CN114170679A - Walking aid continuous gait phase estimation control method based on self-adaptive oscillator - Google Patents
Walking aid continuous gait phase estimation control method based on self-adaptive oscillator Download PDFInfo
- Publication number
- CN114170679A CN114170679A CN202111389744.3A CN202111389744A CN114170679A CN 114170679 A CN114170679 A CN 114170679A CN 202111389744 A CN202111389744 A CN 202111389744A CN 114170679 A CN114170679 A CN 114170679A
- Authority
- CN
- China
- Prior art keywords
- oscillator
- gait
- aos
- convergence
- phase
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/112—Gait analysis
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1121—Determining geometric values, e.g. centre of rotation or angular range of movement
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7271—Specific aspects of physiological measurement analysis
- A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H3/00—Appliances for aiding patients or disabled persons to walk about
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H3/00—Appliances for aiding patients or disabled persons to walk about
- A61H2003/007—Appliances for aiding patients or disabled persons to walk about secured to the patient, e.g. with belts
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
- A61H2201/12—Driving means
- A61H2201/1207—Driving means with electric or magnetic drive
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
- A61H2201/16—Physical interface with patient
- A61H2201/1602—Physical interface with patient kind of interface, e.g. head rest, knee support or lumbar support
- A61H2201/164—Feet or leg, e.g. pedal
- A61H2201/1642—Holding means therefor
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
- A61H2201/16—Physical interface with patient
- A61H2201/1602—Physical interface with patient kind of interface, e.g. head rest, knee support or lumbar support
- A61H2201/165—Wearable interfaces
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
- A61H2201/50—Control means thereof
- A61H2201/5007—Control means thereof computer controlled
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2205/00—Devices for specific parts of the body
- A61H2205/10—Leg
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2230/00—Measuring physical parameters of the user
- A61H2230/62—Posture
- A61H2230/625—Posture used as a control parameter for the apparatus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Physiology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Dentistry (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Algebra (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Epidemiology (AREA)
Abstract
The invention discloses a walking aid continuous gait phase estimation control method, in particular to a walking aid continuous gait phase estimation control method based on a self-adaptive oscillator, which adopts an AOs algorithm model, an acceleration convergence module, a gait phase estimation detection module, a gait event alignment module and a power assisting control module; AOs model is composed of a group of Hopf oscillator units with omega (t) as fundamental frequency; the acceleration convergence module accelerates the convergence speed of the AO algorithm under periodic and quasi-periodic gait signals; the gait phase estimation detection module provides a detection method and a corresponding mechanism for the gait phase estimation failure of the AOs algorithm; the gait event alignment module proposes AOs a smoothing solution that has an estimated phase corresponding to the gait event; the power assist control module provides a predefined power assist value based on the currently estimated gait phase. The invention solves the technical problem that the power-assisted control of the walking aid is better suitable for daily activities. The safety and the reliability of the assistance of the walking aid are ensured.
Description
Technical Field
The invention relates to a walking aid continuous gait phase estimation control method, in particular to a walking aid continuous gait phase estimation control method based on a self-adaptive oscillator.
Background
The wearable exoskeleton robot is used for assisting the old or the cerebral apoplexy patients to realize walking exercise and rehabilitation training and improving the walking ability of the old or the cerebral apoplexy patients, and is an effective means for treating the group movement function decline or disorder. The exoskeleton auxiliary exercise and training mode with the assistance of the lower limb joints can compensate joint driving force lacking in the lower limbs in the walking motion process, and meanwhile, the safety of the motion process is guaranteed.
At present, the research on the power-assisted control strategy of the walking aid at home and abroad can be mainly divided into three ideas: firstly, discrete gait events (such as a swing phase and a support phase) are switched through a State transition condition based on a Finite State Machine (FSM) method, and a corresponding power control strategy is adopted in each gait event, for example, a patent with the patent number of cn202011088505.x uses an IMU on the thigh side to detect the moment when the heel lifts off; patent No. CN201310009489.4 uses the lower limb joint information to determine the swing phase and the support phase. Secondly, the gait is intercepted in the whole period based on an oscillator (Oscillators), and the gait phase ratio identification from 0 to 100 percent is carried out in each period. And thirdly, estimating a gait phase value based on a machine learning method of big data. In the three methods, the hip joint power-assisted control is carried out on the walking aid by acquiring the current walking human body kinematics data and processing the data to obtain the power-assisted control quantity.
