Multi-scene-oriented lower limb exoskeleton gait control method and system

Abstract

The invention discloses a multi-scene-oriented lower limb exoskeleton gait control method and system, and belongs to the technical field of gait recognition control, and the method comprises the steps: carrying out the discrete wavelet transformation of the collected hip joint angles of two legs of a lower limb exoskeleton, determining the gait event types and walking mode types of the two legs, and judging whether to generate a reset signal or not; estimating gait phases of the two legs by adopting double self-adaptive oscillators; when it is detected that the type of the generated gait event is the maximum hip joint stretching event, whether reset signals generated by the corresponding legs are detected at the same time or not is judged, and reset and correction are correspondingly carried out; according to the walking mode category of each leg, dynamically generating a reference auxiliary torque of the corresponding leg to complete gait control of the lower limb exoskeleton; and updating the parameters of the dual-adaptive oscillator, and returning to execute gait phase estimation until the process is ended. The method can be used for multi-scene walking aid, solves the problem of oscillator failure during motion mode switching, does not need training, is small in calculation amount, and has high robustness and accuracy.

Description

Technical Field

[0001] This invention belongs to the field of gait recognition and control technology, specifically relating to a lower limb exoskeleton gait control method and system for multiple scenarios. Background Technology

[0002] Lower limb exoskeleton robots rely on timely and accurate recognition of human movement intentions to provide coordinated assist torques. Existing gait control methods are mainly divided into discrete and continuous categories. Discrete gait control methods, such as those based on finite state machines, depend on specific gait events to divide walking stages, but the assist torques provided to the exoskeleton during the same walking stage are preset and lack adjustability, making it difficult to achieve personalized assistance for the exoskeleton. In contrast, continuous gait control methods, especially those based on adaptive oscillators, can obtain continuous gait phase percentages, which is beneficial for achieving precise continuous control. However, traditional adaptive oscillators often lack pattern recognition capabilities and are only suitable for single, uniform, and non-switching gait scenarios. When encountering complex walking environments or sudden changes in walking patterns, they are prone to slow convergence, decreased accuracy, or even divergence.

[0003] To address this problem, researchers have employed machine learning or neural network methods to identify walking patterns. However, these methods heavily rely on training with large datasets of motion data, which requires significant human and time investment, and the generalization ability of well-trained models is limited.

[0004] Therefore, a gait control method is provided that requires no data training, has low computational cost, and maintains high robustness and accuracy even when switching between multiple scenarios. Summary of the Invention

[0005] The purpose of this invention is to address the above-mentioned problems by proposing a gait control method and system for lower limb exoskeletons in multiple scenarios. The system aims to identify the walking patterns of the lower limb exoskeleton and estimate the gait phase in real time and accurately. In particular, it solves the problem of failure of traditional adaptive oscillators when switching movement modes. It requires no training, has low computational load, can assist walking in multiple scenarios, and has high robustness and accuracy.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] This invention proposes a gait control method for lower limb exoskeletons in multiple scenarios, comprising the following steps:

[0008] S1. Real-time acquisition of the left and right hip joint angles of the lower limb exoskeleton when worn by the human body using a sliding window, forming data to be processed;

[0009] S2. Perform discrete wavelet transform on the data to be processed, that is, decompose the data to be processed using Daubechies mother wavelet and Symlets mother wavelet respectively.

[0010] S3. Determine the current time based on the data to be processed after discrete wavelet transform. The gait event types and walking pattern categories of both legs are identified, and a reset signal is determined.

[0011] S4. Use a dual adaptive oscillator to estimate the current time. Gait phase of both legs;

[0012] S5, upon detecting the current moment When the gait event type is a maximum hip extension event, determine the current moment. Did the corresponding leg generate a reset signal at the same time? If not, for the current moment... The gait phase of the corresponding leg is corrected to obtain the corrected gait phase, and step S6 is executed. If so, the parameters of the corresponding leg in the dual adaptive oscillator are reset to the initial parameters. When the next maximum hip extension event is detected in the left or right leg, the gait phase of the dual adaptive oscillator is synchronously set to the preset initial gait phase at the time of occurrence of the next maximum hip extension event detected in the left or right leg, and the process returns to step S4.

[0013] S6. Based on the current time The walking pattern category of each leg is dynamically generated to correspond to the leg at the current moment. The reference auxiliary torque enables gait control of the lower limb exoskeleton;

[0014] S7, Regarding the current time The parameters of the dual adaptive oscillator are updated, and the updated parameters are used as the initial parameters of the dual adaptive oscillator in the next sampling period. The initial time of the next sampling period is used as the current time, and the process returns to step S4 until the process ends.

[0015] Preferably, before performing discrete wavelet transform on the data to be processed, Butterworth low-pass filtering is also performed on the data to be processed.

[0016] The wavelet mother function of the discrete wavelet transform is given by the following formula:

[0017]

[0018] In the formula, For the current moment wavelet mother function, To correspond to the mother wavelet, For scale parameters, These are translation parameters;

[0019] The decomposition result of the Daubechies mother wavelet is the mid-to-high frequency components after the third-level decomposition.

[0020] The decomposition result of the Symlets mother wavelet is the high-frequency component after first-level decomposition.

[0021] Preferably, the current time is determined based on the data to be processed after discrete wavelet transform. The system identifies the gait event types and walking pattern categories for both legs and determines whether a reset signal should be generated, as detailed below:

[0022] S31, Based on the current time The decomposition result of the Daubechies mother wavelet corresponding to the hip joint angle of the leg determines the gait event type of the corresponding leg, that is, the current moment. The state of the Daubechies mother wavelet decomposition result corresponding to the hip joint angle of the leg when it passes through the zero point: when it changes from positive to negative, the hip joint angle of the corresponding leg reaches a maximum value, and the gait event type of the corresponding leg is determined to be the maximum hip extension event; when it changes from negative to positive, the hip joint angle of the corresponding leg reaches a minimum value, and the gait event type of the corresponding leg is determined to be the maximum hip flexion event.

[0023] S32, Based on the current time The Symlets mother wavelet decomposition results of the corresponding leg hip joint angle, the hip joint angle of the most recent maximum hip extension event of the corresponding leg, and the hip joint angle of the most recent maximum hip flexion event of the corresponding leg determine the walking pattern category of the corresponding leg.

[0024] S33. Determine the current time respectively. If the downward trend of the corresponding leg exceeds the preset trend threshold, it is considered that the corresponding leg has a falling edge and a reset signal is generated for the corresponding leg; otherwise, it is considered that the corresponding leg does not have a falling edge and a reset signal is not generated for the corresponding leg.

[0025] Preferably, based on the current time The Symlets mother wavelet decomposition results of the corresponding leg's hip joint angle, the hip joint angle of the most recent maximum hip extension event of the corresponding leg, and the hip joint angle of the most recent maximum hip flexion event of the corresponding leg determine the walking pattern category of the corresponding leg, as follows:

[0026] S321. Calculate the discriminant features of walking patterns. :

[0027]

[0028] in,

[0029]

[0030] In the formula, express Norm, The Symlets mother wavelet decomposition results correspond to the hip joint angle of the leg. For the corresponding leg and angular difference, The hip angle represents the most recent maximum hip extension event of the corresponding leg. The hip angle is the most recent maximum hip flexion event of the corresponding leg.

[0031] S322. The walking pattern category of the corresponding leg is determined using a multiple threshold classification method, as shown below:

[0032]

[0033] In the formula, This indicates the walking pattern category of the corresponding leg. Indicates going upstairs. Indicates flat land. Indicates going downstairs. Indicates standing. This represents the first preset threshold. This indicates the second preset threshold. This indicates the third preset threshold. This indicates the fourth preset threshold.

