A method for generating exoskeleton bionic assistance curves for non-stationary human motion

Abstract

The present invention relates to a method for generating an exoskeleton bionic power-assistance curve for non-stationary motion of the human body, and improves the synchronization and bionic capabilities of the power-assistance curve of the ankle joint exoskeleton robot for non-stationary motion of the human body. The invention uses an inertial measurement unit to collect the angular data of the human calf, and uses the data as the independent variable of the power-assistance curve; designs a power-assistance curve model modeling method based on a double Gaussian function to achieve shape planning of the power-assistance curve; uses the measurement data of two inertial measurement units of the human calf and the instep of the foot to achieve the estimation of model parameters, and improves the similarity between the exoskeleton power-assistance curve and the human body's biological torque curve. The present invention is designed from two dimensions: the independent variable and the model of the power-assistance curve, respectively ensuring the human-machine motion synchronization capability of the power-assistance curve under non-stationary motion of the human body, and the bionic capability of the power-assistance curve and the human body's biological torque curve, which helps to promote the application of exoskeleton robots in complex and unstructured task scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of exoskeleton robots, and in particular to a method for generating an exoskeleton bionic power-assistance curve for non-stationary motion of the human body. Background Art

[0002] Exoskeleton robots are wearable, intelligent devices that assist patients with limb movement and enhance the physical functions of healthy individuals. However, current exoskeleton technology has drawbacks, primarily due to the generated assistance curves being out of sync with human-machine motion and failing to meet the biomechanical requirements of the human body. This phenomenon is particularly evident in the irregular and non-periodic nature of non-stationary human motion, such as sudden speed changes, gait transitions, and obstacle crossing. Therefore, to promote the application of exoskeleton robots in complex scenarios, it is crucial to generate exoskeleton-inspired assistance curves for non-stationary human motion.

[0003] At present, there are two methods for generating the power assistance curve of exoskeleton robots: one is to first identify the type of human activity and then switch to different power assistance curves, and the other is to collect sensor data and directly adjust the power assistance curve in stages. However, there are two problems with the current methods: (1) they often rely on equipment and instruments in a laboratory environment, such as respirators, functional treadmills, and motion capture systems; (2) they only consider the optimization of the longitudinal dimension of the power assistance curve (y-axis, curve amplitude and shape), while ignoring the synchronization problem between the transverse dimension (x-axis, curve model independent variable) and human movement. At present, some work has begun to try to use small devices such as electromyography sensors, foot pressure sensors, and inertial measurement units to achieve the optimization design of the power assistance curve, but there is still no simultaneous consideration of the optimization design problem of the two dimensions of the power assistance curve. When time is used as the independent variable of the curve model, if the human body suddenly changes speed, it is difficult to synchronize the power assistance curve points with the human body movement points, that is, the power assistance curve no longer meets the needs of the human body. Therefore, some studies have proposed using the human body state quantity as the independent variable of the power assistance curve to improve the synchronization between the power assistance curve and human movement. The human body state quantity is a potential feasible solution for optimizing the horizontal dimension of the power assistance curve, but what is urgently needed now is to conduct simultaneous and integrated analysis of the optimization design of the two dimensions of the power assistance curve, and to design a power assistance curve that can simultaneously meet the human power assistance needs in the vertical dimension and continuously synchronize human movement in the horizontal dimension. The generation and implementation of this power assistance curve will greatly enhance the adaptability of the exoskeleton robot to complex scenarios and extreme tasks. Summary of the Invention

[0004] In view of the above-mentioned shortcomings and deficiencies in the existing technology, the present invention aims to provide a method for generating an exoskeleton biomimetic assistance curve for non-stationary human motion. This method, applied to an exoskeleton robot system that includes human ankle joint assistance, enhances the synchronization and biomimetic capabilities of the exoskeleton's assistance curve for non-stationary human motion. This method improves the exoskeleton's ability to synchronize motion with a human during non-stationary motion and complex tasks, achieving biomimetic assistance for human joints.

[0005] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion, which is used to improve the synchronization and bionic ability of the exoskeleton robot power-assistance curve for non-stationary human motion, comprising the following steps:

[0006] Two inertial measurement units are used to collect data on the calf swing angle and the dorsum of the foot during human movement.

