Flexible lower limb exoskeleton power-assistance control method based on multi-source information fusion

Abstract

The present invention belongs to the technical field of lower limb exoskeleton control, and specifically is a method for assisting the control of a flexible lower limb exoskeleton based on multi-source information fusion. First, plantar pressure, hip joint acceleration and angular velocity, and knee joint acceleration and angular velocity are collected, and the hip joint acceleration and angular velocity, and knee joint acceleration and angular velocity are fused to obtain the hip and knee joint angles; the hip and knee joint angles and plantar pressure are normalized to form a multi-source information matrix; the multi-source information matrix is ​​input into a gait phase recognition model to perform gait phase recognition and obtain the gait phase; then, based on the gait phase, joint angles, and joint torques, the Bowden line that plays an assisting role is determined; the expected change length of the Bowden line under hip and knee joint assistance is calculated respectively; finally, the expected change length of the Bowden line under hip and knee joint assistance is converted into the expected rotation angle of the hip and knee joint motors, and the hip and knee joint motors are controlled using fuzzy PID control based on the rotation angles. This method solves the problem of low gait phase recognition accuracy and difficulty in performing precise assist control in lower limb exoskeleton control.

Description

Technical Field

[0001] The present invention belongs to the technical field of lower limb exoskeleton control, and in particular relates to a flexible lower limb exoskeleton power-assisted control method based on multi-source information fusion. Background Art

[0002] In recent years, assisted exoskeletons designed to enhance the user's coordination and athletic ability while reducing metabolic rate have garnered widespread attention, playing a crucial role in rehabilitation medicine, military operations, and emergency rescue. Flexible exoskeletons consist of a mechanical system and a control system. The control system uses sensors to monitor the user's posture, movements, and force feedback, controlling the exoskeleton's movements.

[0003] Currently, most lower-limb exoskeletons suffer from poor human-machine interaction and insufficient gait intention perception, which impacts control system decision-making and prevents them from achieving the desired assistance effect. There are two main types of existing lower-limb exoskeleton assistance control systems: one that relies on the interaction between the human body and the exoskeleton, and the other that relies on a predetermined trajectory based on the movement state (e.g., running, climbing stairs, walking on flat ground). Both of these assistance control systems focus on the exoskeleton as the main body, failing to consider the human's gait phase and neglecting the human's role as the main body in the human-machine interaction process. This results in poor human-machine interaction and poor assistance effects.

[0004] Existing research on gait phase perception mainly considers single modal information, resulting in inaccurate gait phase division. For example, the gait phase of a single leg can be divided into the pre-stance phase, mid-phase, late phase, and swing phase based on plantar pressure. However, since the plantar pressure in the swing phase is always zero, it cannot be further subdivided. Gait phase division based on joint angles cannot determine whether the lower limb has touched the ground and entered the stance phase, because the angles of each joint are different when different people touch the ground. Failure to accurately identify the gait phase will affect the precise control of the lower limb exoskeleton.

[0005] Furthermore, during the power-assistance process, the precise modeling of the Bowden cable trajectory and the calculation of its varying length directly impact the motion of the joint motor. When switching between front and rear Bowden cable power assistance, the joint motor motion exhibits a step-like motion, further increasing the difficulty of power-assistance control. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the technical problem that the present invention intends to solve is to provide a flexible lower limb exoskeleton power-assistance control method based on multi-source information fusion, so as to solve the problem of low accuracy of human gait phase recognition in lower limb exoskeleton control, making it difficult to perform precise control and provide appropriate power assistance.

[0007] The present invention solves the technical problem by adopting the following technical solutions:

[0008] A method for assisting control of a flexible lower limb exoskeleton based on multi-source information fusion, characterized in that the method comprises the following steps:

[0009] Step S1: Collect plantar pressure, hip joint acceleration and angular velocity, knee joint acceleration and angular velocity, fuse the hip joint acceleration and angular velocity, and knee joint acceleration and angular velocity, respectively, to obtain hip and knee joint angles; normalize the hip and knee joint angles and plantar pressure to form a multi-source information matrix; input the multi-source information matrix into a gait phase recognition model to perform gait phase recognition and obtain gait phase;

[0010] Step S2: Determine the Bowden line that provides assistance based on the gait phase, joint angle, and joint torque; perform kinematic modeling on the flexible lower limb exoskeleton to obtain the length of the Bowden line without assistance. ; Calculate the tensile deformation of the Bowden cable under power assistance according to the following formula;