However, with the adoption of a wide FSM (finite state machine) power-assisted control method, all states of the walking aid must be mastered in advance, the current state is judged according to the sensor, power-assisted control is provided, and once an unknown state occurs, the algorithm is disordered to cause a safety problem; the most advanced machine learning method needs to collect a large amount of data covering all use scenes, and the model belongs to a black box and cannot be theoretically explained. The Oscillator algorithm with many advantages is basically divided into two types, namely an Adaptive Oscillator (AOs) and a Phase Oscillator (PO), wherein the PO performs Phase estimation through a gait angle and an angular velocity, but the obtained Phase angle has large nonlinear distortion, discomfort can be caused during walking assistance, and a shaking problem exists during the process that the angle of the lower limb of a human body is 0. AOs belongs to an iterative algorithm, which can predict the periodic and quasi-periodic motion of human body, but has the condition of difficult convergence under non-periodic motion and slow convergence speed under periodicity; secondly AOs easily converge into a periodic signal component that is a multiple of the gait cycle; AOs when the human body moves, it will not be converged or the convergence is disturbed when it is converted from one class period/periodic motion to another class period/periodic motion with different frequency; finally AOs there is a certain random offset between the gait phase and the actual human motion state, and the random offset can not be predicted in advance.
Disclosure of Invention
The invention aims to solve the defects of the prior art and provide a walking aid continuous gait phase estimation control method based on an adaptive oscillator.
Specifically, the control method adopts an AOs algorithm model, an acceleration convergence module, a gait phase estimation detection module, a gait event alignment module and a power assisting control module.
AOs model consists of a set of Hopf Oscillator (HO) units with ω (t) as the fundamental frequency, and the formula is as follows:
mu is the amplitude of the oscillator, omega is the frequency characteristic of the oscillator, x and y are the coordinate values of the oscillator in a Cartesian coordinate system respectively,v is the corresponding coupling coefficient for the periodic input outside the oscillator. When the external θ (t) frequency coincides with the oscillator frequency, a phenomenon similar to "resonance" occurs, and the coordinates (x, y) will continue to move around the origin at an angular velocity of ω.
Will be provided withThe periodic input signal is regarded as To ensure that the oscillator has the ability to be synchronised to external disturbances, i.e.A new state variable ω (t) is introduced into the Hopf oscillator to form an adaptive HO oscillator, and feedback is introduced so that the oscillator "resonates" with any periodic input signal. While for HO to input signal amplitudeAnd phaseAlso introduces the corresponding state variable alpha (t),into HO. The output of the HO oscillator is expressed as
Wherein the content of the first and second substances,in correspondence with the phase of the HO,corresponding to the amplitude of the Hopf oscillator.Is an input gait signal thetar(t) output in synchronism with the oscillatorAs a feedback quantity.νω,vηRepresenting the feedback learning parameters in the AOs model, determines the convergence rate of the AOs model synchronous input signal. The feedback control ensures that the state variable of the oscillator is maintained for a limited timeOmega, alpha can converge to parameters of the externally coupled input signal
An oscillation pool, the AOs algorithm, is formed using a plurality of the above-described adaptive HO oscillator units. AOs each HO oscillator contains amplitude alpha (t), frequency i omega (t) and phaseThree state parameters, wherein i is the number of HO oscillators, if the number of oscillators isn, i is 0. By learning the characteristics of the input period and the period-like signal, the AO has the ability to dynamically synchronize the input signal:
wherein the content of the first and second substances,is to input a gait signalThe error of θ (t) is output in synchronization with the oscillator as a feedback amount.
The acceleration convergence module is used for improving the problem that the convergence speed of the AO algorithm is low under periodic and quasi-periodic gait signals. It dynamically estimates the period of the input gait signal in real time, and accordingly sets AOs the learning parameters of the algorithm to an optimal value, achieving a faster convergence speed.