[0007] Identify the gait phase and events of the current limb side based on the dorsum of the foot state data, obtain the calf swing angle values ​​at the two moments of foot landing and foot leaving the ground; calculate the human ankle joint angle on the current limb side;

[0008] Using data on the human ankle joint angle during the entire stance phase of gait, we analyzed the consistent correspondence between the maximum ankle dorsiflexion angle and the maximum ankle plantar flexion moment, and extracted the calf swing angle corresponding to the moment of maximum ankle dorsiflexion angle.

[0009] The parameters of a power assistance curve model based on a double Gaussian function are estimated using the calf swing angle corresponding to the foot landing event, foot lift-off event, and ankle joint maximum dorsiflexion angle to generate an ideal curve model.

[0010] The ideal power assist curve model is used to estimate the power assist value of the exoskeleton robot that meets the synchronization and bionic requirements for any calf swing angle.

[0011] The calf inertial measurement unit is set at any position between the ankle and the knee joint to measure the swing angle of the calf in the sagittal plane, and the dorsum inertial measurement unit is set at any position of the foot to measure foot status data.

[0012] The inertial measurement units are all arranged at locations where there is little muscle, skin and soft tissue that are not easily moved.

[0013] The inertial measurement unit is reset to zero when the human body stands still and upright.

[0014] The gait phase and gait event recognition refers to the use of the foot movement angle, angular velocity and acceleration data collected by the inertial measurement unit of the dorsum of the foot, and the use of pattern recognition methods to realize the recognition of the stance phase and swing phase, as well as the foot landing and foot leaving events.

[0015] The human ankle joint angle measurement is obtained by subtracting the human sagittal plane direction angles measured by the two inertial measurement units of the calf and the dorsum of the foot, under the premise of ensuring that the data measurements meet the synchronization.

[0016] The moment of maximum ankle joint dorsiflexion angle is obtained by obtaining the extreme point of the ankle joint angle data curve through an extreme value recognition method to obtain the corresponding moment.

[0017] The power assistance curve model based on the double Gaussian function is:

[0018]

[0019] Where τ(θ) represents the power assist curve value when the calf swing angle is θ, where the calf swing angle θ is negative when the ankle joint is in the backward position relative to the knee joint, positive when it is in the forward position, and zero when it is in the same vertical plane; A represents the maximum value of the power assist curve, μ represents the middle position of the power assist curve, σ1 and σ2 determine the shapes of the power assist curve in the ascending stage and the descending stage, respectively, and θ TO represents the calf swing angle at the moment of foot leaving the ground, θ HS represents the shank swing angle at the moment of foot contact event.

[0020] The estimated parameters of the double Gaussian boost curve model refer to the online estimation of the three parameters μ, σ1, and σ2.

[0021] Furthermore, the online parameter estimation method for estimating the double Gaussian boost curve model is:

[0022] (1) where μ is estimated to be the calf swing angle corresponding to the maximum ankle dorsiflexion angle, and the symbol is θ MDF ;

[0023] (2) where σ1 is estimated as:

[0024]

[0025] (3) where σ2 is estimated as:

[0026]

[0027] The exoskeleton robot assist value for any calf swing angle is estimated by using an ideal curve model, by inputting the latest measured calf swing angle θ, and calculating the exoskeleton robot assist value τ(θ) corresponding to the calf swing angle, which is used as at least one of a rotational assist torque or a linear assist force for the exoskeleton robot to provide walking assistance.

[0028] The present invention has the following beneficial effects and advantages:

[0029] The present invention only uses two inertial measurement units, the calf and the instep, and proposes a method for generating an exoskeleton bionic power assistance curve for non-stationary human motion. This method innovatively uses the calf swing angle as the process variable (horizontal axis variable or function independent variable) of the power assistance curve to ensure the continuous synchronization of the human-machine motion state; a double Gaussian function is used as the model function of the power assistance curve, and the online estimation and update of the model parameters can be realized. This ensures that even in complex and non-stationary motion scenarios such as sudden acceleration and deceleration, sudden stop, and crossing obstacles of the human body, the synchronization of the exoskeleton robot and the human body movement can still be guaranteed, and the bionic nature of the power assistance curve to the human body's biological torque can be guaranteed, thereby greatly improving the adaptability of the exoskeleton robot, meeting the diverse assistance needs of users, and promoting the daily application of the exoskeleton robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is an overall flow chart of the implementation process of the present invention and a schematic diagram of the position of the inertial measurement unit;

