Power-assisted robot joint motor compliance control method

Abstract

The invention discloses a compliance control method for a joint motor of a power-assisted robot. The compliance control method comprises the following steps: 1) acquiring human body motion data by using an IMU (Inertial Measurement Unit) and a plantar pressure sensor; 2) processing the human body motion data, and predicting the current gait cycle of the human body; 3) based on the current gait cycle of the human body, generating an expected power-assisted curve, and taking the maximum value of the expected power-assisted curve as an expected power-assisted basic value; 4) calculating the fuzzy increment entropy of the electromyographic signal; 5) determining a human body fatigue stage according to the electromyographic signal fuzzy increment entropy; (6) based on the human body fatigue stage, the expected assistance basic value is adjusted, and an expected assistance value is generated; (7) based on the expected assistance value, the joint motor control quantity of the exoskeleton robot is generated; and 8) based on the joint motor control quantity of the exoskeleton robot, driving a motor to perform joint control of the exoskeleton robot. Potential safety hazards caused by man-machine mismatching are eliminated, the flexibility of joints is improved, and better assistance effect and safety are achieved.

Description

Technical Field

[0001] The invention relates to the field of robot joint motor control, and in particular to a compliant control method for a power-assisted robot joint motor. Background Art

[0002] Power-assisted robots are mechanical devices that can enhance human working ability and are widely used in industries such as industry, medical treatment and services. They achieve precise operation through power systems, sensors and control systems, and can assist in completing repetitive, tedious or high-precision work. Power-assisted robots have the advantages of improving work efficiency, reducing labor costs and improving safety, but they also face challenges in terms of technical maturity, cost and ethical laws. Among them, exoskeleton robots are a highly integrated biomechanical and electrical integration system that cleverly integrates technologies from multiple fields such as computers and mechanical engineering. They can intelligently drive the power system based on human motion information, realize real-time information exchange and perception, and precise human-machine collaboration. This system significantly enhances human physical efficiency and expands the boundaries of human capabilities by providing motion assistance to the wearer. As a representative of them, lower limb exoskeleton robots are widely used in military, medical rehabilitation and industry. They not only reduce the combat burden of soldiers and improve the rehabilitation effect of disabled people, but also effectively protect the health of the spine and lumbar spine of industrial workers, fully demonstrating their great potential in enhancing human functions, improving work efficiency and ensuring human safety.

[0003] The compliance control of the exoskeleton robot joints can control the exoskeleton to move in the desired direction and position, while adjusting the position gain of the joints according to the stiffness of the contact force, so that the exoskeleton joint angle control is adapted to the stiffness of the force. This control method can make the exoskeleton more flexible during movement, reduce the interference force between the wearer and the wearer, and improve the wearer's comfort, thereby helping patients with nerve damage or limb disabilities to conduct rehabilitation training and restore motor function. During the rehabilitation process, the exoskeleton can provide appropriate assistance and support based on the patient's movement intentions and force conditions.

[0004] The active-assisted lower limb rehabilitation exoskeleton robot is a device that assists disabled patients in lower limb rehabilitation training. It uses advanced robotics technology and intelligent control systems to help patients with gait training and muscle strength recovery by providing power and assistance, thereby improving patients' walking ability and quality of life. At present, the generation of joint assistance curves for exoskeleton robots mainly relies on the biological torque of the lower limb joints of the human body. During the walking process of the human body, the joint torque will also have periodic changes in a gait cycle. The use of Gaussian models to fit the assistance curve online has strict requirements on data distribution, and the influence of outliers is obvious, and the calculation complexity is high. Therefore, the present invention uses a sine curve to perform a simple fitting, and the assistance trajectory of the lower limb exoskeleton can be quickly obtained. In addition, traditional assistance strategies usually require manual adjustment of parameters to meet the needs of different scenarios or motion states. The present invention can automatically adjust the assistance curve according to different states of the human body.

[0005] Fatigue can cause a decrease in human muscle strength and endurance, thereby increasing the demand for exoskeleton robot joint assistance. In a fatigued state, the human body needs more external support to maintain normal motor function. Fatigue can change the human body's movement pattern, including joint angles, muscle activity patterns, and force output. These changes may reduce the power-assistance effect of the exoskeleton, so that the original power-assistance curve may no longer match the human body's movement pattern. Fatigue may also cause a decrease in coordination between the human body and the exoskeleton, affecting the human body's reaction speed and judgment ability, making it difficult to accurately control the movement of the exoskeleton. This will lead to an increase in the deviation between the power-assistance curve and actual demand, and may even cause the movement of the exoskeleton to conflict with the movement of the human body. In order to ensure the power-assistance effect, the power-assistance curve of the exoskeleton needs to be able to be dynamically adjusted according to the fatigue state of the human body.