To simplify the analysis, consider the adaptive HO oscillator equation (2),
assume AOs has approached a stable value, i.e., converged to the input signal's characteristics:obtaining:
is provided with
Substituting e (t) into the formula (2) to obtain the input and output state equation with the input beingThe output is ω, α:
Bω=[1 0 0 0]T
assuming that the HO oscillator has approached the convergence value at this time, then Using the principle of averagingThe averaging technique simplifies the results of equation (7) to yield:
the amplitude transfer function of the HO adaptive oscillator is:
the time constant of the transfer function is:
τα=2/η (10)
the larger η, the smaller the time constant, and the faster the convergence speed.
The frequency transfer function of the HO adaptive oscillator is:
selecting
The above formula (11) can be simplified to
A step response constant of
Based on the relations (10), (12), (14) of the feedback learning parameters and the convergence time constants, the optimum learning parameters can be found to achieve the desired convergence time τωAnd τα. Theoretically, the convergence time constant τ is secured under the convergence conditionωAnd ταThe smaller the better. Because two hypothesis assumptions are introduced, 1. hypothesis AOs has approached a stable value, i.e., converged to the characteristics of the input signal; 2.simplified toIs assumed to be averaged. The above relation has a certain error. In practice, it has been found that the convergence constant is at 0.2T ≦ τα≤5T,0.2T≤τωThe above formula can better meet the AOs convergence characteristic and can successfully converge to the external signal. Thus taking τα=0.2T,τω=0.2T。
The optimal learning parameters finally derived are as follows:
v 'therein'η,ν′ωAnd are andis the optimal learning parameter; tau isα τωThe convergence time constants of amplitude and frequency, respectively, and T is the estimated gait cycle.
The gait phase estimation detection module provides a detection method and a corresponding mechanism for the gait phase estimation failure of the AOs algorithm: the state parameter amplitude alpha (t), frequency omega (t) · i and phase of each oscillator are obtained by estimating AOs algorithmAnd three state parameters are used for predicting future gait signal parameters, comparing the actual semaphore with the predicted semaphore error, and if the actual semaphore exceeds a set threshold, indicating that the gait phase estimation of the AOs algorithm has larger error and wrong phase estimation. While avoiding AOs convergence to other frequency components, α1The proportion of (t) to the amplitude of the actual signal also needs to be up to 50%. After the module detects the problem, the initial value of the oscillator and the learning parameter are automatically endowed again, wherein the initial value of the phase and the amplitude of the oscillator is randomly endowed, the initial value of the frequency is the estimated frequency of the input signal, and the learning parameter is the optimal parameter in the acceleration convergence module. To achieve a secure and robust estimation result. In order to prevent frequent reassignment of AOs new values of the model parameters, so that the interval between two adjacent assignment operations is larger than the set time interval, a timer timing method is adopted. After each assignment operation, the timer is started, and the assignment operation cannot be repeated until the timer reaches the set time.
The gait event alignment module proposes AOs a smoothing solution that has an estimated phase corresponding to the gait event. In the practice of the algorithm of AOs, since AOs does not limit the sign of the ω (t) state within AOs, it often occurs that ω (t) is negative, resulting in a calculated ω (t) valueA decrease from 100% to 0% may occur. Therefore, the human gait phase estimated from the AOs convergence value needs to be considered separately for the positive and negative frequencies ω (t).
Generally, the human gait is initially from the toe-off point, and the gait phase at this moment is set to 0%. An estimated phase value is then obtained for each gait cycle at that momentI.e. the phase that needs to be compensated. The real-time online error compensation method ensuresSmoothness of estimated gait phase:
in order to learn the error compensation amount on-line,to compensate for the gait phase, subsequent walking assistance may be used. The Filter may employ IIR and FIR filters.
The power control module provides a predefined power value to the wearer according to the currently estimated gait phase, so as to realize power assistance in daily activities.
The walking aid control method based on the self-adaptive oscillator provided by the invention can realize rapid convergence on the basis of keeping the advantages of AOs, such as smooth power assistance, simple sensor arrangement and the like. Secondly, the fast convergence can be realized in the switching process of different daily periodic movements of the human body. Finally, the invention can realize the judgment of AOs whether the walking aid works normally or not through the prediction error, thereby ensuring the safety and reliability of the assistance of the walking aid.