[0031] Figure 2 is a corresponding relationship diagram between the ankle joint angle and the joint torque in the present invention;

[0032] Figure 3 This is a diagram proving that the present invention is applicable to activities such as walking on flat ground, running, and walking on a slope;

[0033] Figure 4 This is a schematic diagram of generating a power assist curve based on a double Gaussian function in the present invention;

[0034] Figure 5 This is a graph showing the actual generation of power assistance curves in experiments with different types of activities;

[0035] Figure 6 It is a similarity comparison diagram of the power assistance curve and the human body torque curve when facing non-stationary motion in different types of activities. DETAILED DESCRIPTION

[0036] To make the above-mentioned objects, features, and advantages of the present invention more readily apparent, the specific implementation methods of the present invention are described in detail below with reference to the accompanying drawings. The following description sets forth many specific details to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways than those described herein, and those skilled in the art may make similar modifications without departing from the scope of the invention. Therefore, the present invention is not limited to the specific implementation methods disclosed below.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art of the art to which the present invention pertains. The terms used in the specification of the invention herein are for the purpose of describing specific embodiments only and are not intended to limit the present invention.

[0038] The method of the present invention is used for an exoskeleton robot system including human ankle joint motion assistance, and is used to improve the synchronization and biomimetic ability of the exoskeleton robot's assistance curve for the non-stationary motion of the human body. The method uses an inertial measurement unit to collect the directional angle data of the human calf, and uses the data as the independent variable of the assistance curve; designs a power assistance curve model modeling method based on a double Gaussian function to achieve shape planning of the assistance curve; uses the measurement data of two inertial measurement units of the human calf and the instep of the foot to estimate the parameters of the double Gaussian function, and the estimated frequency is equivalent to the gait frequency, thereby improving the similarity between the exoskeleton robot's assistance curve and the human ankle joint biological torque curve, that is, biomimetic. The present invention is designed from two dimensions: the independent variable and the model of the assistance curve, respectively ensuring the human-machine motion synchronization ability of the assistance curve under the non-stationary motion of the human body, and ensuring the biomimetic ability of the assistance curve and the human biological torque curve, which helps to promote the application of exoskeleton robots in complex and unstructured task scenarios and realize the daily and practical application of exoskeleton robots.

[0039] The power assist curve of the exoskeleton robot is completely determined by two aspects: (1) the independent variable of the curve function, and (2) the model of the curve function, and the output is the power assist value. The independent variable determines the change process of the curve and determines the synchronization of human-machine movement, while the function model determines the shape of the curve and determines the bionics of the power assist curve. The current independent variables of the power assist curve are time and gait cycle percentage. However, because the non-stationary motion of the human body has temporal non-uniformity, that is, the motion process is not linearly increasing in time, considering that the slope of time is 1 and the slope of the gait cycle percentage is 1 / T, where T is the current gait cycle value, both are constants, so they cannot be used to describe the non-stationary motion process. For example Figure 1 As shown, the method in the invention uses the inertial measurement unit of the calf to collect the calf swing angle of the human body during movement, and collects and analyzes the calf swing angle between the moment the foot touches the ground and the moment the foot leaves the ground. Taking into account the monotonicity of the calf swing angle in this time period, the invention uses the calf swing angle as the independent variable of the power assistance curve, wherein the start and end times of the time period can be determined by common pattern recognition methods, such as KNN, hidden Markov model, etc., by identifying gait events to obtain the endpoint time.

[0040] The bionic characteristics of the exoskeleton robot's power-assistance curve are the user's expectations and needs. In terms of the bionic shape of the power-assistance curve, the invention utilizes the continuous correspondence between the maximum dorsiflexion angle of the human ankle joint and the maximum plantar flexion torque of the ankle joint. Figure 2 As shown in , during walking on flat ground, the maximum dorsiflexion moment of the human ankle joint corresponds to the maximum plantar flexion moment, and this correspondence is also applicable in other types of human activities. Figure 3 As shown in the figure, during human activities such as walking on flat ground with variable speed, walking on a slope with variable gradient, and running on flat ground with variable speed, the maximum dorsiflexion angle of the human ankle joint and the maximum plantar flexion torque maintain a corresponding pattern. Therefore, the method of the invention is to correspond the maximum moment of the power assistance curve with the maximum moment of the ankle joint dorsiflexion angle, that is, the model parameter μ is estimated as the calf swing angle θ corresponding to the maximum moment of the dorsiflexion angle. MDF , thereby achieving bionics at the peak moment.