[0011] (2)

[0012] Where, is the tension of joint k, when hour, is the hip joint pull; when hour, It is the tension of knee joint; The tensile deformation of the Bowden cable under the assistance of joint k;

[0013] Calculate the expected joint torque according to the following formula:

[0014] (3)

[0015] Where, is the expected assist torque of joint k, is the angle of joint k, u is the coefficient, and d is the constant term;

[0016] Calculate the expected length change of the Bowden cable under hip and knee joint assistance according to the following formula;

[0017] (5)

[0018] Where, The expected change in length of the Bowden cable under the assistance of joint k, is the power lever arm of joint k;

[0019] Step S3: Convert the expected length change of the Bowden cable under the power assist of the hip and knee joints into the expected rotation angles of the hip and knee joint motors respectively, and control the hip and knee joint motors using fuzzy PID control according to the rotation angles.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] The present invention fully considers the gait phase of a person's walking process and performs gait phase identification based on multimodal data (plantar pressure, hip joint angle, and knee joint angle). This facilitates the precise determination of the Bowden cable that provides power, enabling precise power-assisted control. After determining the Bowden cable that provides power, the expected length change of the Bowden cable under power assistance is calculated, and this expected length change of the Bowden cable under power assistance is converted into the expected rotation angle of the joint motor, thereby controlling the joint motor. Because the joint motor movement exhibits a step phenomenon when switching between the front and rear Bowden cable power assistance, fuzzy PID control is used to control the joint motor to alleviate the step phenomenon. This method achieves precise power-assisted control of the flexible lower limb exoskeleton, improving the human-machine collaborative control capabilities of the entire exoskeleton system, thereby enabling the exoskeleton to accurately provide power assistance. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is the overall flow chart of the present invention;

[0023] Figure 2 is a flow chart of the perception layer of the present invention;

[0024] Figure 3 is the gait phase division diagram;

[0025] Figure 4 is a flow chart of the conversion layer of the present invention;

[0026] Figure 5 It is a flow chart of the control layer of the present invention. DETAILED DESCRIPTION

[0027] Specific embodiments are given below in conjunction with the accompanying drawings. The specific embodiments are only used to introduce the technical solutions of the present invention in detail and are not intended to limit the scope of protection of the present application.

[0028] The present invention provides a flexible lower limb exoskeleton power-assisted control method based on multi-source information fusion (hereinafter referred to as the method, see Figures 1 to 5 ), including the following steps:

[0029] Step S1: The sensing layer collects plantar pressure, hip joint acceleration and angular velocity, and knee joint acceleration and angular velocity, and fuses the hip joint acceleration and angular velocity and knee joint acceleration and angular velocity respectively to obtain the hip and knee joint angles; the multi-source information matrix is ​​input into the gait phase recognition model to perform gait phase recognition and realize gait phase perception;

[0030] S11. Plantar pressure is collected in real time through a pressure sensor. The acceleration and angular velocity of the hip joint are collected through an accelerometer and gyroscope located at the hip joint. The acceleration and angular velocity of the knee joint are collected through an accelerometer and gyroscope located at the knee joint. Since muscle movement and vibration caused by heel contact during walking can lead to inaccurate sensor measurements, and the joint angle solved by a single acceleration is easily affected by high-frequency noise, which causes error accumulation and inaccurate solution results. Therefore, an extended Kalman filter is used to fuse the acceleration and angular velocity of the hip joint and the acceleration and angular velocity of the knee joint to obtain the hip joint angle and the knee joint angle.

[0031] S12. Normalize the plantar pressure and joint angle separately, scaling the data to a uniform range to eliminate dimensional differences or deviations between different features. The normalization calculation formula is shown in Equation (1). The normalized plantar pressure, hip joint angle, and knee joint angle form a multi-source information matrix.

[0032] (1)

[0033] Where, 、 are the data before and after normalization, 、 Normalize the maximum and minimum values ​​in the data separately;

[0034] S13. Use the kernel principal component analysis method to reduce the dimensionality of the multi-source information matrix to obtain the reduced dimensionality multi-source information matrix, retaining the main structural features of the data, which is conducive to improving the efficiency of the gait phase recognition model; select the least squares support vector machine as the gait phase recognition model, input the reduced dimensionality multi-source information matrix into the gait phase recognition model, and identify the gait phase; during the recognition process, use the particle swarm optimization algorithm to optimize the regularization parameter and kernel width of the least squares support vector machine.