[0006] In order to achieve smooth control of the joint motors of power-assisted robots, current research on lower limb exoskeleton power-assistance strategies mostly focuses on the normal state of the human body, and rarely explores the impact of muscle fatigue on the power-assistance curve. Summary of the invention

[0007] The object of the present invention is to provide a method for compliant control of a joint motor of a power-assist robot, comprising the following steps:

[0008] 1) Use IMU unit and plantar pressure sensor to obtain human motion data;

[0009] 2) Processing human motion data to predict the current gait cycle of the human body;

[0010] 3) Based on the current gait cycle of the human body, an expected power assistance curve is generated, and the maximum value of the expected power assistance curve is used as the expected power assistance base value;

[0011] 4) Calculate the fuzzy incremental entropy of the electromyographic signal;

[0012] 5) Determine the fatigue stage of the human body based on the fuzzy incremental entropy of the electromyographic signal;

[0013] 6) Based on the fatigue stage of the human body, the expected assistance base value is adjusted to generate the expected assistance value;

[0014] 7) Based on the expected power value, generate the exoskeleton robot joint motor control quantity;

[0015] 8) Based on the exoskeleton robot joint motor control quantity, the motor is driven to control the exoskeleton robot joint.

[0016] Furthermore, the human motion data includes plantar pressure information, posture angles of human joints, angular velocities and accelerations of hip joints, knee joints and ankle joints of left and right legs of the human body.

[0017] Further, in step 1), the step of obtaining human motion data includes:

[0018] 1.1) Use IMU units attached to the hip joint, knee joint and ankle joint of the left and right legs of the human body to collect the angular velocity and acceleration of each joint; use the plantar pressure sensor to collect plantar pressure information;

[0019] 1.2) Use Kalman filtering or complementary filtering method to process angular velocity and acceleration to obtain the posture angle of human joints.

[0020] Furthermore, a gait cycle refers to the difference between two consecutive time points when the hip joint extension angle of the same side of the lower limb reaches the maximum.

[0021] Further, in step 2), the step of predicting the current gait cycle of the human body includes:

[0022] 2.1) Calculate the gait period T i-j ,Right now:

[0023] T i-j =S i-j -S i-j-1 (1)

[0024] In the formula, S i-j S is the time for the ij step to reach the maximum hip extension angle; i-j-1 The time for ij-1 step to reach the maximum hip extension angle; the initial value of j is 1;

[0025] 2.2) Determine whether j≥m is true. If not, set j=j+1 and return to step 2.1). If yes, proceed to step 2.3);

[0026] 2.3) Predict the current gait cycle Ti of the human body, that is:

[0027] T i=ω1T i-1 +ω2T i-2 +...+ω n T i-n (2)

[0028] In the formula, ω j is the weight coefficient; j=1,2,…,m.

[0029] Furthermore, the expected power curve F is as follows:

[0030]

[0031] Where Fn is the maximum value of the expected power assist curve; t o is the starting time of the assist, t e is the end time of the assistance curve, and x is the time variable.

[0032] Further, in step 4), the step of calculating the fuzzy incremental entropy of the electromyographic signal includes:

[0033] 4.1) taking the i-th motion signal segment (electromyographic signal motion activity segment) as a separate time series {u1(i): 1≤i≤N1}; the i-th motion signal segment includes hip joint information;

[0034] A vector sequence in time series is represented as ,Right now:

[0035]

[0036] in, u0 is the time series of the EMG signal amplitude;

[0037] 4.2) Calculate the vector sequence in the time series With vector sequence The distance between Right now:

[0038]

[0039] 4.4) Calculate vector sequence With vector sequence Similarity Right now:

[0040]

[0041] In the formula, r = k·std(T) represents the constant tolerance value; k is a constant greater than 0; the function std(·) represents the standard deviation of the time series;

[0042] 4.5) Calculate the function φ m (r), that is:

[0043]

[0044] In the formula, for and The similarity between

[0045] 4.6) Construct the approximate entropy fApEn(m,r) expression of the time series, that is:

[0046] fApEn(m,r,N)=φ m (r)-φ m+1 (r) (9)

[0047] 4.7) Using Gaussian function as the fuzzy membership function to solve formula (9), we can get the approximate entropy f corresponding to the i-th action signal segment. i ,Right now:

[0048]

[0049] 4.8) Construct a time series X1 with n-1 zero values, that is:

[0050] X1={f1,0,…,0} (11)

[0051] 4.9) Estimate the fApEn value for the time series X1 and obtain the fuzzy incremental entropy sequence h1;

[0052] 4.10) Repeat steps 4.1)-4.9) to obtain a fuzzy incremental entropy sequence h1,h2,…,h consisting of n fuzzy incremental entropies. n .

