Oscillator-based lower extremity hip exoskeleton hybrid gait phase estimation and assist control method
Patent Information
- Application Number
- CN202210780706.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-04
- Publication Date
- 2026-09-18
- Estimated Expiration
- 2042-07-04
AI Technical Summary
[0004]但如今国内外骨骼产品在控制方法上仍然存在许多缺点，例如：人体步态信息感知依赖大量传感器、助力力矩滞后、泛化性能差及人机协调度低等
[0057] By adopting the above technical solution, compared with the prior art, the present invention can autonomously learn the gait characteristics of the exoskeleton wearer and automatically detect changes in the wearer's walking speed. It selects different phase estimators according to the degree of speed change, which overcomes the problem of inaccurate phase estimation caused by frequent switching of walking speed in traditional oscillator-based gait phase estimation assistance methods. Moreover, the present invention can calculate the assist torque that conforms to the human gait characteristics without complex dynamic modeling, resulting in higher human-machine coordination.
Smart Images
Figure CN115581594B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of rehabilitation medical exoskeleton robot assistance, and specifically relates to a control method for a hip joint exoskeleton robot based on an adaptive oscillator. Background Technology
[0002] As people age, their physiological functions decline, their limbs become less flexible, and various diseases, such as stroke, acute cerebrovascular disease, rheumatoid arthritis, and muscle atrophy, can lead to motor dysfunction in the elderly, seriously endangering their health and affecting their normal lives.
[0003] In recent years, the emergence of lower limb exoskeleton robots has overcome the shortcomings of traditional rehabilitation training methods, such as low efficiency and slow results. Lower limb exoskeleton robots can flexibly adjust the rehabilitation training method according to the individual conditions of different patients, making rehabilitation training more scientific and effective. Robotics technology has brought good news to rehabilitation training. Therefore, the development of lower limb exoskeleton robot control technology has important application value in realizing the automation of rehabilitation training, helping patients with lower limb motor dysfunction to regain their walking ability, and improving the level of rehabilitation medicine in my country.
[0004] However, there are still many shortcomings in the control methods of domestic and foreign bone products, such as: human gait information perception relies on a large number of sensors, the assist torque is lagging, the generalization performance is poor, and the human-machine coordination is low. Summary of the Invention
[0005] The purpose of this invention is to provide a hybrid gait phase estimation-assisted control method for a lower limb hip joint exoskeleton based on an oscillator. It can autonomously learn the gait characteristics of the exoskeleton wearer and automatically detect changes in the wearer's walking speed. Based on the degree of speed change, it selects different phase estimators, ensuring the accuracy of phase estimation for each gait during variable-speed walking, the smoothness of the assistance, and human-machine coordination, even with only angle sensors and plantar pressure sensors available.
[0006] The technical solution of this invention: a hybrid gait phase estimation-assisted control method for lower limb hip joint exoskeleton based on oscillators, comprising the following steps:
[0007] 1) First, acquire angle sensor data and plantar pressure sensor data of the hip joint. Then, based on the acquired angle data and plantar pressure data, process the acquired data through a filter to remove high-frequency signals. The filtered data is then used to divide the gait phases through a phase divider to obtain the gait phase division results.
[0008] 2) Based on the results of the gait phase division, the phase estimation selector calculates the results by combining the filtered angle data and the output signal of the adaptive oscillator, and selects whether to use the angle model phase estimator or the adaptive oscillator for phase estimation based on the results.
[0009] 2.1) When using an adaptive oscillator, the adaptive oscillator calculates the normalized phase and the adaptive oscillator output signal based on the filtered angle signal and the above results. The normalized phase is input into the torque generator, and the adaptive oscillator output signal is returned to the phase estimation selector.
[0010] 2.2) When the angle model phase estimator is selected, the angle model phase estimator calculates the phase of the gait cycle change stage based on the input signal and inputs it into the torque generator;
[0011] 3) The torque generator calculates the assist torque based on the received signal; the assist torque controls the exoskeleton motor actuator.