[0035] like Figure 3As shown in the figure, the gait cycle of a single leg of the human body is divided into 6 phases, namely the early, middle and late stance phase and the early, middle and late swing phase. The gait phase recognition model recognizes the gait phase according to the following basis: (a) Early stance phase: from the beginning of the heel touching the ground to the sole of the foot touching the ground, the hip joint angle decreases, the knee joint angle increases, and both the hip and knee joints generate extension torque; (b) Middle stance phase: during the stage of full sole landing, the heel pressure gradually decreases, the sole and toe pressure gradually increase, and the hip and knee joints change from extension torque to flexion torque; (c) Late stance phase: from the heel off the ground to the toe off the ground, During the entire process, plantar pressure is mainly distributed in the sole and toe areas, and both the hip and knee joints generate flexion torques. When the toes leave the ground, the support phase ends and the swing phase begins. (d) Early swing phase: As the toes leave the ground, the hip joint drives the knee joint to flex until the flexion angle of the knee and hip joints is maximum. The hip joint generates a flexion torque and the knee joint generates an extension torque. (e) Mid-swing phase: The knee joint begins to extend from the maximum flexion state until the calf is in an upright position. (f) Late swing phase: Starting from the upright position of the calf until the heel touches the ground, the hip joint generates an extension torque and the knee joint generates a flexion torque in the mid-to-late swing phase.

[0036] Step S2: The conversion layer determines the Bowden line that plays an assisting role based on the gait phase, joint angle, and joint torque obtained by the perception layer; calculates the expected change length of the Bowden line under hip joint assistance and the expected change length of the Bowden line under knee joint assistance, such as Figure 4 shown.

[0037] S21. Determine the Bowden line that provides assistance based on gait phase, joint angles, and joint torques;

[0038] S22. Kinematic modeling of the flexible lower limb exoskeleton. Based on the principle of non-interference between the front and rear Bowden lines and the change of joint angles, the length of the Bowden line under no assistance is obtained. ;

[0039] S23. The wearer stands with his feet crossed, with the unassisted leg 50 cm behind the assisted leg. The dynamometer is placed at the end of the Bowden cable. The hip joint motor and the knee joint motor are respectively stretched by 30 mm in front and back of the Bowden cable. The position of the Bowden cable and the joint tension are recorded. Repeat this process several times. The quadratic polynomial is used to fit the tensile deformation of the assisted Bowden cable and the joint tension using the least squares method. The coefficients a, b, and c of the quadratic polynomial are determined. The relationship between the joint tension and the tensile deformation of the assisted Bowden cable is obtained as shown in formula (2). The tensile deformation of the assisted Bowden cable is calculated based on this relationship.

[0040] (2)

[0041] Where, is the tension of joint k, when hour, is the hip joint pull; when hour, It is the tension of knee joint; The tensile deformation of the Bowden cable under the assistance of joint k;

[0042] S24, there is a linear relationship between the joint expected assist torque and the joint angle, so the joint expected assist torque and the joint angle are fitted to obtain a fitting equation; taking 50% of the joint torque in the current gait phase as the benchmark, with the minimum root mean square error and assist torque continuity as the goal, the coefficient u of the fitting equation is determined, and then the relationship between the joint expected assist torque and the joint angle as shown in formula (3) is obtained, and the joint expected assist torque is calculated based on this relationship;

[0043] (3)

[0044] (4)

[0045] Where, is the expected assist torque of joint k, is the angle of joint k, RMSE is the root mean square error, is the expected assist torque of joint k at time m, is 50% of the moment of joint k at time m, and n is the number of data sampling times;

[0046] S25, converting the expected power torque of the joint into the expected power size, and calculating the expected change length of the Bowden line under power assistance of the hip joint and knee joint respectively by formula (5);

[0047] (5)

[0048] Where, The expected change in length of the Bowden cable under the assistance of joint k, is the assist lever arm of joint k.

[0049] Step S3: The execution layer converts the expected change length of the Bowden cable under power assistance into the expected rotation angle of the joint motor. According to the expected rotation angle of the joint motor, the joint motor is controlled by fuzzy PID control to alleviate the step phenomenon of the joint motor motion trajectory caused by the switching of the front and rear Bowden cable power assistance, such as Figure 5 shown.