Drawings
FIG. 1 is a system diagram of the present control method;
FIG. 2 is a schematic diagram of a sensor arrangement;
FIG. 3 is a schematic diagram of a walker robotic system architecture;
FIG. 4 is an experimental photograph of a walking-aid robot;
FIG. 5 is a comparison between the conventional AOs algorithm and the experimental results of gait phase estimation control proposed in this patent (convergence speed) in case of uniform walking; the vertical dotted line is the corresponding convergence time;
FIG. 6 is a safety detection and response result of the gait phase estimation algorithm in daily activities;
fig. 7 is a graph of the alignment effect of the estimated gait phase with the actual gait event under accelerated and decelerated walking.
Detailed Description
The invention discloses a walking aid continuous gait phase estimation control method based on a self-adaptive oscillator, which adopts an AOs algorithm model, an acceleration convergence module, a gait phase estimation detection module, a gait event alignment module and a power assisting control module, and is shown in figure 1.
AOs model consists of a set of Hopf Oscillator (HO) units with ω (t) as the fundamental frequency, and the formula is as follows:
mu is the amplitude of the oscillator, omega is the frequency characteristic of the oscillator, x and y are the coordinate values of the oscillator in a Cartesian coordinate system respectively,v is the corresponding coupling coefficient for the periodic input outside the oscillator. When the external θ (t) frequency coincides with the oscillator frequency, a phenomenon similar to "resonance" occurs, and the coordinates (x, y) will continue to move around the origin at an angular velocity of ω.
Will be provided withThe periodic input signal is regarded as To ensure that the oscillator has the ability to be synchronised to external disturbances, i.e.A new state variable ω (t) is introduced into the Hopf oscillator to form an adaptive HO oscillator, and feedback is introduced so that the oscillator "resonates" with any periodic input signal. While for HO to input signal amplitudeAnd phaseAlso introduces the corresponding state variable alpha (t),into HO. The output of the HO oscillator is expressed as
Wherein the content of the first and second substances,in correspondence with the phase of the HO,corresponding to the amplitude of the Hopf oscillator.Is an input gait signal thetar(t) output in synchronism with the oscillatorAs a feedback quantity.νω,νηRepresenting the feedback learning parameters in the AOs model, determines the convergence rate of the AOs model synchronous input signal. The feedback control ensures that the state variable of the oscillator is maintained for a limited timeOmega, alpha can converge to parameters of the externally coupled input signal
An oscillation pool, the AOs algorithm, is formed using a plurality of the above-described adaptive HO oscillator units. AOs each HO oscillator contains amplitude alpha (t), frequency i omega (t) and phaseAnd three state parameters, wherein i is the sequence number of the HO oscillator, and if the number of the oscillators is n, i is 0. By learning the characteristics of the input period and the period-like signal, the AO has the ability to dynamically synchronize the input signal:
wherein the content of the first and second substances,is to input a gait signalThe error of θ (t) is output in synchronization with the oscillator as a feedback amount.
The acceleration convergence module is used for improving the problem that the convergence speed of the AO algorithm is low under periodic and quasi-periodic gait signals. It dynamically estimates the period of the input gait signal in real time, and accordingly sets AOs the learning parameters of the algorithm to an optimal value, achieving a faster convergence speed.
To simplify the analysis, consider the adaptive HO oscillator equation (2),
assume AOs has approached a stable value, i.e., converged to the input signal's characteristics:obtaining:
is provided with
Substituting e (t) into the formula (2) to obtain the input and output state equation with the input beingThe output is ω, α:
Bω=[1 0 0 0]T
assuming that the HO oscillator has approached the convergence value at this time, then The results of equation (7) are simplified using the averaging technique to yield:
the amplitude transfer function of the HO adaptive oscillator is:
the time constant of the transfer function is:
τα=2/η (10)
the larger η, the smaller the time constant, and the faster the convergence speed.