[0012] The filter described is a Butterworth low-pass filter. A second-order Butterworth low-pass filter is designed to filter out high-frequency noise in the data. The gain of the Butterworth low-pass filter is given by formula (1.1).
[0013] (1.1)
[0014] Where ω is the frequency of the input signal, ω c The cutoff frequency is 50Hz, n is the filter order (2nd), and A is the transfer function gain.
[0015] The phase divider is calculated based on the filtered angle signal and the above results, obtaining a normalized phase and an adaptive oscillator output signal. Specifically, it preprocesses the filtered hip joint angle... Perform the normalization process shown in formula (2);
[0016] (2)
[0017] in, , These are the minimum and maximum hip joint angles after filtering preprocessing, respectively. The hip joint angle is the normalized value.
[0018] Introducing sliding window sequences:
[0019] (1.2)
[0020] Where n is an odd number, and ;
[0021] Based on the above sliding window sequence (1.2), the local peaks and valleys of the hip joint angle curve are given by formula (1.3):
[0022] (1.3)
[0023] Where VL represents the trough, PK represents the peak, ZE represents the zero-crossing point, Min represents the function for finding the minimum value, Max represents the function for finding the maximum value, and ε is a positive minimum value.
[0024] To better describe changes in plantar pressure, the collected real-time plantar pressure data were normalized as shown in formula (1.4):
[0025] (1.4)
[0026] in, Represents the ball of the foot and the heel. The pressure value is measured in real time by the plantar pressure sensor. This is the minimum pressure value measured by the plantar pressure sensor. The maximum pressure measured by the plantar pressure sensor. The normalized plantar pressure;
[0027] Based on formulas (1.3) and (1.4), a fuzzy logic reasoning system for gait phase division was constructed. For the hip joint angle input variable, three membership functions were defined: LOW: low, MID: medium, HIGH: high. For the plantar pressure input variable, five membership functions were defined: L: small, M: medium, PH: relatively high, H: high, VH: extremely high. For the output variables of gait phase division, IW: initial swing phase, FW: full swing phase, IS: initial support phase, FS: full support phase.
[0028] The phase estimation selector records the time of occurrence of the initial swing phase of each current gait based on the derivation results of the phase divider. An adaptive oscillator can dynamically learn, track, and synchronize each harmonic component by performing a Fourier transform on any periodic signal, outputting the sum of multiple harmonic components. It describes the periodic signal using four state variables: amplitude, phase, frequency, and offset. Based on the error between the input and output signals, it continuously adjusts and constructs a nonlinear combination of dynamically changing state variables to approximate the input signal. Eventually, the amplitude, frequency, and offset converge to stable values, while the phase changes periodically over time. The formula for the adaptive oscillator is defined as follows:
[0029] (1.6)
[0030] (1.7)
[0031] (1.8)
[0032] (1.9)
[0033] in The fundamental frequency of the tracked signal. and The first The phase and amplitude of each oscillator The offset of the tracked signal, The tracked signal and output signal The error value between , and These are the learning rates for phase, frequency, and amplitude, respectively. Setting an appropriate learning rate can improve the adaptive oscillator's ability to track and synchronize with the input signal. is a natural number representing the number of oscillators that learn the harmonic order of the input signal.
[0034] The adaptive oscillator output signal for
[0035] (1.10)
[0036] Phase estimation in steady state of adaptive oscillator for
[0037] (1.11)
[0038] Based on the phase divider's derivation results, the time when the initial swing phase of the current gait occurs is denoted as... Then the phase of the adaptive oscillator can be estimated. Normalization to Normalized phase The following formula
[0039] (1.12)
[0040] Where mod is the modulo function.