[0041] As shown in Figure 4, this method models the power assist curve as a double Gaussian function, where the red dashed line represents the first segment of the Gaussian function, whose action time is from the moment of foot contact to the moment of ankle dorsiflexion angle maximum; and the black solid line represents the second segment of the Gaussian function, whose action time is from the moment of ankle dorsiflexion angle maximum to the moment of foot lift-off. The mathematical expression of the overall power assist curve is as follows:

[0042]

[0043] Among them, τ and θ represent the power value and the calf swing angle respectively. The calf swing angle is 0 when the human body is standing upright, negative when the ankle joint is located behind the knee joint in the sagittal plane, and positive when the ankle joint is located in front of the knee joint in the sagittal plane. Therefore, during the development of the human foot stance phase, the calf swing angle gradually transitions from positive to negative; and the power effect interval is θ TO and θ HS , respectively, represent the calf swing angle corresponding to the moment of foot departure and the moment of foot contact; the double Gaussian function has four parameters: A, μ, σ1, and σ2. Parameter A determines the amplitude of the entire power curve, which is specified as 15% of the subject's current weight in this method; and the parameter estimation method of μ has been explained above, which is equivalent to the calf swing angle θ corresponding to the moment of maximum dorsiflexion angle. MDF ; The parameter estimation methods of σ1 and σ2 are as follows Figure 4 As shown below:

[0044]

[0045]

[0046] The division by 4 is to reduce the numerical jump of the power assist curve at the endpoints on both sides, so as to achieve a smooth transition of the power assist phase. This value can be changed according to actual needs, and can even be changed online to adapt to more tasks and demand scenarios.

[0047] A human motion experiment was conducted on the above-mentioned power-assistance curve generation method to demonstrate (1) the curve generation capability of the method in the face of different types of human activities, and (2) the similarity between the generated curve and the human biological torque curve when encountering non-stationary human motion, that is, the method has biomimetic capabilities. A total of 7 subjects participated in the experiment (6 males and 1 female; human body information is expressed in the form of mean ± standard deviation: age 26.9 ± 2.7 years, height 1.72 ± 0.07 m, weight 62.9 ± 3.8 kg). All subjects were required to wear a flexible ankle joint power-assistance exoskeleton robot and walk on a treadmill. The exoskeleton robot has inertial measurement units on the lower leg and the dorsum of the foot, thus meeting the sensor quantity and position requirements required for the implementation of the present invention. The treadmill speed was selected to be the optimal speed for the subject. Before the experiment began, participants were familiarized with walking on the treadmill and determined their optimal speed. (Once the optimal speed was determined, the treadmill speed remained constant for that participant throughout the experiment, but participants were free to modify their preferred walking rhythm and gait pattern, such as by changing step length or phase ratio.) Participants were asked to perform four types of activities: walking on level ground, running on level ground, walking uphill (+10 degrees), and walking downhill (-10 degrees).

[0048] like Figure 5 Figure 2 shows the assistance curves generated by the exoskeleton robot during the experiment. The multiple curves in the figure represent data from multiple gait cycles. This figure demonstrates that this technique can generate different types of assistance curves based on the type of activity. The curve's range and shape can be modified using the online estimation of the curve model parameters μ, σ1, and σ2, as well as the measured calf swing angle, to meet the subject's assistance needs for the given activity.