[0034] The calculation is based on the downward trend of the corresponding leg, as follows:

[0035]

[0036] In the formula, To correspond to the downward trend of the legs, Initial time Symlets mother wavelet decomposition results corresponding to the hip joint angle. Norm, For the current moment Symlets mother wavelet decomposition results corresponding to the hip joint angle. Norm, h is a preset constant.

[0037] Preferably, a dual adaptive oscillator is used to estimate the current time. The gait phase of both legs is specifically determined by executing the following set of dynamic control equations using dual adaptive oscillators:

[0038]

[0039] In the formula, For the current moment Reconstructed left hip joint angle For the current moment Reconstructed right hip joint angle, For the current moment Left leg offset For the current moment Right leg offset. For the current moment left leg The amplitude of each harmonic. For the current moment Right leg The amplitude of each harmonic; For the current moment left leg The gait phase of a harmonic, For the current moment Right leg The gait phase of a harmonic, , This represents the total number of harmonics.

[0040] Preferably, for the current time The parameters of the dual adaptive oscillator are updated as follows:

[0041] S71. Calculate the current time respectively. Left leg estimation error and the current moment Right leg estimation error :

[0042]

[0043] In the formula, For the current moment Left hip joint angle, For the current moment Right leg hip joint angle;

[0044] S72, Based on the current time Left leg estimation error and the current moment Right leg estimation error The dual adaptive oscillator is corrected in real time using the following formula:

[0045]

[0046] In the formula, For the current moment left leg The update rate of the gait phase of each harmonic. For the current moment Right leg The update rate of the gait phase of each harmonic. For the current moment Update rate of left leg offset For the current moment Update rate of right leg offset For the current moment left leg The update rate of the amplitude of each harmonic. For the current moment Right leg The update rate of the amplitude of each harmonic. For the current moment The update rate of the base frequency, For the current moment The fundamental frequency, The preset gait phase learning rate, The preset offset learning rate or amplitude learning rate, The preset base frequency learning rate;

[0047] S73. The modified formula is discretized using the first-order forward Euler method to obtain the updated parameters. The first-order forward Euler method involves discretizing the current time step by step. The update rate of each parameter of the dual adaptive oscillator and the preset sampling period Multiply and then add up to the current time. The corresponding parameters of the dual adaptive oscillator.

[0048] Preferably, for the current time Corrected gait phase is obtained by correcting the gait phase of the corresponding leg, as follows:

[0049] S511, Calculate the corresponding leg at time [time]. Gait phase error The formula is as follows:

[0050]

[0051] In the formula, This represents the moment of the most recent maximum hip extension event within the current sampling period. Represents the corresponding leg at time The initial gait phase, i.e., the time of the dual adaptive oscillator output. Corresponding leg The gait phase of each harmonic;

[0052] S512, Calculate the current time of each leg. Gait phase correction error term :

[0053]

[0054] in,

[0055]

[0056] In the formula, For the corresponding leg at time The gait phase correction error term, For the corresponding leg at time Gait phase error gain, For the current moment The fundamental frequency, The preset gain constant;

[0057] S513, based on the current time of each leg The gait phase correction error term corrects the gait phase of the corresponding leg, thus obtaining the gait phase of the corresponding leg at the current moment. Corrected gait phase The calculation is as follows:

[0058]

[0059] In the formula, For modulo operation, For the corresponding leg at the current moment The initial gait phase.

[0060] Preferably, based on the current time The walking pattern category of each leg is dynamically generated to correspond to the leg at the current moment. The reference auxiliary torque is used to complete the gait control of the lower limb exoskeleton, as follows:

[0061] S61, at the current moment The corresponding leg walking pattern category is At that time, provide the corresponding leg at the current moment. The reference auxiliary torque is zero; at the current moment The corresponding leg walking pattern category is , and At that time, a weighted summation of Gaussian radial basis functions is performed to obtain the corresponding leg at the current time. Reference auxiliary torque The formula is as follows:

[0062]

[0063] in, For the corresponding leg at the current moment Corrected gait phase, For the first The peak value of a Gaussian radial basis function For the first The mean of the Gaussian radial basis functions For the first The standard deviation of the Gaussian radial basis functions This represents the number of Gaussian radial basis functions. For global coefficients, Indicates going upstairs. Indicates flat land. Indicates going downstairs. Indicates standing;

[0064] S62, based on the current time of each leg The reference auxiliary torque is calculated in reverse using a force-potential model to determine the current moment. The target deformation angle of the drive motor corresponding to the hip joint of the leg;

[0065] S63, Current time The target deformation angle of the drive motor of each leg hip joint and the angle of each leg hip joint are used as the final reference position input to the PID controller to perform closed-loop control of the position of both legs, thus completing the gait control of the lower limb exoskeleton.

[0066] Preferably, the hip joint drive unit of the lower limb exoskeleton includes a housing, a drive motor, a pinion, a gear, a spring, a preload adjustment unit, a rocker mechanism, a first guide rod slider mechanism, and a thigh link. The rocker mechanism includes a first link and a second link. The housing is fixed to the waist when worn by the human body. The thigh link is rotatably connected to the housing. The gear is fixedly connected to the thigh link. The pinion and gear form an internal meshing gear pair and are eccentrically arranged. The thigh link is fixed to the thigh when worn by the human body. The drive motor is fixedly connected to the housing and is used to drive the pinion to rotate. The preload adjustment unit and the gear are mounted together on the thigh link. One end of the first link is coaxially fixedly connected to the pinion, and the other end is connected to one end of the second link. The other end of the second link is connected to the first... The slider of the guide rod slider mechanism is hinged. The guide rail of the first guide rod slider mechanism is fixedly connected to the thigh connecting rod and parallel to the length direction of the thigh. One end of the spring is connected to the slider of the first guide rod slider mechanism, and the other end is connected to the preload adjustment unit. The preload adjustment unit is fixedly connected to the thigh connecting rod and is used to adjust the preload of the spring. In the initial state, the center line of the spring is collinear with the rotation centers of the pinion and the gear, and does not generate torque on the gear. When the thigh connecting rod rotates around the hip joint with the thigh swing, the gear drives the pinion to rotate, thereby causing the rocker mechanism to stretch or retract the spring and generate an elastic torque on the gear. At the same time, the drive motor drives the pinion to rotate, thereby causing the gear to swing. The force-position model is as follows:

[0067]

[0068] in, For the current moment The target deformation angle of the drive motor corresponding to the hip joint of the leg. For the stiffness of the spring, Let be the radius of the pinion. This is the preload force of the spring in its initial state.

[0069] A gait control system for a lower limb exoskeleton oriented to multiple scenarios includes a controller and two inertial measurement units (IMUs). The two IMUs respectively detect the hip joint angle of the left leg and the hip joint angle of the right leg. The controller is used to execute the gait control method for a lower limb exoskeleton oriented to multiple scenarios as described in any one of claims 1-9.