[0041] The angle model phase estimator is used to calculate the phase during the gait cycle change phase. The model formula is defined as follows:
[0042] (1.13)
[0043] in, For hip joint angle data, It is the weighted average of the maximum hip extension angles during the first three gait cycles. The weighted average of the minimum buckling angles. The gait phase at the beginning of the initial support phase, at the end of the complete swing phase;
[0044] , The calculation method is as shown in formulas (1.14) and (1.15).
[0045] (1.14)
[0046] (1.15)
[0047] in, These represent the maximum extension and minimum flexion of the hip joint, respectively. These represent the maximum extension and minimum flexion angles for each of the first three gait cycles. These are the angle-weighted average parameters.
[0048] The calculation method is as follows:
[0049] (1.16)
[0050] (1.17)
[0051] in, For each of the first three gait cycles At the moment of occurrence, These are the phase-weighted average parameters;
[0052] Therefore, the gait phase estimated by the oscillator-based lower limb exoskeleton hybrid gait phase estimation method is... Described by formula (1.18):
[0053] (1.18).
[0054] The torque generator calculates the assist torque and joint torque execution layer based on the received signal: the estimated gait phase, although from 0 to 1 in each gait cycle... Linear growth, but due to individual stride differences, the phase at the end of the full swing phase and the beginning of the initial support phase are different. Not at all Therefore, a segmented assist torque generator was designed, with the following formula:
[0055] (1.19)
[0056] in, To assist in increasing the amplitude of the auxiliary torque, The gait phase is estimated by a hybrid gait phase estimation method for lower limb exoskeleton based on oscillators.
[0057] By adopting the above technical solution, compared with the prior art, the present invention can autonomously learn the gait characteristics of the exoskeleton wearer and automatically detect changes in the wearer's walking speed. It selects different phase estimators according to the degree of speed change, which overcomes the problem of inaccurate phase estimation caused by frequent switching of walking speed in traditional oscillator-based gait phase estimation assistance methods. Moreover, the present invention can calculate the assist torque that conforms to the human gait characteristics without complex dynamic modeling, resulting in higher human-machine coordination. Attached Figure Description
[0058] Figure 1 A schematic diagram of the technical framework of a hybrid gait phase estimation-assisted control technology for a lower limb hip joint exoskeleton based on an oscillator;
[0059] Figure 2 A fuzzy logic deduction system for dividing gait phase stages;
[0060] Figure 3 The results of the hybrid gait phase estimation test for the lower limb hip exoskeleton based on oscillator;
[0061] Figure 4 Test results of the assist torque calculated for the oscillator-based lower limb hip exoskeleton hybrid gait phase estimation assist control technology; Detailed Implementation
[0062] The present invention will be further described below with reference to embodiments, but these embodiments are not intended to limit the scope of the invention.
[0063] Example 1: A hybrid gait phase estimation-assisted control method based on an oscillator-driven lower limb hip joint exoskeleton, comprising the following steps:
[0064] 1) First, acquire angle sensor data and plantar pressure sensor data of the hip joint. Then, based on the acquired angle data and plantar pressure data, process the acquired data through a filter to remove high-frequency signals. The filtered data is then used to divide the gait phases through a phase divider to obtain the gait phase division results.
[0065] 2) Based on the results of the gait phase division, the phase estimation selector calculates the results by combining the filtered angle data and the output signal of the adaptive oscillator, and selects whether to use the angle model phase estimator or the adaptive oscillator for phase estimation based on the results.
[0066] 2.1) When using an adaptive oscillator, the adaptive oscillator calculates the normalized phase and the adaptive oscillator output signal based on the filtered angle signal and the above results. The normalized phase is input into the torque generator, and the adaptive oscillator output signal is returned to the phase estimation selector.
[0067] 2.2) When the angle model phase estimator is selected, the angle model phase estimator calculates the phase of the gait cycle change stage based on the input signal and inputs it into the torque generator;
[0068] 3) The torque generator calculates the assist torque based on the received signal; the assist torque controls the exoskeleton motor actuator.