[0049] At the same time, in order to further verify the bionic function of the curve generation method in the face of non-stationary human motion, the experiment requires the treadmill to randomly perform rapid acceleration (increasing the treadmill speed by 80% within 0.05 seconds and then returning to the original speed within the next 0.05 seconds) and rapid deceleration (the same as the rapid acceleration process), inducing non-stationary motion of the human body through the rapid change of the treadmill speed. Figure 6The figure shows the force-assisted curve generated by the exoskeleton robot and the ankle biotorque curve of the subject (derived from human data collected using motion capture and a plantar pressure treadmill, followed by biomechanical analysis software). The values ​​are normalized to facilitate similarity or biomimetic comparison. The similarity between the two curves is quantified using the Pearson Correlation Coefficient. The biomimetic coefficients for rapid deceleration and acceleration during level walking are 0.90 and 0.92, respectively; for level running, they are 0.94 and 0.98; for uphill walking, they are 0.90 and 0.97; and for downhill walking, they are 0.90 and 0.96. A coefficient greater than 0.6-0.8 indicates clear similarity between the two curves, demonstrating that the curves generated by this method maintain biomimetic consistency with the torque curves of the target human joints, even in different activities and under various non-stationary motion conditions.

[0050] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should be regarded as within the scope of protection of the present invention.

Claims

1. A method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion, characterized in that: The method is used to improve the synchronization and bionic capabilities of the exoskeleton robot's power curve for the non-stationary motion of the human body, including the following steps: Two inertial measurement units are used to collect data on the calf swing angle and the dorsum of the foot during human movement. Identify the gait phase and events of the current limb side based on the dorsum of the foot state data, obtain the calf swing angle values ​​at the two moments of foot landing and foot leaving the ground; calculate the human ankle joint angle on the current limb side; Using data on the human ankle joint angle during the entire stance phase of gait, we analyzed the consistent correspondence between the maximum ankle dorsiflexion angle and the maximum ankle plantar flexion moment, and extracted the calf swing angle corresponding to the moment of maximum ankle dorsiflexion angle. The parameters of a power assistance curve model based on a double Gaussian function are estimated using the calf swing angle corresponding to the foot landing event, foot lift-off event, and ankle joint maximum dorsiflexion angle to generate an ideal power assistance curve model. The ideal power assist curve model is used to estimate the power assist value of the exoskeleton robot that meets the synchronization and bionic requirements for any calf swing angle.

2. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: The calf inertial measurement unit is set at any position between the ankle and knee joint to measure the swing angle of the calf in the sagittal plane, and the dorsum inertial measurement unit is set at any position of the foot to measure the foot status data.

3. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: The inertial measurement units are all arranged at locations where there is little muscle, skin and soft tissue that are not easily moved.

4. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: The inertial measurement unit is reset to zero when the human body stands still and upright.

5. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: The gait phase and gait event recognition refers to the use of the foot movement angle, angular velocity and acceleration data collected by the inertial measurement unit of the dorsum of the foot, and the use of pattern recognition methods to realize the recognition of the stance phase and swing phase, as well as the foot landing and foot leaving events.

6. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: The human ankle joint angle measurement is obtained by subtracting the human sagittal plane direction angles measured by the two inertial measurement units of the calf and the dorsum of the foot, under the premise of ensuring that the data measurements meet the synchronization.

7. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: The moment of maximum ankle joint dorsiflexion angle is obtained by obtaining the extreme point of the ankle joint angle data curve through an extreme value recognition method to obtain the corresponding moment.

8. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: The ideal power curve model based on the double Gaussian function is: Where τ(θ) represents the power assist curve value when the calf swing angle is θ, where the calf swing angle θ is negative when the ankle joint is in the backward position relative to the knee joint, positive when it is in the forward position, and zero when it is in the same vertical plane; A represents the maximum value of the power assist curve, μ represents the middle position of the power assist curve, σ1 and σ2 determine the shapes of the power assist curve in the ascending stage and the descending stage, respectively, and θ TO represents the calf swing angle at the moment of foot leaving the ground, θ HS Indicates the calf swing angle at the moment of foot landing.

9. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: Estimating the parameters of the double Gaussian boost curve model refers to the online estimation of the three parameters μ, σ1, and σ2.

10. The method for generating an exoskeleton bionic power-assistance curve for non-stationary human motion according to claim 1, characterized in that: Estimating the exoskeleton robot's assist value for any calf swing angle refers to using an ideal assist curve model to input the latest measured calf swing angle θ, and calculate the exoskeleton robot's assist value τ(θ) corresponding to the calf swing angle, which is used as at least one of a rotational assist torque or a linear assist force for the exoskeleton robot to provide walking assistance.

Images