[0050] S31, will help the expected change in length of the Bowden cable Convert it into the expected rotation angle of the joint motor to obtain the expected rotation angle of the hip joint motor and the expected rotation angle of the knee joint motor. The calculation formula is as follows:

[0051] (6)

[0052] Where, is the desired rotation angle of the joint motor, is the radius of the turntable where the joint motor and the Bowden cable connect.

[0053] S32. Fuzzy PID control is used to control the hip joint motor and the knee joint motor respectively. Taking the control of the hip joint motor as an example, the rotation angle error of the hip joint motor (the difference between the expected rotation angle and the actual rotation angle of the hip joint motor) and the error change rate are used as inputs of the fuzzy controller. The fuzzy controller adaptively adjusts the PID parameters according to the fuzzy rules to obtain the adjusted PID parameters, which are as follows:

[0054] (7)

[0055] Where, For the manual setting value of PID parameters, is the PID parameter change, are the PID parameters after tuning;

[0056] The adjusted PID parameters are applied to the PID controller. The PID controller outputs a current signal for controlling the hip joint motor. The hip joint motor is controlled according to the current signal output by the PID controller. The output of the PID controller is expressed as:

[0057] (8)

[0058] Where u(t) is the current signal output by the PID controller, and e(t) is the rotation angle error of the hip joint motor;

[0059] Similarly, the knee joint motor is controlled to complete the control of the flexible lower limb exoskeleton.

[0060] The DC motor control based on fuzzy PID has the characteristics of fast response speed, small overshoot and strong stability. Using fuzzy PID control to control the joint motor can effectively alleviate the step phenomenon of the joint motor motion trajectory caused by the switching of the front and rear Bowden cable power assistance, reduce the peak error and total error of the motor motion trajectory, reduce the overshoot phenomenon, and effectively provide more precise power assistance.

[0061] The present invention does not describe any prior art applications.

Claims

1. A flexible lower limb exoskeleton power-assisted control method based on multi-source information fusion, characterized in that: The method comprises the following steps: Step S1: Collect plantar pressure, hip joint acceleration and angular velocity, knee joint acceleration and angular velocity, fuse the hip joint acceleration and angular velocity, and knee joint acceleration and angular velocity, respectively, to obtain hip and knee joint angles; normalize the hip and knee joint angles and plantar pressure to form a multi-source information matrix; input the multi-source information matrix into a gait phase recognition model to perform gait phase recognition and obtain gait phase; Step S2: Determine the Bowden line that provides assistance based on the gait phase, joint angle, and joint torque; perform kinematic modeling on the flexible lower limb exoskeleton to obtain the length of the Bowden line without assistance. ; Calculate the tensile deformation of the Bowden cable under power assistance according to the following formula; (2) Where, is the tension of joint k, when hour, is the hip joint pull; when hour, It is the tension of knee joint; The tensile deformation of the Bowden cable under the assistance of joint k; Calculate the expected joint torque according to the following formula: (3) Where, is the expected assist torque of joint k, is the angle of joint k, u is the coefficient, and d is the constant term; Calculate the expected length change of the Bowden cable under hip and knee joint assistance according to the following formula; (5) Where, The expected change in length of the Bowden cable under the assistance of joint k, is the power lever arm of joint k; Step S3: Convert the expected length change of the Bowden cable under the power assist of the hip and knee joints into the expected rotation angles of the hip and knee joint motors respectively, and control the hip and knee joint motors using fuzzy PID control according to the rotation angles.

2. The flexible lower limb exoskeleton power-assisted control method based on multi-source information fusion according to claim 1 is characterized in that: The gait phase recognition model adopts a least squares support vector machine, and uses a particle swarm optimization algorithm to optimize the regularization parameter and kernel width of the least squares support vector machine.

3. The flexible lower limb exoskeleton power-assisted control method based on multi-source information fusion according to claim 1 or 2, characterized in that: In the process of fitting equation (3), the coefficient u is determined by the following equation: (4) Where RMSE is the root mean square error, is the expected assist torque of joint k at time m, is 50% of the torque of joint k at time m, and n is the number of data sampling times.

4. The flexible lower limb exoskeleton power-assisted control method based on multi-source information fusion according to claim 3 is characterized in that: The extended Kalman filter is used to fuse the acceleration and angular velocity of the hip joint and the acceleration and angular velocity of the knee joint to obtain the hip and knee joint angles.

Images