The frequency transfer function of the HO adaptive oscillator is:
selecting
The above formula (11) can be simplified to
A step response constant of
Based on the relations (10), (12), (14) of the feedback learning parameters and the convergence time constants, the optimum learning parameters can be found to achieve the desired convergence time τωAnd τα. Theoretically, the convergence time constant τ is secured under the convergence conditionωAnd ταThe smaller the better. Because two hypothesis assumptions are introduced, 1. hypothesis AOs has approached a stable value, i.e., converged to the characteristics of the input signal; 2.simplified toIs assumed to be averaged. The above relation has a certain error. In practice, it has been found that the convergence constant is at 0.2T ≦ τα≤5T,0.2T≤τωThe above formula can better meet the AOs convergence characteristic and can successfully converge to the external signal. Thus taking τα=0.2T,τω=0.2T。
The optimal learning parameters finally derived are as follows:
wherein v'η,ν′ωAnd are andis the optimal learning parameter; tau isα τωThe convergence time constants of amplitude and frequency, respectively, and T is the estimated gait cycle.
The gait phase estimation detection module provides a detection method and a corresponding mechanism for the gait phase estimation failure of the AOs algorithm: the state parameter amplitude alpha (t), frequency omega (t) · i and phase of each oscillator are obtained by estimating AOs algorithmAnd three state parameters are used for predicting future gait signal parameters, comparing the actual semaphore with the predicted semaphore error, and if the actual semaphore exceeds a set threshold, indicating that the gait phase estimation of the AOs algorithm has larger error and wrong phase estimation. While avoiding AOs convergence to other frequency components, α1The proportion of (t) to the amplitude of the actual signal also needs to be up to 50%. After the module detects the problem, the initial value of the oscillator and the learning parameter are automatically endowed again, wherein the initial value of the phase and the amplitude of the oscillator is randomly endowed, the initial value of the frequency is the estimated frequency of the input signal, and the learning parameter is the optimal parameter in the acceleration convergence module. To achieve a secure and robust estimation result. In order to prevent frequent reassignment of AOs new values of the model parameters, so that the interval between two adjacent assignment operations is larger than the set time interval, a timer timing method is adopted. After each assignment operation, the timer is started, and the assignment operation cannot be repeated until the timer reaches the set time.
The gait event alignment module proposes AOs a smoothing solution that has an estimated phase corresponding to the gait event. In the practice of the algorithm of AOs, since AOs does not limit the sign of the ω (t) state within AOs, it often occurs that ω (t) is negative, resulting in a calculated ω (t) valueA decrease from 100% to 0% may occur. Therefore, the human gait phase estimated from the AOs convergence value needs to be considered separately for the positive and negative frequencies ω (t).
Generally, the human gait is initially from the toe-off point, and the gait phase at this moment is set to 0%. An estimated phase value is then obtained for each gait cycle at that momentI.e. the phase that needs to be compensated. The real-time online error compensation method ensures the smoothness of the estimated gait phase as follows:
in order to learn the error compensation amount on-line,to compensate for the gait phase, subsequent walking assistance may be used. The Filter may employ IIR and FIR filters.
The power control module provides a predefined power value to the wearer according to the currently estimated gait phase, so as to realize power assistance in daily activities.
The following are the application examples of the walking-aid robot adopting the gait phase estimation control method provided by the invention, and the comparison of the traditional AOs algorithm and the experimental results of the gait phase estimation control provided by the invention:
as shown in figures 2 and 3, the walking-aid robot prototype system mainly comprises a mechanical body structure, a control circuit module, a binding belt part, a driving joint module, a motion sensing module and an energy supply module. Wherein the mechanical body structure consists of a waist structure, a hip joint flexion and extension structure and a thigh rod; the back of the robot is provided with a microprocessor control circuit, so that the processing of sensing information and the calculation of a power value are realized, and a control command of left and right hip joint moments is output; a waist binding belt is arranged on the waist structure to realize the fixed connection between the waist of the robot and the waist of a human body, and thigh binding belts are arranged on the thigh rods to realize the fixed connection between the leg rods of the robot and the thighs of the human body so as to transfer the assistance; the drive joint module is composed of a Maxon EC 90flat servo motor and a secondary planetary gear reducer, and the motor driver adopts an ESCON 50/5 module to realize torque control; the motion perception module mainly comprises an IMU (inertial measurement Unit) arranged on the lower side of a thigh bandage and a pressure sensor arranged in the thigh bandage, and is used for respectively measuring hip joint flexion and extension angle information and human-computer force interaction information; the power supply module is a 24V lithium battery and respectively supplies power to the motor driver and the microprocessor (24V to 5V). The microprocessor realizes wireless data transmission with the upper computer through the Bluetooth module. The whole algorithm is arranged in the microprocessor STM32F 103. In order to compensate the gait phase of the human heel to 0% at the time of landing, a pressure sensor is arranged on the sole of the human body to detect the landing time of the human heel.