[0070] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0071] This invention is based on discrete wavelet transform and dual adaptive oscillators for walking pattern recognition and gait phase estimation. After error correction, the gait phase is corrected and a reference auxiliary torque is generated. It aims to identify walking patterns and estimate gait phase in real time and accurately. In particular, it solves the problem of traditional adaptive oscillators failing when switching movement modes. It has high computational efficiency and can meet the requirements of real-time, accurate and stable control of lower limb exoskeletons in multi-walking environments. Specifically, in existing technologies, traditional adaptive oscillators rely on periodic learning of joint angles, and parameter adjustments require 4-5 gait cycles to stabilize. When movement mode switching (such as switching from horizontal walking to climbing stairs, or switching between walking and stopping) causes non-periodic abrupt changes in joint angles, the parameter adjustment speed cannot adapt to signal changes, and there is no active reset mechanism or redundant verification. Errors will continue to accumulate or even diverge, eventually losing their control reference value. In contrast, this application uses the time-frequency analysis characteristics of discrete wavelet transform for walking mode recognition and gait event detection, eliminating the need for training on large datasets, significantly reducing deployment costs and time. Furthermore, it can capture signal abrupt changes during walking mode switching in real time (such as the descending edge of the movement trend and the maximum hip extension event), quickly identifying the walking mode and triggering active parameter reset of the dual adaptive oscillators, avoiding the limitations of traditional adaptive oscillators. Due to the convergence lag of the oscillator, this invention effectively solves the problem of traditional adaptive oscillators being easily disturbed and diverging when switching between different movement modes such as flat ground, climbing stairs, and descending stairs on the lower limb exoskeleton. At the same time, the dual adaptive oscillators process the joint angle signals of the left and right legs respectively, and realize the linkage constraint of the parameters of the two legs by utilizing the biological characteristic of the human body walking with the same frequency. The information of the two legs is fused for estimation. Compared with the single-leg adaptive oscillator, the accuracy and stability of gait phase estimation are higher. In particular, it can maintain high robustness in switching between multiple scenarios (such as flat ground, climbing stairs, and descending stairs). Even if the signal of one leg is disturbed, the other leg can still provide a stable reference, eliminating the divergence problem caused by the lack of calibration of one leg. Walking pattern recognition and gait phase estimation can be realized by relying only on the hip joint angle data of the lower limb exoskeleton, reducing the need for the number and types of sensors of the lower limb exoskeleton. Attached Figure Description

[0072] Figure 1 This is a flowchart of the lower limb exoskeleton gait control method for multiple scenarios according to the present invention;

[0073] Figure 2 This is a schematic diagram of the gait phase estimation and walking pattern recognition results of the present invention;

[0074] Figure 3 This is a comparison chart of the average phase error and recognition accuracy of the present invention;

[0075] Figure 4 This is a schematic diagram of the hip joint drive unit of the lower limb exoskeleton of the present invention;

[0076] Figure 5This is a schematic diagram illustrating the working principle of the hip joint drive unit of the lower limb exoskeleton of the present invention.

[0077] Explanation of reference numerals in the attached drawings: 1. Housing; 2. Small gear; 3. Large gear; 4. Spring; 5. Preload adjustment unit; 6. Thigh linkage; 7. Rocker mechanism; 8. First guide rod slider mechanism; 9. Drive motor. Detailed Implementation

[0078] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0079] It should be noted that when a component is referred to as being "connected" to another component, it can be directly connected to the other component or there may be an intervening component. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application.

[0080] The lower limb exoskeleton gait control method proposed in this invention mainly consists of two parts: walking information detection and exoskeleton control. Through the coordinated work of these parts, accurate estimation of gait phase and stable control of the lower limb exoskeleton are achieved in various walking environments. Walking information detection integrates three core operations: walking pattern recognition, gait phase estimation, and auxiliary torque generation control. Specifically, walking pattern recognition is based on Discrete Wavelet Transform (DWT) detection, which calculates the DWT coefficients of the hip joint angle within a sliding window to achieve the set function. In this embodiment, two different mother wavelets (e.g., db1 and sym4) are used to decompose the hip joint angle and extract the corresponding frequency domain features, thereby completing walking pattern classification and gait event detection. Gait phase estimation uses a dual adaptive oscillator to process the hip joint angle and acquire the gait phase in real time, which is then used for subsequent gait phase error correction based on detected gait events. Auxiliary torque generation generates a lower limb exoskeleton reference torque based on the detected gait information, i.e., corrected gait phase and identified movement patterns, and implements control through a PID controller to ensure coordinated movement between the lower limb exoskeleton and the wearer.

[0081] Example 1:

[0082] like Figures 1-5 As shown, a gait control method for lower limb exoskeletons in multiple scenarios includes the following steps:

[0083] S1. Based on the sliding window, the left and right hip joint angles of the lower limb exoskeleton are collected in real time to form data to be processed.

[0084] Specifically, in this embodiment, one IMU sensor (Inertial Measurement Unit) is installed at the center of the left thigh's outer thigh, and another IMU sensor is installed at the center of the right thigh's outer thigh. The two IMU sensors acquire angle signals (hip joint angle) in real time and transmit the data to the controller (Jetson Orin Nano) via a USB interface. The data transmission delay is <10ms, ensuring signal timeliness. The controller uses a sliding window with 150 sampling points, with a sliding step size of one sampling point. Each time new hip joint angle data is received, earlier data within the sliding window is discarded, maintaining a constant amount of data within the window. This setup balances computational efficiency and analytical accuracy, avoiding the real-time performance degradation caused by an excessively long window.

[0085] S2. Perform discrete wavelet transform on the data to be processed, that is, decompose the data to be processed using Daubechies mother wavelet and Symlets mother wavelet respectively.

[0086] In one embodiment, before performing discrete wavelet transform on the data to be processed, Butterworth low-pass filtering is also performed on the data to be processed.

[0087] The wavelet mother function of the discrete wavelet transform is given by the following formula:

[0088]

[0089] In the formula, For the current moment wavelet mother function, To correspond to the mother wavelet, For scale parameters, These are translation parameters;

[0090] The decomposition result of the Daubechies mother wavelet is the mid-to-high frequency components after the third-level decomposition.

[0091] The decomposition result of the Symlets mother wavelet is the high-frequency component after first-level decomposition.

[0092] Specifically, the angle signal within the sliding window is subjected to Butterworth low-pass filtering (cutoff frequency 5Hz) to eliminate high-frequency noise interference. The filtered data, after undergoing Discrete Wavelet Transform (DWT), is directly used as input for subsequent pattern recognition and event detection. The DWT extraction of the filtered data employs two different mother wavelets to perform different detection tasks, including gait event type and walking pattern category recognition. Specifically:

[0093] DWT parameter configuration and coefficient extraction. Two mother wavelets (Daubechies and Symlets mother wavelets) are used for processing. The Daubechies-N1 (db1) mother wavelet is sensitive to extreme points and is therefore used for critical gait event type detection. The Symlet-N4 (sym4) mother wavelet has superior time-frequency localization characteristics, reducing phase distortion, and is therefore used for gait pattern classification. The angle signal within the filtered sliding window is decomposed. The db1 mother wavelet undergoes a three-level decomposition to extract the third-level detail coefficients (cD3), i.e., mid-to-high frequency components (LH and HL components). The sym4 mother wavelet undergoes a one-level decomposition to extract the first-level detail coefficients (cD1), i.e., high-frequency components (HH components). The decomposition can be implemented using the PyWavelets library to avoid boundary effects; this is a well-known technique and will not be elaborated upon here.

[0094] S3. Determine the current time based on the data to be processed after discrete wavelet transform. The system identifies the gait event types and walking pattern categories for both legs and determines whether a reset signal should be generated.

[0095] In one embodiment, the current time is determined based on the data to be processed after discrete wavelet transform. The system identifies the gait event types and walking pattern categories for both legs and determines whether a reset signal should be generated, as detailed below:

[0096] S31, Based on the current time The decomposition result of the Daubechies mother wavelet corresponding to the hip joint angle of the leg determines the gait event type of the corresponding leg, that is, the current moment. The state of the Daubechies mother wavelet decomposition result corresponding to the hip joint angle of the leg when it passes through the zero point: when it changes from positive to negative, the hip joint angle of the corresponding leg reaches a maximum value, and the gait event type of the corresponding leg is determined to be the maximum hip extension event; when it changes from negative to positive, the hip joint angle of the corresponding leg reaches a minimum value, and the gait event type of the corresponding leg is determined to be the maximum hip flexion event.