[0053] Furthermore, the human body fatigue stage includes a mild fatigue stage, a moderate fatigue stage and an exhaustion stage;

[0054] When the fuzzy incremental entropy satisfies α≤h i When ≤β1, the human body is in the stage of mild fatigue;

[0055] When the fuzzy incremental entropy satisfies β1≤h i When ≤β2, the human body’s fatigue stage is the moderate fatigue stage;

[0056] When the fuzzy incremental entropy satisfies β2≤h i When ≤γ, the human body’s fatigue stage is the exhaustion stage.

[0057] Further, when the human body fatigue stage is a slight fatigue stage, the expected power assist value is Fn1 = Fn + 10% Fn;

[0058] When the human body is in the moderate fatigue stage, the expected power value is Fn1 = Fn + 15% Fn;

[0059] When the human body's fatigue stage is the exhaustion stage, the expected assistance value is Fn1 = Fn + 30% Fn.

[0060] Further, in step 7), the step of generating the exoskeleton robot joint motor control quantity includes:

[0061] 1) Obtain the error e between the expected power and the actual power of the exoskeleton f ;

[0062] 2) Using the admittance controller to solve the relationship equation between the error and the displacement adjustment amount, the displacement adjustment amount Δx is obtained, and the displacement adjustment amount is used as the control amount of the exoskeleton robot joint motor;

[0063] The relationship between error and displacement adjustment is shown below:

[0064]

[0065] Among them, M is the inertia parameter; B is the damping parameter; K is the stiffness parameter.

[0066] The technical effect of the present invention is unquestionable. The present invention generates an expected assistance curve for the human-machine system that conforms to the natural movement laws of the human body, ensuring that the assistance of the exoskeleton is highly consistent with the wearer's movement intention. At the same time, the electromyographic signals are collected and processed using a fuzzy incremental entropy algorithm to quantitatively evaluate the wearer's fatigue level.

[0067] According to this evaluation result, the entire fatigue process is scientifically divided into three different fatigue degree categories. This classification helps to provide accurate compensation signals for different fatigue stages during the control process of the exoskeleton robot, thereby optimizing the power-assisting effect and achieving better human-machine collaboration. Finally, based on the above-mentioned gait cycle prediction, fatigue degree assessment, and generation of the expected power-assisting curve, the present invention will send corresponding control signals to the driver according to different power-assisting requirements to drive the motor for precise exoskeleton control. This process not only ensures that the power-assisting effect of the exoskeleton robot joints matches the actual needs of the wearer, but also achieves smooth control of the power-assisting robot joints through flexible adjustment strategies, greatly improving the wearer's comfort and user experience.

[0068] The present invention obtains human motion state data collected by inertial measurement sensors and plantar pressure sensors, including acceleration, angular velocity values ​​and plantar pressure information of the hip joints, knee joints and ankle joints of the left and right legs, to identify the motion intention of the human body in the current gait cycle, including gait event recognition, gait cycle division and gait cycle prediction, and then generates the exoskeleton robot joint motor assist curve. At the same time, the human body electromyographic signal is collected, and the fatigue state of the current muscle is identified by calculating its fuzzy incremental entropy, the corresponding assist mode is selected, the original assist value is adjusted, and finally, according to different assist modes, a control signal is sent to the driver to drive the motor for exoskeleton control, so as to realize the compliant control of the robot joint motor.

[0069] In the present invention, the generation of the power-assistance curve of the exoskeleton robot joint motor adopts IMU and plantar pressure sensor to identify the gait cycle, and the power-assistance timing and power-assistance size of the power-assistance robot motor are determined by analyzing the angle change and torque change within the cycle, and the desired power-assistance curve is generated by sinusoidal function fitting. The precise power-assistance timing and size can reduce the impact and vibration of the human body during exercise, thereby reducing the risk of sports injuries. Especially in key stages such as gait conversion, appropriate power-assistance can effectively relieve the burden on joints and muscles. The invention can better solve the interaction between man and machine and improve the coordination between man and machine.