[0069] The filter described is a Butterworth low-pass filter. A second-order Butterworth low-pass filter is designed to filter out high-frequency noise in the data. The gain of the Butterworth low-pass filter is given by formula (1.1).
[0070] (1.1)
[0071] Where ω is the frequency of the input signal, ω c The cutoff frequency is 50Hz, n is the filter order (2nd), and A is the transfer function gain.
[0072] The phase divider is calculated based on the filtered angle signal and the above results, obtaining a normalized phase and an adaptive oscillator output signal. Specifically, it preprocesses the filtered hip joint angle... Perform the normalization process shown in formula (2);
[0073] (2)
[0074] in, , These are the minimum and maximum hip joint angles after filtering preprocessing, respectively. The hip joint angle is the normalized value.
[0075] Introducing sliding window sequences:
[0076] (1.2)
[0077] Where n is an odd number, and ;
[0078] Based on the above sliding window sequence (1.2), the local peaks and valleys of the hip joint angle curve are given by formula (1.3):
[0079] (1.3)
[0080] Where VL represents the trough, PK represents the peak, ZE represents the zero-crossing point, Min represents the function for finding the minimum value, Max represents the function for finding the maximum value, and ε is a positive minimum value.
[0081] To better describe changes in plantar pressure, the collected real-time plantar pressure data were normalized as shown in formula (1.4):
[0082] (1.4)
[0083] in, Represents the ball of the foot and the heel. The pressure value is measured in real time by the plantar pressure sensor. This is the minimum pressure value measured by the plantar pressure sensor. The maximum pressure measured by the plantar pressure sensor. The normalized plantar pressure;
[0084] Based on formulas (1.3) and (1.4), a fuzzy logic reasoning system for gait phase segmentation was constructed. For the hip joint angle input variable, three membership functions were defined: LOW (low), MID (medium), and HIGH (high). For the plantar pressure input variable, five membership functions were defined: L (small), M (medium), PH (high), H (high), and VH (very high). For the output variables of gait phase segmentation, IW (initial swing phase), FW (full swing phase), IS (initial support phase), and FS (full support phase) were defined. (See Table 1 for details.)
[0085] Table 1
[0086]
[0087] The phase estimation selector records the time of occurrence of the initial swing phase of each current gait based on the derivation results of the phase divider. An adaptive oscillator can dynamically learn, track, and synchronize each harmonic component by performing a Fourier transform on any periodic signal, outputting the sum of multiple harmonic components. It describes the periodic signal using four state variables: amplitude, phase, frequency, and offset. Based on the error between the input and output signals, it continuously adjusts and constructs a nonlinear combination of dynamically changing state variables to approximate the input signal. Eventually, the amplitude, frequency, and offset converge to stable values, while the phase changes periodically over time. The formula for the adaptive oscillator is defined as follows:
[0088] (1.6)
[0089] (1.7)
[0090] (1.8)
[0091] (1.9)
[0092] in The fundamental frequency of the tracked signal. and The first The phase and amplitude of each oscillator The offset of the tracked signal, The tracked signal and output signal The error value between , and These are the learning rates for phase, frequency, and amplitude, respectively. Setting an appropriate learning rate can improve the adaptive oscillator's ability to track and synchronize with the input signal. is a natural number representing the number of oscillators that learn the harmonic order of the input signal.
[0093] The adaptive oscillator output signal for
[0094] (1.10)
[0095] Phase estimation in steady state of adaptive oscillator for
[0096] (1.11)
[0097] Based on the phase divider's derivation results, the time when the initial swing phase of the current gait occurs is denoted as... Then the phase of the adaptive oscillator can be estimated. Normalization to Normalized phase The following formula
[0098] (1.12)
[0099] Where mod is the modulo function.