The swing angle information of the thigh is processed by the algorithm, and the estimation of the gait cycle is obtained by detecting the peak value interval of the angle curve of the adjacent hip joint; the filter () function for smoothly aligning the gait phases uses a first order butterworth filter. The algorithm can obtain the phase angle in the walking processAnd the reciprocating oscillation changes in each gait cycle, so that the phase angle is used as the calculation input of the power assisting moment, a sine function is selected for normalization processing, and the adjustment of the power assisting peak value is realized through the power assisting adjusting coefficient kappa. And if the algorithm predicts that the future angle error exceeds a certain threshold, different common assistance forces adopt a following mode. The boost equation can be expressed as:
the wearer walks at a constant speed, accelerates and decelerates and daily activities while wearing the walker device, as shown in fig. 4. And (4) acquiring IMU signals in real time, and performing gait phase estimation and power-assisted operation. The results of the conventional AOs algorithm and the gait phase estimation control experiment proposed by the invention are shown in fig. 5-7. Fig. 5 shows a comparison result between the conventional AOs algorithm under constant walking and the gait phase estimation control experiment proposed in this patent, which shows that the algorithm can obtain the optimal convergence speed after being optimized, and the convergence can be achieved after two steps. Fig. 6 shows the safety detection and handling results of the gait phase estimation algorithm in daily activities, and it can be seen that in the process of running-rest-re-walking, running and walking are both cycle-like activities, but AOs algorithm is difficult to converge in the process of walking, and the algorithm is more suitable for the daily activity state. The method can obviously find that the prediction error is larger when the algorithm is not converged and the phase estimation is wrong, and can be used for detecting the accuracy and safety of gait phase estimation of the algorithm and determining the time point of initializing internal parameters, thereby ensuring safer assistance. Fig. 7 is a graph showing the alignment effect of the estimated gait phase and the heel strike time actually detected by the pressure sensor in the acceleration-deceleration walking.
Claims (1)
1. A walking aid continuous gait phase estimation control method based on a self-adaptive oscillator is characterized by comprising an AOs algorithm model, an acceleration convergence module, a gait phase estimation detection module, a gait event alignment module and a power assisting control module;
the AOs model is composed of a group of Hopf oscillator units with ω (t) as fundamental frequency, and the formula is as follows:
wherein mu is the amplitude of the oscillator, omega is the frequency characteristic of the oscillator, x and y are coordinate values of the oscillator in a Cartesian coordinate system respectively,is a periodic input external to the oscillator, v is the corresponding coupling coefficient; when the external theta (t) frequency is consistent with the frequency of the oscillator, a phenomenon similar to resonance occurs, and the coordinates (x, y) continuously move and oscillate around the origin at the angular speed of omega;
will be provided withThe periodic input signal is regarded asTo ensure that the oscillator has the ability to be synchronised to external disturbances, i.e.Introducing a new state variable omega (t) into the Hopf oscillator to form an adaptive Hopf oscillator, and introducing feedback so that the oscillator and any periodic input signal generate resonance; at the same time, to realize the amplitude of input signal by Hopf oscillatorAnd phaseSynchronization/convergence of, also introducing corresponding state variablesInto a Hopf oscillator; the output of the Hopf oscillator is represented as
Wherein the content of the first and second substances,corresponding to the phase of the Hopf oscillator,corresponding to the amplitude of the Hopf oscillator,is an input gait signal thetar(t) output in synchronism with the oscillatorAs a feedback quantity,vω,vηrepresenting AOs feedback learning parameters in the model, and determining AOs convergence rate of the synchronous input signal of the model; the feedback control ensures that the state variable of the oscillator is maintained for a limited timeOmega, alpha can converge to parameters of the externally coupled input signal