[0070] The present invention adopts a fatigue characterization algorithm based on fuzzy incremental entropy to identify the fatigue state of human muscles, which can objectively and effectively divide the fatigue stages, so as to provide corresponding compensation signals according to different fatigue stages in the exoskeleton robot motor control, so as to better achieve the purpose of compliant control and human-machine integration.

[0071] The flexible control method for the joint motor of the power-assisted robot proposed in the present invention provides an power-assisted curve for the exoskeleton motor that conforms to the human body's movement intention on the basis of generating a basic power-assisted curve and accurately identifying the human body's fatigue stage, thereby reducing the human body's movement consumption, eliminating the safety hazards caused by human-machine mismatch, improving the flexibility of the joints, and having better power-assisted effect and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a flow chart of a compliant control strategy of a joint motor of a power-assist robot according to the present invention;

[0073] Figure 2 This is a flow chart of fatigue calculation based on fuzzy incremental entropy in the present invention. DETAILED DESCRIPTION

[0074] The present invention is further described below in conjunction with the embodiments, but it should not be understood that the above subject matter of the present invention is limited to the following embodiments. Without departing from the above technical ideas of the present invention, various substitutions and changes are made according to the common technical knowledge and customary means in the art, which should all be included in the protection scope of the present invention.

[0075] Embodiment 1:

[0076] See also Figure 1 to Figure 2 , a compliant control method for a joint motor of a power-assist robot, comprising the following steps:

[0077] 1) Use IMU unit and plantar pressure sensor to obtain human motion data;

[0078] 2) Processing human motion data to predict the current gait cycle of the human body;

[0079] 3) Based on the current gait cycle of the human body, an expected power assistance curve is generated, and the maximum value of the expected power assistance curve is used as the expected power assistance base value;

[0080] 4) Calculate the fuzzy incremental entropy of the electromyographic signal;

[0081] 5) Determine the fatigue stage of the human body based on the fuzzy incremental entropy of the electromyographic signal;

[0082] 6) Based on the fatigue stage of the human body, the expected assistance base value is adjusted to generate the expected assistance value;

[0083] 7) Based on the expected power value, generate the exoskeleton robot joint motor control quantity;

[0084] 8) Based on the exoskeleton robot joint motor control quantity, the motor is driven to control the exoskeleton robot joint.

[0085] The human motion data includes plantar pressure information, posture angles of human joints, angular velocities and accelerations of the hip joints, knee joints and ankle joints of the left and right legs of the human body.

[0086] In step 1), the step of obtaining human motion data includes:

[0087] 1.1) Use IMU units attached to the hip joint, knee joint and ankle joint of the left and right legs of the human body to collect the angular velocity and acceleration of each joint; use the plantar pressure sensor to collect plantar pressure information;

[0088] 1.2) Process the angular velocity and acceleration to obtain the posture angle of the human joint;

[0089] The processing steps are: use a fusion algorithm (such as Kalman filtering, complementary filtering, etc.) to estimate the attitude angle based on the dynamic response of the angular velocity data and the static accuracy of the acceleration data, while reducing the influence of integral drift and linear acceleration interference.

[0090] A gait cycle refers to the difference between two consecutive time points when the hip joint extension angle on the same side of the lower limb reaches the maximum.

[0091] In step 2), the step of predicting the current gait cycle of the human body includes:

[0092] 2.1) Calculate the gait period T i-j ,Right now:

[0093] T i-j =S i-j -S i-j-1 (1)

[0094] In the formula, S i-j S is the time for the ij step to reach the maximum hip extension angle; i-j-1 The time for ij-1 step to reach the maximum hip extension angle; the initial value of j is 1;

[0095] 2.2) Determine whether j≥m is true. If not, set j=j+1 and return to step 2.1). If yes, proceed to step 2.3);

[0096] 2.3) Predict the current gait cycle Ti of the human body, that is:

[0097] T i =ω1T i-1 +ω2T i-2 +...+ω n T i-n (2)

[0098] In the formula, ω j is the weight coefficient; j=1,2,…,m.

[0099] The expected power curve F is as follows:

[0100]

[0101] Where Fn is the maximum value of the expected power assist curve; t o is the starting time of the assist, t e is the end time of the assistance curve, and x is the time variable.