[0100] The angle model phase estimator is used to calculate the phase during the gait cycle change phase. The model formula is defined as follows:
[0101] (1.13)
[0102] in, For hip joint angle data, It is the weighted average of the maximum hip extension angles during the first three gait cycles. The weighted average of the minimum buckling angles. The gait phase at the beginning of the initial support phase, at the end of the complete swing phase;
[0103] , The calculation method is as shown in formulas (1.14) and (1.15).
[0104] (1.14)
[0105] (1.15)
[0106] in, These represent the maximum extension and minimum flexion of the hip joint, respectively. These represent the maximum extension and minimum flexion angles for each of the first three gait cycles. These are the angle-weighted average parameters.
[0107] The calculation method is as follows:
[0108] (1.16)
[0109] (1.17)
[0110] in, For each of the first three gait cycles At the moment of occurrence, These are the phase-weighted average parameters;
[0111] Therefore, the gait phase estimated by the oscillator-based lower limb exoskeleton hybrid gait phase estimation method is... Described by formula (1.18):
[0112] (1.18).
[0113] The torque generator calculates the assist torque and joint torque execution layer based on the received signal: the estimated gait phase, although from 0 to 1 in each gait cycle... Linear growth, but due to individual stride differences, the phase at the end of the full swing phase and the beginning of the initial support phase are different. Not necessarily Therefore, a segmented assist torque generator was designed, with the following formula:
[0114] (1.19)
[0115] in, To assist in increasing the amplitude of the auxiliary torque, The gait phase is estimated by the oscillator-based lower limb exoskeleton hybrid gait phase estimation method.
Claims
1. A hybrid gait phase estimation method for the lower limb hip joint exoskeleton based on an oscillator, characterized in that, Includes the following steps: 1) First, acquire angle sensor data and plantar pressure sensor data of the hip joint. Then, based on the acquired angle data and plantar pressure data, process the acquired data through a filter to remove high-frequency signals. The filtered data is then used to divide the gait phases through a phase divider to obtain the gait phase division results. 2) The results of gait phase segmentation, the filtered angle data from the phase estimation selector, and the output signal of the adaptive oscillator are used to calculate the results. Based on these results, the angle model phase estimator or the adaptive oscillator is selected for phase estimation. 2.1) When using an adaptive oscillator, the adaptive oscillator calculates the normalized phase and the adaptive oscillator output signal based on the filtered angle signal and the above results. The normalized phase is input into the torque generator, and the adaptive oscillator output signal is returned to the phase estimation selector. 2.2) When the angle model phase estimator is selected, the angle model phase estimator calculates the phase of the gait cycle change stage based on the input signal and inputs it into the torque generator; 3) The torque generator calculates the assist torque based on the received signal; the assist torque is used to control the exoskeleton motor actuator; The phase divider is calculated based on the filtered angle signal and the above results, obtaining a normalized phase and an adaptive oscillator output signal. Specifically, it preprocesses the filtered hip joint angle... Perform the normalization process shown in formula (2); （2） in, , These are the minimum and maximum hip joint angles after filtering preprocessing, respectively. The hip joint angle is the normalized value. Introducing sliding window sequences: （1.2） Where n is an odd number, and ; Based on the above sliding window sequence (1.2), the local peaks and valleys of the hip joint angle curve are given by formula (1.3): （1.3） Where VL represents the trough, PK represents the peak, ZE represents the zero-crossing point, Min represents the function for finding the minimum value, Max represents the function for finding the maximum value, and ε is a positive minimum value. To better describe changes in plantar pressure, the collected real-time plantar pressure data were normalized as shown in formula (1.4): （1.4） in, Represents the ball of the foot and the heel. The pressure value is measured in real time by the plantar pressure sensor. This is the minimum pressure value measured by the plantar pressure sensor. The maximum pressure measured by the plantar pressure sensor. The normalized plantar pressure; Based on formulas (1.3) and (1.4), a fuzzy logic reasoning system for gait phase division was constructed. For the hip joint angle