[0097] S32, Based on the current time The Symlets mother wavelet decomposition results of the corresponding leg hip joint angle, the hip joint angle of the most recent maximum hip extension event of the corresponding leg, and the hip joint angle of the most recent maximum hip flexion event of the corresponding leg determine the walking pattern category of the corresponding leg.

[0098] S33. Determine the current time respectively. If the downward trend of the corresponding leg exceeds the preset trend threshold, it is considered that the corresponding leg has a falling edge and a reset signal is generated for the corresponding leg; otherwise, it is considered that the corresponding leg does not have a falling edge and a reset signal is not generated for the corresponding leg.

[0099] In one embodiment, based on the current time The Symlets mother wavelet decomposition results of the corresponding leg's hip joint angle, the hip joint angle of the most recent maximum hip extension event of the corresponding leg, and the hip joint angle of the most recent maximum hip flexion event of the corresponding leg determine the walking pattern category of the corresponding leg, as follows:

[0100] S321. Calculate the discriminant features of walking patterns. :

[0101]

[0102] in,

[0103]

[0104] In the formula, express Norm, The Symlets mother wavelet decomposition results correspond to the hip joint angle of the leg. For the corresponding leg and angular difference, The hip angle represents the most recent maximum hip extension event of the corresponding leg. The hip angle is the most recent maximum hip flexion event of the corresponding leg.

[0105] S322. The walking pattern category of the corresponding leg is determined using a multiple threshold classification method, as shown below:

[0106]

[0107] In the formula, This indicates the walking pattern category of the corresponding leg. Indicates going upstairs. Indicates flat land. Indicates going downstairs. Indicates standing. This represents the first preset threshold. This indicates the second preset threshold. This indicates the third preset threshold. This indicates the fourth preset threshold.

[0108] The calculation is based on the downward trend of the corresponding leg, as follows:

[0109]

[0110] In the formula, To correspond to the downward trend of the legs, Initial time Symlets mother wavelet decomposition results corresponding to the hip joint angle. Norm, For the current moment Symlets mother wavelet decomposition results corresponding to the hip joint angle. Norm, h is a preset constant.

[0111] Discrete wavelet transform (DWT) is used for parallel detection of desired gait events (MFA) for the left and right legs. These two event signals are used to trigger gait phase correction, independently correcting the initial gait phase output from the dual adaptive oscillators, ultimately outputting high-precision dual-leg gait phases. Based on the data to be processed after DWT, the current time is determined. The system identifies the gait event types and walking pattern categories for both legs and determines whether a reset signal should be generated, as detailed below:

[0112] 3.1) Detection of critical gait events (MFA and MEA). The detection logic is that when the cD3 coefficient of db1 decomposition passes through the extreme point of the hip joint angle corresponding to the zero point, the angle reaches a maximum value when cD3 changes from positive to negative, which is determined as the maximum hip extension event (MFA); when cD3 changes from negative to positive, the angle reaches a minimum value, which is determined as the maximum hip flexion event (MEA).

[0113] 3.2) Walking pattern classification. Walking pattern recognition preferably uses the Symlet-N4 (sym4) mother wavelet to extract the first-level detail coefficients (cD1) and calculate the cD1 coefficients. of Manhattan norm (MAN) and calculate the hip angle corresponding to the most recently detected MFA event. Angle corresponding to the MEA event The difference between Multiplying the two together yields the discriminative features of the walking pattern. By comparing the discriminative features of walking patterns With preset threshold ( The relationship between walking patterns can be used to classify them, such as according to... , , , Four preset thresholds will correspond to the leg's walking pattern category. Divided into going upstairs ( ), flat land ), downstairs ( Standing () Four categories.

[0114] 3.3) Motion mode switching recognition. For recognizing a downward motion trend, the cD1 coefficient is monitored. of Norm from initial value up to the current value downward trend ,when Exceeding the preset trend threshold If the current leg has a descending edge, it is considered to be in a downward trend and a reset signal is generated. Otherwise, if the current leg does not have a descending edge, it is considered not to have a downward trend and a reset signal is not generated for the current leg.

[0115] S4. Use a dual adaptive oscillator to estimate the current time. Gait phase of both legs.

[0116] In one embodiment, a dual adaptive oscillator is used to estimate the current time. The gait phase of both legs is specifically determined by executing the following set of dynamic control equations using dual adaptive oscillators:

[0117]

[0118] In the formula, For the current moment Reconstructed left hip joint angle For the current moment Reconstructed right hip joint angle, For the current moment Left leg offset For the current moment Right leg offset. For the current moment left leg The amplitude of each harmonic. For the current moment Right leg The amplitude of each harmonic; For the current moment left leg The gait phase of a harmonic, For the current moment Right leg The gait phase of a harmonic, , This represents the total number of harmonics.

[0119] Specifically, a dual adaptive oscillator (DAFO) is used to estimate gait phase, leveraging the biomechanical characteristic of legs having the same frequency but different phases to improve the robustness of gait phase estimation. The current time... Reconstructed left hip joint angle and the current moment Reconstructed right hip joint angle Simultaneously, a dual adaptive oscillator is input to execute a set of dynamic control equations. For example, in this embodiment, the reconstructed hip joint angle of the left or right leg is the hip joint angle after Butterworth low-pass filtering.

[0120] S5, upon detecting the current moment When the gait event type is a maximum hip extension event, determine the current moment. Did the corresponding leg generate a reset signal at the same time? If not, for the current moment... The gait phase of the corresponding leg is corrected to obtain the corrected gait phase, and step S6 is executed. If so, the parameters of the corresponding leg in the dual adaptive oscillator are reset to the initial parameters. When the next maximum hip extension event is detected in the left or right leg, the gait phase of the dual adaptive oscillator is synchronously set to the preset initial gait phase at the time of occurrence of the next maximum hip extension event detected in the left or right leg as the initial time of the next sampling period, and the process returns to step S4.

[0121] In one embodiment, for the current time Corrected gait phase is obtained by correcting the gait phase of the corresponding leg, as follows:

[0122] S511, Calculate the corresponding leg at time [time]. Gait phase error The formula is as follows:

[0123]

[0124] In the formula, This represents the moment of the most recent maximum hip extension event within the current sampling period. Represents the corresponding leg at time The initial gait phase, i.e., the time of the dual adaptive oscillator output. Corresponding leg The gait phase of each harmonic;

[0125] S512, Calculate the current time of each leg. Gait phase correction error term :

[0126]

[0127] in,

[0128]

[0129] In the formula, For the corresponding leg at time The gait phase correction error term, For the corresponding leg at time Gait phase error gain, For the current moment The fundamental frequency, The preset gain constant;

[0130] S513, based on the current time of each leg The gait phase correction error term corrects the gait phase of the corresponding leg, thus obtaining the gait phase of the corresponding leg at the current moment. Corrected gait phase The calculation is as follows:

[0131]

[0132] In the formula, For modulo operation, For the corresponding leg at the current moment The initial gait phase.

[0133] Specifically, gait phase correction and dual adaptive oscillator reset address the divergence problem of traditional adaptive oscillators during walking mode switching.

[0134] 5.1) Gait Phase Correction. When a maximum hip extension event of the left or right leg is detected, and no reset signal is detected for the corresponding leg, gait phase correction for that leg is initiated. The corresponding leg's gait phase at time [time value missing] is calculated. initial gait phase Phase with standard gait ( The error of the gait phase at time = 0 is used as the corresponding leg at time... Gait phase error To calculate the current time of each leg Gait phase correction error term This is used to smoothly correct the initial gait phase.