Forming an oscillation pool by adopting a plurality of self-adaptive Hopf oscillator units, namely AOs algorithm; AOs each Hopf oscillator contains an amplitude α (t), a frequency i ω (t) and a phaseThree state parameters, wherein i is the serial number of the Hopf oscillator, and if the number of the oscillators is n, i is 0.. n-1; by learning the characteristics of the input period and the period-like signal, the AO has the ability to dynamically synchronize the input signal:
wherein the content of the first and second substances,is to input a gait signalOutput in synchronism with oscillatorAs a feedback quantity;
wherein the convergence acceleration module dynamically estimates the period of the input gait signal in real time, and accordingly sets AOs the learning parameters of the algorithm to an optimal value to achieve a faster convergence speed, specifically,
to simplify the analysis, consider the adaptive HO oscillator equation (2),
assume AOs has approached a stable value, i.e., converged to the input signal's characteristics:obtaining:
is provided with
Substituting e (t) into the formula (2) to obtain the input and output state equation with the input beingThe output is ω, α:
Bω=[1 0 0 0]T
assuming that the Hopf oscillator is already close to the convergence value at this time, then The results of equation (7) are simplified using the averaging technique to yield:
the amplitude transfer function of the Hopf adaptive oscillator is:
the time constant of the transfer function is:
τα=2/η (10)
wherein, the larger eta is, the smaller the time constant is, and the faster the convergence speed is;
the frequency transfer function of the Hopf adaptive oscillator is:
selecting
The above equation (11) is simplified as:
the step response constant is:
based on the relations (10), (12), (14) of the feedback learning parameters and the convergence time constants, the optimum learning parameters can be found to achieve the desired convergence time τωAnd τα(ii) a Theoretically, under the condition of ensuring convergenceConvergence time constant τωAnd ταThe smaller the better; because two hypothesis assumptions are introduced, 1. hypothesis AOs has approached a stable value, i.e., converged to the characteristics of the input signal; 2.simplified toAverage hypothesis of (d); the above relation has certain error; in practice, it has been found that the convergence constant is at 0.2T ≦ τα≤5T,0.2T≤τωThe above formula can better meet the AOs convergence characteristic and can successfully converge to the characteristic of an external signal when the T is less than or equal to 5T; thus taking τα=0.2T,τω=0.2T;
The optimal learning parameters finally derived are as follows:
wherein vη′,νω', andis the optimal learning parameter; tau isα τωConvergence time constants of amplitude and frequency, respectively, T being the estimated gait cycle;
the gait phase estimation detection module estimates and obtains the state parameter amplitude alpha (t), the frequency omega (t) · i and the phase of each oscillator through AOs algorithmThree state parameters are used for predicting future gait signal parameters, the actual semaphore is compared with the predicted semaphore error, and if the actual semaphore exceeds a set threshold value, the gait phase estimation of the AOs algorithm has a large error and the phase estimation is wrong; while avoiding AOs convergence to other frequency components, α1(t) accounting for actual signalThe proportion of the amplitude also needs to reach a proportion of 50%; after the module detects the problem, the initial values of the phase and the amplitude of the oscillator are automatically endowed again, wherein the initial values of the phase and the amplitude of the oscillator are randomly endowed, the initial values of the frequency are the estimated frequency of the input signal, and the learning parameters are the optimal parameters in the accelerated convergence module, so that the safe and stable estimation result is realized; in order to prevent frequently endowing AOs model parameters with new values again, so that the interval between two adjacent assignment operations is larger than the set time interval, a timer timing method is adopted, the timer is started after each assignment operation, and the assignment operation cannot be repeated until the timer reaches the set time;
the gait event alignment module ensures the smoothness of the estimated gait phase; in the practice of the algorithm of AOs, since AOs does not limit the sign of the ω (t) state within AOs, it often occurs that ω (t) is negative, resulting in a calculated ω (t) valueThe situation of reducing from 100% to 0% occurs, so the human gait phase estimated from the AOs convergence value needs to be considered according to the positive and negative of the frequency ω (t);
generally, the human gait is defaulted to be started at the toe-off moment, the gait phase at the moment is set to be 0%, and then the estimated phase value at the moment of each gait cycle is acquiredI.e. the phase that needs to be compensated; the real-time online error compensation method comprises the following steps:
in order to learn the error compensation amount on-line,for the compensated gait phase, subsequent walking assistance can be used, and an IIR (infinite impulse response) Filter and an FIR (finite impulse response) Filter are adopted by a Filter;
the power control module provides a predefined power value to the wearer according to the currently estimated gait phase, so as to realize power assistance in daily activities.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111389744.3A CN114170679A (en) | 2021-11-22 | 2021-11-22 | Walking aid continuous gait phase estimation control method based on self-adaptive oscillator |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111389744.3A CN114170679A (en) | 2021-11-22 | 2021-11-22 | Walking aid continuous gait phase estimation control method based on self-adaptive oscillator |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114170679A true CN114170679A (en) | 2022-03-11 |
Family
ID=80480171
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111389744.3A Withdrawn CN114170679A (en) | 2021-11-22 | 2021-11-22 | Walking aid continuous gait phase estimation control method based on self-adaptive oscillator |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114170679A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024029606A1 (en) * | 2022-08-05 | 2024-02-08 | 株式会社国際電気通信基礎技術研究所 | Model generation method, model generation device, phase estimation method, control method, and control device |
-
2021
- 2021-11-22 CN CN202111389744.3A patent/CN114170679A/en not_active Withdrawn
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024029606A1 (en) * | 2022-08-05 | 2024-02-08 | 株式会社国際電気通信基礎技術研究所 | Model generation method, model generation device, phase estimation method, control method, and control device |
JP7433561B1 (en) | 2022-08-05 | 2024-02-19 | 株式会社国際電気通信基礎技術研究所 | Model generation method, model generation device, phase estimation method, control method, and control device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109223444B (en) | Rehabilitation robot and man-machine cooperative interaction force control method thereof | |
CN108392795A (en) | A kind of healing robot Multimode Controlling Method based on Multi-information acquisition | |
CN110834329B (en) | Exoskeleton control method and device | |
CN110251898B (en) | Closed-loop control system of wrist rehabilitation exoskeleton based on force feedback | |
CN109330819B (en) | Master-slave type upper limb exoskeleton rehabilitation robot control system and control method thereof | |
CN106110587B (en) | lower limb exoskeleton rehabilitation system and method based on man-machine cooperation | |
CN109223453B (en) | Power-assisted exoskeleton device based on regular walking gait learning | |
CN113633521B (en) | Upper limb exoskeleton rehabilitation robot control system and control method | |
CN114170679A (en) | Walking aid continuous gait phase estimation control method based on self-adaptive oscillator | |
CN112842824B (en) | Training method for rehabilitation and recovery of lower limbs | |
Yan et al. | A novel adaptive oscillators-based control for a powered multi-joint lower-limb orthosis | |
Tu et al. | A data-driven reinforcement learning solution framework for optimal and adaptive personalization of a hip exoskeleton | |
Xue et al. | A new delayless adaptive oscillator for gait assistance | |
CN114129392A (en) | Self-adaptive redundant driving exoskeleton rehabilitation robot capable of regulating and controlling terminal fingertip force | |
CN111728827B (en) | Power lower limb exoskeleton control method, device and system | |
CN110587613B (en) | Real-time feedback and closed-loop control method for negative-pressure pneumatic flexible knee joint exoskeleton | |
Kiguchi et al. | A study of a 4DOF upper-limb power-assist intelligent exoskeleton with visual information for perception-assist | |
Xue et al. | Resonance impedance shaping control of hip robotic exoskeleton | |
CN114147687A (en) | Method and system for real-time man-machine dominant switching cooperative control of lower limb exoskeleton | |
CN114366553A (en) | Exoskeleton rehabilitation training method and system | |
CN111360815B (en) | Human-computer interaction motion control method based on electromyographic signals and joint stress | |
Zhang et al. | Fuzzy adaptive sliding mode control of lower limb exoskeleton rehabilitation robot | |
CN112440267A (en) | Gait phase identification method based on inertial sensor | |
CN111135012A (en) | Training method based on hand rehabilitation training device | |
Enya et al. | Upper limb flexion assistance based on minimum-jerk trajectory using wearable motion-assist robot |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220311 |