input variable, three membership functions were defined: LOW: low, MID: medium, HIGH: high. For the plantar pressure input variable, five membership functions were defined: L: small, M: medium, PH: relatively high, H: high, VH: extremely high. For the output variables of gait phase division, IW: initial swing phase, FW: full swing phase, IS: initial support phase, FS: full support phase. The angle model phase estimator is used to calculate the phase during the gait cycle change phase. The model formula is defined as follows: （1.13） in, For hip joint angle data, It is the weighted average of the maximum hip extension angles during the first three gait cycles. The weighted average of the minimum buckling angles. The gait phase at the beginning of the initial support phase, at the end of the complete swing phase; , The calculation method is as shown in formulas (1.14) and (1.15). （1.14） （1.15） in, These represent the maximum extension and minimum flexion of the hip joint, respectively. These represent the maximum extension and minimum flexion angles for each of the first three gait cycles. These are the angle-weighted average parameters; The calculation method is as follows: （1.16） （1.17） in, For each of the first three gait cycles At the moment of occurrence, These are the phase-weighted average parameters; Therefore, the gait phase estimated by the oscillator-based lower limb exoskeleton hybrid gait phase estimation method is... Described by formula (1.18): （1.18）。 2. The oscillator-based lower limb hip joint exoskeleton hybrid gait phase estimation method according to claim 1, characterized in that: The filter described is a Butterworth low-pass filter. A second-order Butterworth low-pass filter is designed to filter out high-frequency noise in the data. The gain of the Butterworth low-pass filter is given by formula (1.1). （1.1） Where ω is the frequency of the input signal, ω c The cutoff frequency is 50Hz, n is the filter order (2nd), and A is the transfer function gain.
3. The oscillator-based lower limb hip joint exoskeleton hybrid gait phase estimation method according to claim 1, characterized in that: The phase estimation selector records the time of occurrence of the initial swing phase of each current gait based on the derivation results of the phase divider. Any periodic signal is Fourier transformed to output the sum of multiple harmonic components. The adaptive oscillator can dynamically learn, track, and synchronize each harmonic, and describe the periodic signal with four state variables: amplitude, phase, frequency, and offset. Based on the error between the input and output signals, the nonlinear combination of dynamically changing state variables is continuously adjusted to approximate the input signal. Eventually, the amplitude, frequency, and offset will converge to stable values, while the phase will change periodically over time. The formula for an adaptive oscillator is defined as follows: （1.6） （1.7） （1.8） （1.9） in The fundamental frequency of the tracked signal. and The first The phase and amplitude of each oscillator The offset of the tracked signal, The tracked signal and output signal The error value between , and These are the learning rates for phase, frequency, and amplitude, respectively. Setting an appropriate learning rate can improve the adaptive oscillator's ability to track and synchronize with the input signal. is a natural number representing the number of oscillators that learn the harmonic order of the input signal.
4. The gait phase estimation method for lower limb hip joint exoskeleton based on oscillator according to claim 1, characterized in that: The adaptive oscillator output signal for （1.10） Phase estimation in steady state of adaptive oscillator for （1.11） Based on the phase divider's derivation results, the time when the initial swing phase of the current gait occurs is denoted as... Then the phase of the adaptive oscillator can be estimated. Normalization to Normalized phase The following formula （1.12） Where mod is the modulo function.
5. The gait phase estimation method for lower limb hip joint exoskeleton based on oscillator according to claim 1, characterized in that: The torque generator calculates the assist torque and joint torque execution layer based on the received signal: the estimated gait phase, although from 0 to 1 in each gait cycle... Linear growth, but due to individual stride differences, the phase at the end of the full swing phase and the beginning of the initial support phase are different. Not at all Therefore, a segmented assist torque generator was designed, with the following formula: （1.19） in, To assist in increasing the amplitude of the auxiliary torque, The gait phase is estimated by a hybrid gait phase estimation method for lower limb exoskeleton based on oscillators.