[0135] 5.2) Dual Adaptive Oscillator Reset. When a falling edge of the movement trend is detected by the DWT, it serves as the trigger signal for resetting the dual adaptive oscillator. Immediately after triggering, the internal parameters of the dual adaptive oscillator, namely frequency, amplitude, and phase, are reset to their initial values, and parameter learning is paused. When the DWT detects the next MFA event (whether it is the left or right leg), the occurrence time of this MFA event is used as the initial time of the next sampling period. The gait phase of both legs is set to the preset initial gait phase, and the dynamic control equations of the dual adaptive oscillator will resume operation based on this, achieving rapid gait phase convergence. The dual adaptive oscillator reset is triggered when a walking mode switch is detected to eliminate interference from aperiodic signals; and restarts the dual adaptive oscillator when the next critical gait event occurs, achieving rapid gait phase convergence.

[0136] S6. Based on the current time The walking pattern category of each leg is dynamically generated to correspond to the leg at the current moment. The reference auxiliary torque enables gait control of the lower limb exoskeleton.

[0137] In one embodiment, based on the current time The walking pattern category of each leg is dynamically generated to correspond to the leg at the current moment. The reference auxiliary torque is used to complete the gait control of the lower limb exoskeleton, as follows:

[0138] S61, at the current moment The corresponding leg walking pattern category is At that time, provide the corresponding leg at the current moment. The reference auxiliary torque is zero; at the current moment The corresponding leg walking pattern category is , and At that time, a weighted summation of Gaussian radial basis functions is performed to obtain the corresponding leg at the current time. Reference auxiliary torque The formula is as follows:

[0139]

[0140] in, For the corresponding leg at the current moment Corrected gait phase, For the first The peak value of a Gaussian radial basis function For the first The mean of the Gaussian radial basis functions For the first The standard deviation of the Gaussian radial basis functions This represents the number of Gaussian radial basis functions. For global coefficients, Indicates going upstairs. Indicates flat land. Indicates going downstairs. Indicates standing;

[0141] S62, based on the current time of each leg The reference auxiliary torque is calculated in reverse using a force-potential model to determine the current moment. The target deformation angle of the drive motor corresponding to the hip joint of the leg;

[0142] S63, Current time The target deformation angle of the drive motor of each leg hip joint and the angle of each leg hip joint are used as the final reference position input to the PID controller to perform closed-loop control of the position of both legs, thus completing the gait control of the lower limb exoskeleton.

[0143] In one embodiment, the hip joint drive unit of the lower limb exoskeleton includes a housing, a drive motor, a pinion, a gear, a spring, a preload adjustment unit, a rocker mechanism, a first guide rod slider mechanism, and a thigh link. The rocker mechanism includes a first link and a second link. The housing is fixed to the waist when worn by the human body. The thigh link is rotatably connected to the housing. The gear is fixedly connected to the thigh link. The pinion and gear form an internal meshing gear pair and are eccentrically arranged. The thigh link is fixed to the thigh when worn by the human body. The drive motor is fixedly connected to the housing and is used to drive the pinion to rotate. The preload adjustment unit and the gear are mounted together on the thigh link. One end of the first link is coaxially fixedly connected to the pinion, and the other end is connected to one end of the second link. The other end of the second link is connected to the first guide rod slider mechanism. The slider of the first guide rod slider mechanism is hinged. The guide rail of the first guide rod slider mechanism is fixedly connected to the thigh connecting rod and parallel to the length direction of the thigh. One end of the spring is connected to the slider of the first guide rod slider mechanism, and the other end is connected to the preload adjustment unit. The preload adjustment unit is fixedly connected to the thigh connecting rod and is used to adjust the preload of the spring. In the initial state, the center line of the spring is collinear with the rotation centers of the pinion and the gear, and does not generate torque on the gear. When the thigh connecting rod rotates around the hip joint with the swing of the thigh, the gear drives the pinion to rotate, thereby causing the rocker mechanism to stretch or retract the spring and generate an elastic torque on the gear. At the same time, the drive motor drives the pinion to rotate, thereby causing the gear to swing the thigh connecting rod. The force-position model is as follows:

[0144]

[0145] in, For the current moment The target deformation angle of the drive motor corresponding to the hip joint of the leg. For the stiffness of the spring, Let be the radius of the pinion. This is the preload force of the spring in its initial state.

[0146] Specifically, a personalized reference torque is generated based on the walking pattern and the phase of the corrected gait to achieve coordinated assistance. Details are as follows:

[0147] 6.1) Reference Auxiliary Torque Generation. When the walking mode category identified by DWT is LG, SA, or SD, the corresponding reference auxiliary torque generation function is selected. This reference auxiliary torque generation function is preferably a linear combination of multiple Gaussian functions (weighted summation of Gaussian radial basis functions). The preset fitting parameters for different walking mode categories can be adjusted according to actual needs. For example... It can be personalized according to the wearer's weight, specifically set to 10% to 15% of the wearer's weight to ensure a proper fit. When the walking pattern category recognized by DWT is... When the reference assist torque is not provided, the output will be ignored, and the corresponding hip joint will be in standby mode to ensure smooth and safe switching of walking modes. When a walking restart (reset signal) is detected, that is, when the next MFA event occurs (i.e., the next maximum hip extension event occurs in the left or right leg), the standby state will be immediately released and the corresponding reference assist torque will be quickly restored.

[0148] 5.2) Implementation of Reference Auxiliary Torque. The calculated reference auxiliary torque for both legs is used to inversely calculate the target deformation angle of the drive motor through a force-position model. The target deformation angle of the corresponding leg is combined with the current hip joint angle detected by the IMU to form the final reference position. Finally, a PID controller is used to implement independent position closed-loop control of both legs. Each PID controller dynamically adjusts its output to generate deformation based on the position deviation between this final reference position and the position information fed back by the corresponding drive motor in real time, thereby indirectly achieving accurate tracking of the target deformation angle. It is easy to understand that the PID controller collects the actual rotation angle of the drive motor in real time through the encoder integrated into the drive motor, compares the final reference position with the actual rotation angle of the drive motor to calculate the angle deviation, and then converts the angle deviation into a PWM drive signal output to the drive motor through proportional (P), integral (I), and derivative (D) adjustment rules. The proportional gain ranges from 5.0 to 15.0, the integral gain ranges from 0.1 to 1.0, and the derivative gain ranges from 0.01 to 0.1. In the specific experiment of this embodiment, the proportional gain is set to 8.0, the integral gain to 0.3, and the derivative gain to 0.05. The drive motor rotates to the target deformation angle according to the drive signal, and drives the spring to generate the corresponding deformation through gear transmission. According to Hooke's law, the deformation of the spring is converted into elastic force, which is finally transmitted as the actual auxiliary torque at the hip joint, realizing accurate tracking of the reference auxiliary torque and completing the gait control of the lower limb exoskeleton.

