

9. The adaptive exoskeleton control method based on model predictive control and finite element modeling according to claim 5, characterized in that: The MPC algorithm is based on a quadratic programming solver, with a rolling optimization objective function: Where y is the reference vector, y i is the actual vector of the i-th step, and Q is the weight matrix.

Images