[0149] Specifically, such as Figure 4As shown, the hip joint drive unit of the lower limb exoskeleton includes a housing 1, a drive motor 9, a pinion 2, a large gear 3, a spring 4, a preload adjustment unit 5, a rocker mechanism 7, a first guide rod slider mechanism 8, and a thigh link 6. The housing 1 is fixed to the waist when worn by the human body and is used to support and install the hip joint drive unit. The drive motor 9 is fixedly installed on the housing 1, and its output shaft is connected to the pinion 2 for driving the pinion 2 to rotate. The pinion 2 and the large gear 3 form an internal meshing gear pair, and the large gear 3 is installed on the housing 1 through a rotating shaft and bearing, so that it can rotate relative to the housing 1. The thigh link 6 is the output component of the hip joint. The pinion 2 meshes with the large gear 3 in an eccentric manner. The preload adjustment unit 5 is connected to the thigh link 6 and can be a spring preload adjustment mechanism in the prior art. For example, the preload adjustment unit includes a threaded fixed seat and a bolt. The fixed seat is connected and fixed to the thigh link 6. The bolt is threadedly connected to the other end of the spring 4. Rotating the bolt can adjust the preload of the spring 4. Spring 4 is positioned along the length of the bolt in the preload adjustment unit. One end is fixedly connected to the slider of the first guide rod slider mechanism 8, and the other end is fixedly connected to the preload adjustment unit 5. This allows the large gear 3 to stretch or retract during rotation, generating an elastic torque acting on the large gear 3. The other end of spring 4 can also be connected to the preload adjustment unit 5 via the second guide rod slider mechanism for guiding the spring's extension and retraction. The sliding directions of the second guide rod slider mechanism and the first guide rod slider mechanism 8 are parallel to each other. In the initial state, the centerline of spring 4 is collinear with the rotation centers of the small gear 2 and the large gear 3, so that spring 4 does not generate an elastic torque on the large gear 3 in the initial state. When the large gear 3 rotates, the line of action of spring 4 deviates from the rotation centers of the small gear 2 and the large gear 3, thus forming a nonlinear elastic torque that varies with the rotation angle. The nonlinear elastic torque and the rotation angle of the large gear 3 satisfy the force-displacement model. In this embodiment, the radius ratio of the large gear 3 to the small gear 2 is 2:1. When the small gear 2 rotates, it drives the spring 4 to stretch or compress, achieving accurate output of the reference auxiliary torque, and using a force-position model to calculate the target deformation angle of the drive motor 9 in reverse. Figure 5 As shown, Let the rotation center of pinion 2 be and the radius be . , The rotation center of the large gear 3 has a radius of 2r. The swing angle of the hip joint (i.e., the target deformation angle of the drive motor corresponding to the hip joint of the leg) is defined by point A, which is the connection point between spring 4 and rocker mechanism 7, and point B, which is the connection point between spring 4 and preload adjustment unit 5. When the thigh swings, pinion 2 drives point A to rotate... point.

[0150] The gait event types and walking pattern categories detected by each leg are corrected for errors. The corrected effective gait events and effective walking patterns are then passed to the next level for reference auxiliary torque generation to improve detection accuracy.

[0151] S7, Regarding the current time The parameters of the dual adaptive oscillator are updated, and the updated parameters are used as the initial parameters of the dual adaptive oscillator in the next sampling period. The initial time of the next sampling period is used as the current time, and the process returns to step S4 until the process ends.

[0152] In one embodiment, for the current time The parameters of the dual adaptive oscillator are updated as follows:

[0153] S71. Calculate the current time respectively. Left leg estimation error and the current moment Right leg estimation error :

[0154]

[0155] In the formula, For the current moment Left hip joint angle, For the current moment Right leg hip joint angle;

[0156] S72, Based on the current time Left leg estimation error and the current moment Right leg estimation error The dual adaptive oscillator is corrected in real time using the following formula:

[0157]

[0158] In the formula, For the current moment left leg The update rate of the gait phase of each harmonic. For the current moment Right leg The update rate of the gait phase of each harmonic. For the current moment Update rate of left leg offset For the current moment Update rate of right leg offset For the current moment left leg The update rate of the amplitude of each harmonic. For the current moment Right leg The update rate of the amplitude of each harmonic. For the current moment The update rate of the base frequency, For the current moment The fundamental frequency, The preset gait phase learning rate, The preset offset learning rate or amplitude learning rate, The preset base frequency learning rate;

[0159] S73. The modified formula is discretized using the first-order forward Euler method to obtain the updated parameters. The first-order forward Euler method involves discretizing the current time step by step. The update rate of each parameter of the dual adaptive oscillator and the preset sampling period Multiply and then add up to the current time. The corresponding parameters of the dual adaptive oscillator.

[0160] The objective of the dual adaptive oscillator is to minimize the estimation error between the left and right legs. and To achieve this goal, based on dynamic control rules, utilizing... and By learning rate , and Simultaneously, it independently updates the gait phase, amplitude, and offset of each leg, and collectively updates a shared fundamental frequency. This shared fundamental frequency... The linkage constraint mechanism utilizes the synchronous frequency characteristic of human bipedal walking. This linkage constraint mechanism ensures that even if a single leg signal (such as...) is transmitted, the signal will still be transmitted even if the signal is transmitted through the same leg. Interference causes increased error, affecting the stable signal of the other leg (such as...). It can also provide a reference, through The average fusion effect constrains the shared base frequency. To maintain stability and eliminate divergence issues caused by uncalibrated single-leg walking, the system employs several mechanisms: a base frequency update to capture the user's walking rhythm for human-machine synchronization, amplitude and offset updates to fit the motion characteristics of joint angles, and gait phase updates to ensure the real-time accuracy of gait phase prediction.

[0161] The correction formula is discretized using the first-order forward Euler method, that is, using the current time step. The calculated update rate of each parameter and the preset sampling period The product of the ... The parameters (including gait phase, amplitude, offset, fundamental frequency, etc.) are accumulated and updated to obtain the updated parameters. The updated parameters are used as the initial parameters for the next sampling period, and the initial time of the next sampling period is used as the current time. The process returns to step S4 to achieve adaptive looping. It is easy to understand that the first-order forward Euler method is an existing technique for solving discrete values ​​of differential equations by those skilled in the art.

[0162] Reference Figure 2 and Figure 3 The effectiveness of the proposed method was verified through experiments on human wearing a lower limb exoskeleton.

[0163] Reference Figure 2 The results show the experimental results using DAFO and DWT detection. Figure 2 (a) is an upstairs (SA) and horizontal ground (LG) environment. Figure 2 (b) is the environment of downstairs (SD) and horizontal ground (LG). Figure 2 (a) Figure 2 (b) The top curve shows the angles, including the left hip angle (left leg data) and the right hip angle (right leg data). The middle curve represents the gait phase estimated by DAFO; the solid line represents the gait phase of the left leg, and the dashed line represents the gait phase of the right leg. It can be seen that at the walking mode (or mode) switching points (e.g., 38s, 48s, 104s, 114s), the gait phase estimation curve maintains a stable and linear increase without divergence. The bottom curve represents the walking mode result identified by DWT. It can be seen that the switching of walking modes can be accurately identified, with a delay of approximately half a gait cycle. A gait cycle typically begins with a heel strike, i.e., from the heel strike of one foot until the heel strikes the same foot again. In this embodiment, the sampling frequency is set to 300Hz, corresponding to a sampling period t = 1 / 300. The gait cycle refers to the complete process from the first heel strike on one side to the next heel strike on the same side, i.e., the process of completing one full walking motion. The real-time updated base frequency is the reciprocal of the gait cycle. Within one gait cycle, approximately 300 to 450 sampling calculations are performed (e.g., 300). The parameters of the dual adaptive oscillator are iterated based on the bi-leg estimation error of each sampling cycle, meaning that the parameters undergo 300 fine-tunings after each gait cycle, thus achieving rapid convergence within seconds.

[0164] Reference Figure 3 The data shows a performance comparison between the present invention and existing technologies. Figure 3(a) The average phase error (average gait phase error) of the dual adaptive oscillator (DAFO, gray bar) of the present invention was compared with that of the conventional single-leg adaptive oscillator (AOs, black bar). In the three modes of LG, SA, and SD, the errors of the dual adaptive oscillator (0.08, 0.12, and 0.12, respectively) were significantly lower than those of the conventional single-leg adaptive oscillator. Figure 3 (b) illustrates the accuracy achieved by the present invention using DWT (DWT, gray bars), where the error bars (I) represent the error range of the recognition accuracy, such as standard deviation or standard error. In LG, SA, and SD modes, the recognition accuracies are 95.28% ± 1.85%, 92.44% ± 2.27%, and 89.75% ± 4.59%, respectively. Compared to the Random Forest method (RFc, black bars) which requires training, the present invention achieves similar or even higher recognition accuracy without data training.

[0165] Example 2:

[0166] A gait control system for a lower limb exoskeleton oriented to multiple scenarios includes a controller and two inertial measurement units (IMUs). The two IMUs detect the hip joint angle of the left leg and the hip joint angle of the right leg, respectively. The controller is used to execute the gait control method for a lower limb exoskeleton oriented to multiple scenarios as described in Example 1.

[0167] The two inertial measurement units are used to detect the hip joint angles of both legs. The controller is a Jeston Orin Nano. The drive unit of the lower limb exoskeleton uses a GIM8115 disc motor. This drive unit can also integrate a driver, an angle encoder, and a torque detection unit. The driver is used to calculate the angle deviation by comparing the final reference position with the actual rotation angle of the drive motor to realize the movement of the drive motor. The angle encoder is used to detect the actual rotation angle of the drive motor. The torque detection unit is used to detect the actual auxiliary torque at the hip joint.

[0168] This system is based on discrete wavelet transform and dual adaptive oscillators for walking pattern recognition and gait phase estimation. After error correction, the gait phase is corrected and a reference auxiliary torque is generated. It aims to identify walking patterns and estimate gait phase in real time and accurately. In particular, it solves the problem of traditional adaptive oscillators failing when switching movement modes. It has high computational efficiency and can meet the requirements of real-time, accurate and stable control of the lower limb exoskeleton in multi-walking environments. Specifically, in existing technologies, traditional adaptive oscillators rely on periodic learning of joint angles, and parameter adjustments require 4-5 gait cycles to stabilize. When movement mode switching (such as switching from horizontal walking to climbing stairs, or switching between walking and stopping) causes non-periodic abrupt changes in joint angles, the parameter adjustment speed cannot adapt to signal changes, and there is no active reset mechanism or redundant verification. Errors will continue to accumulate or even diverge, eventually losing their control reference value. In contrast, this application uses the time-frequency analysis characteristics of discrete wavelet transform for walking mode recognition and gait event detection, eliminating the need for training on large datasets, significantly reducing deployment costs and time. Furthermore, it can capture signal abrupt changes during walking mode switching in real time (such as the descending edge of the movement trend and the maximum hip extension event), quickly identifying the walking mode and triggering active parameter reset of the dual adaptive oscillators, avoiding the limitations of traditional adaptive oscillators. Due to the convergence lag of the oscillator, this invention effectively solves the problem of traditional adaptive oscillators being easily disturbed and diverging when switching between different movement modes such as flat ground, climbing stairs, and descending stairs on the lower limb exoskeleton. At the same time, the dual adaptive oscillators process the joint angle signals of the left and right legs respectively, and realize the linkage constraint of the parameters of the two legs by utilizing the biological characteristic of the human body walking with the same frequency. The information of the two legs is fused for estimation. Compared with the single-leg adaptive oscillator, the accuracy and stability of gait phase estimation are higher. In particular, it can maintain high robustness in switching between multiple scenarios (such as flat ground, climbing stairs, and descending stairs). Even if the signal of one leg is disturbed, the other leg can still provide a stable reference, eliminating the divergence problem caused by the lack of calibration of one leg. Walking pattern recognition and gait phase estimation can be realized by relying only on the hip joint angle data of the lower limb exoskeleton, reducing the need for the number and types of sensors of the lower limb exoskeleton.

[0169] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0170] The embodiments described above are merely specific and detailed examples of the embodiments described in this application, and should not be construed as limiting the scope of the application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A gait control method for lower limb exoskeletons in multiple scenarios, characterized in that: Includes the following steps: S1. Real-time acquisition of the left and right hip joint angles of the lower limb exoskeleton when worn by the human body using a sliding window, forming data to be processed; S2. Perform discrete wavelet transform on the data to be processed, that is, decompose the data to be processed using Daubechies mother wavelet and Symlets mother wavelet respectively. S3. Determine the current time based on the data to be processed after discrete wavelet transform. The gait event types and walking pattern categories of both legs are identified, and a reset signal is determined. S4. Use a dual adaptive oscillator to estimate the current time. Gait phase of both legs; S5, upon detecting the current moment When the gait event type is a maximum hip extension event, determine the current moment. Did the corresponding leg generate a reset signal at the same time? If not, for the current moment... The gait phase of the corresponding leg is corrected to obtain the corrected gait phase, and step S6 is executed. If so, the parameters of the corresponding leg in the dual adaptive oscillator are reset to the initial parameters. When the next maximum hip extension event is detected in the left or right leg, the gait phase of the dual adaptive oscillator is synchronously set to the preset initial gait phase at the time of occurrence of the next maximum hip extension event detected in the left or right leg, and the process returns to step S4. S6. Based on the current time The walking pattern category of each leg is dynamically generated to correspond to the leg at the current moment. The reference auxiliary torque enables gait control of the lower limb exoskeleton; S7, Regarding the current time The parameters of the dual adaptive oscillator are updated, and the updated parameters are used as the initial parameters of the dual adaptive oscillator in the next sampling period. The initial time of the next sampling period is used as the current time, and the process returns to step S4 until the process ends.

2. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 1, characterized in that: Before performing discrete wavelet transform on the data to be processed, Butterworth low-pass filtering is also performed on the data to be processed. The wavelet mother function of the discrete wavelet transform is given by the following formula: In the formula, For the current moment wavelet mother function, To correspond to the mother wavelet, For scale parameters, These are translation parameters; The decomposition result of the Daubechies mother wavelet is the mid-to-high frequency components after three-level decomposition; The decomposition result of the Symlets mother wavelet is the high-frequency component after first-level decomposition.

3. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 1, characterized in that: The current time is determined based on the data to be processed after discrete wavelet transform. The system identifies the gait event types and walking pattern categories for both legs and determines whether a reset signal should be generated, as detailed below: S31, Based on the current time The decomposition result of the Daubechies mother wavelet corresponding to the hip joint angle of the leg determines the gait event type of the corresponding leg, that is, the current moment. The state of the Daubechies mother wavelet decomposition result corresponding to the hip joint angle of the leg when it passes through the zero point: when it changes from positive to negative, the hip joint angle of the corresponding leg reaches a maximum value, and the gait event type of the corresponding leg is determined to be the maximum hip extension event; when it changes from negative to positive, the hip joint angle of the corresponding leg reaches a minimum value, and the gait event type of the corresponding leg is determined to be the maximum hip flexion event. S32, Based on the current time The Symlets mother wavelet decomposition results of the corresponding leg hip joint angle, the hip joint angle of the most recent maximum hip extension event of the corresponding leg, and the hip joint angle of the most recent maximum hip flexion event of the corresponding leg determine the walking pattern category of the corresponding leg. S33. Determine the current time respectively. If the downward trend of the corresponding leg exceeds the preset trend threshold, it is considered that the corresponding leg has a falling edge and a reset signal is generated for the corresponding leg; otherwise, it is considered that the corresponding leg does not have a falling edge and a reset signal is not generated for the corresponding leg.

4. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 3, characterized in that: According to the current time The Symlets mother wavelet decomposition results of the corresponding leg's hip joint angle, the hip joint angle of the most recent maximum hip extension event of the corresponding leg, and the hip joint angle of the most recent maximum hip flexion event of the corresponding leg determine the walking pattern category of the corresponding leg, as follows: S321. Calculate the discriminant features of walking patterns. : in, In the formula, express Norm, The Symlets mother wavelet decomposition results correspond to the hip joint angle of the leg. For the corresponding leg and angular difference, The hip angle represents the most recent maximum hip extension event of the corresponding leg. The hip angle is the most recent maximum hip flexion event of the corresponding leg. S322. The walking pattern category of the corresponding leg is determined using a multiple threshold classification method, as shown below: In the formula, This indicates the walking pattern category of the corresponding leg. Indicates going upstairs. Indicates flat land. Indicates going downstairs. Indicates standing. This represents the first preset threshold. This indicates the second preset threshold. This indicates the third preset threshold. This indicates the fourth preset threshold. The downward trend of the corresponding leg is calculated as follows: In the formula, To correspond to the downward trend of the legs, Initial time Symlets mother wavelet decomposition results corresponding to the hip joint angle. Norm, For the current moment Symlets mother wavelet decomposition results corresponding to the hip joint angle. Norm, h is a preset constant.

5. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 1, characterized in that: The current time is estimated using a dual adaptive oscillator. The gait phase of both legs is specifically determined by executing the following set of dynamic control equations using dual adaptive oscillators: In the formula, For the current moment Reconstructed left hip joint angle For the current moment Reconstructed right hip joint angle, For the current moment Left leg offset For the current moment Right leg offset. For the current moment left leg The amplitude of each harmonic. For the current moment Right leg The amplitude of each harmonic; For the current moment left leg The gait phase of a harmonic, For the current moment Right leg The gait phase of a harmonic, , This represents the total number of harmonics.

6. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 5, characterized in that: The current time The parameters of the dual adaptive oscillator are updated as follows: S71. Calculate the current time respectively. Left leg estimation error and the current moment Right leg estimation error : In the formula, For the current moment Left hip joint angle, For the current moment Right leg hip joint angle; S72, Based on the current time Left leg estimation error and the current moment Right leg estimation error The dual adaptive oscillator is corrected in real time using the following formula: In the formula, For the current moment left leg The update rate of the gait phase of each harmonic. For the current moment Right leg The update rate of the gait phase of each harmonic. For the current moment Update rate of left leg offset For the current moment Update rate of right leg offset For the current moment left leg The update rate of the amplitude of each harmonic. For the current moment Right leg The update rate of the amplitude of each harmonic. For the current moment The update rate of the base frequency, For the current moment The fundamental frequency, The preset gait phase learning rate, The preset offset learning rate or amplitude learning rate, The preset base frequency learning rate; S73. The correction formula is discretized using the first-order forward Euler method to obtain the updated parameters. The first-order forward Euler method involves discretizing the current time step by step. The update rate of each parameter of the dual adaptive oscillator and the preset sampling period Multiply and then add up to the current time. The corresponding parameters of the dual adaptive oscillator.

7. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 1, characterized in that: The current time Corrected gait phase is obtained by correcting the gait phase of the corresponding leg, as follows: S511, Calculate the corresponding leg at time [time]. Gait phase error The formula is as follows: In the formula, This represents the moment of the most recent maximum hip extension event within the current sampling period. Represents the corresponding leg at time The initial gait phase, i.e., the time of the dual adaptive oscillator output. Corresponding leg The gait phase of each harmonic; S512, Calculate the current time of each leg. Gait phase correction error term : in, In the formula, For the corresponding leg at time The gait phase correction error term, For the corresponding leg at time Gait phase error gain, For the current moment The fundamental frequency, The preset gain constant; S513, based on the current time of each leg The gait phase correction error term corrects the gait phase of the corresponding leg, thus obtaining the gait phase of the corresponding leg at the current moment. Corrected gait phase The calculation is as follows: In the formula, For modulo operation, For the corresponding leg at the current moment The initial gait phase.

8. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 1, characterized in that: According to the current time The walking pattern category of each leg is dynamically generated to correspond to the leg at the current moment. The reference auxiliary torque is used to complete the gait control of the lower limb exoskeleton, as follows: S61, at the current moment The corresponding leg walking pattern category is At that time, provide the corresponding leg at the current moment. The reference auxiliary torque is zero; at the current moment The corresponding leg walking pattern category is , and At that time, a weighted summation of Gaussian radial basis functions is performed to obtain the corresponding leg at the current time. Reference auxiliary torque The formula is as follows: in, For the corresponding leg at the current moment Corrected gait phase, For the first The peak value of a Gaussian radial basis function For the first The mean of the Gaussian radial basis functions For the first The standard deviation of the Gaussian radial basis functions This represents the number of Gaussian radial basis functions. For global coefficients, Indicates going upstairs. Indicates flat land. Indicates going downstairs. Indicates standing; S62, based on the current time of each leg The reference auxiliary torque is calculated in reverse using a force-potential model to determine the current moment. The target deformation angle of the drive motor corresponding to the hip joint of the leg; S63, Current time The target deformation angle of the drive motor of each leg hip joint and the angle of each leg hip joint are used as the final reference position input to the PID controller to perform closed-loop control of the position of both legs, thus completing the gait control of the lower limb exoskeleton.

9. The lower limb exoskeleton gait control method for multiple scenarios as described in claim 8, characterized in that: The hip joint drive unit of the lower limb exoskeleton includes a housing, a drive motor, a pinion, a gear, a spring, a pretension adjustment unit, a rocker mechanism, a first guide rod slider mechanism, and a thigh link. The rocker mechanism includes a first link and a second link. The housing is fixed to the waist when worn by the human body. The thigh link is rotatably connected to the housing. The gear is fixedly connected to the thigh link. The pinion and gear form an internal meshing gear pair and are eccentrically arranged. The thigh link is fixed to the thigh when worn by the human body. The drive motor is fixedly connected to the housing and is used to drive the pinion to rotate. The pretension adjustment unit and the gear are mounted together on the thigh link. One end of the first link is coaxially fixedly connected to the pinion, and the other end is connected to one end of the second link. The other end of the second link is connected to the first guide rod slider mechanism. The slider mechanism is hinged, and the guide rail of the first guide rod slider mechanism is fixedly connected to the thigh connecting rod and parallel to the length direction of the thigh. One end of the spring is connected to the slider of the first guide rod slider mechanism, and the other end is connected to the preload adjustment unit. The preload adjustment unit is fixedly connected to the thigh connecting rod and is used to adjust the preload of the spring. In the initial state, the center line of the spring is collinear with the rotation centers of the small gear and the large gear, and does not generate torque on the large gear. When the thigh connecting rod rotates around the hip joint with the swing of the thigh, the large gear drives the small gear to rotate, thereby the rocker mechanism drives the spring to stretch or retract, generating an elastic torque on the large gear. At the same time, the drive motor drives the small gear to rotate, thereby the large gear drives the thigh connecting rod to swing. The force-position model is as follows: in, For the current moment The target deformation angle of the drive motor corresponding to the hip joint of the leg. For the stiffness of the spring, Let be the radius of the pinion. This is the preload force of the spring in its initial state.

10. A lower limb exoskeleton gait control system for multiple scenarios, characterized in that: The device includes a controller and two inertial measurement units (IMUs), which respectively detect the hip joint angles of the left and right legs. The controller is used to execute the lower limb exoskeleton gait control method for multiple scenarios as described in any one of claims 1-9.

Images