[0102] In step 4), the step of calculating the fuzzy incremental entropy of the electromyographic signal includes:

[0103] 4.1) Treat the i-th action signal segment as a separate time series {u1(i):1≤i≤N1};

[0104] A vector sequence in time series is represented as ,Right now:

[0105]

[0106] in, u0 is the time series of the EMG signal amplitude;

[0107] 4.2) Calculate the vector sequence in the time series With vector sequence The distance between Right now:

[0108]

[0109] 4.4) Calculate vector sequence With vector sequence Similarity Right now:

[0110]

[0111] In the formula, r = k·std(T) represents the constant tolerance value; k is a constant greater than 0; the function std(·) represents the standard deviation of the time series;

[0112] 4.5) Calculate the function φ m (r), that is:

[0113]

[0114] In the formula, for and The similarity between

[0115] 4.6) Construct the approximate entropy fApEn(m,r) expression of the time series, that is:

[0116] fApEn(m,r,N)=φ m (r)-φ m+1 (r) (9)

[0117] 4.7) Using Gaussian function as the fuzzy membership function to solve formula (9), we can get the approximate entropy f corresponding to the i-th action signal segment. i ,Right now:

[0118]

[0119] 4.8) Construct a time series X1 with n-1 zero values, that is:

[0120] X1={f1,0,…,0} (11)

[0121] 4.9) Estimate the fApEn value for the time series X1 and obtain the fuzzy incremental entropy sequence h1;

[0122] 4.10) Repeat steps 4.1)-4.9) to obtain a fuzzy incremental entropy sequence h1,h2,…,h consisting of n fuzzy incremental entropies. n .

[0123] The human body fatigue stages include a mild fatigue stage, a moderate fatigue stage and an exhaustion stage;

[0124] When the fuzzy incremental entropy satisfies α≤h i When ≤β1, the human body is in the stage of mild fatigue;

[0125] When the fuzzy incremental entropy satisfies β1≤h i When ≤β2, the human body’s fatigue stage is the moderate fatigue stage;

[0126] When the fuzzy incremental entropy satisfies β2≤h i When ≤γ, the human body’s fatigue stage is the exhaustion stage.

[0127] When the human body is in the mild fatigue stage, the expected power assist value is Fn1=Fn+10%Fn;

[0128] When the human body is in the moderate fatigue stage, the expected power value is Fn1 = Fn + 15% Fn;

[0129] When the human body's fatigue stage is the exhaustion stage, the expected assistance value is Fn1 = Fn + 30% Fn.

[0130] In step 7), the step of generating the exoskeleton robot joint motor control quantity includes:

[0131] 1) Obtain the error e between the expected power and the actual power of the exoskeleton f ;

[0132] 2) Using the admittance controller to solve the relationship equation between the error and the displacement adjustment amount, the displacement adjustment amount Δx is obtained, and the displacement adjustment amount is used as the control amount of the exoskeleton robot joint motor;

[0133] The relationship between error and displacement adjustment is shown below:

[0134]

[0135] Among them, M is the inertia parameter; B is the damping parameter; K is the stiffness parameter.

[0136] Embodiment 2:

[0137] A method for compliant control of a joint motor of an assistive robot comprises the following steps:

[0138] 1) Use IMU unit and plantar pressure sensor to obtain human motion data;

[0139] 2) Processing human motion data to predict the current gait cycle of the human body;

[0140] 3) Based on the current gait cycle of the human body, an expected power assistance curve is generated, and the maximum value of the expected power assistance curve is used as the expected power assistance base value;

[0141] 4) Calculate the fuzzy incremental entropy of the electromyographic signal;

[0142] 5) Determine the fatigue stage of the human body based on the fuzzy incremental entropy of the electromyographic signal;

[0143] 6) Based on the fatigue stage of the human body, the expected assistance base value is adjusted to generate the expected assistance value;

[0144] 7) Based on the expected power value, generate the exoskeleton robot joint motor control quantity;

[0145] 8) Based on the exoskeleton robot joint motor control quantity, the motor is driven to control the exoskeleton robot joint.

[0146] Embodiment 3:

[0147] A method for compliant control of joint motors of an assistive robot, the technical content of which is the same as that of Example 2, and further, the human motion data includes plantar pressure information, posture angles of human joints, angular velocities and accelerations of the hip joints, knee joints and ankle joints of the left and right legs of the human body.

[0148] Embodiment 4:

[0149] A method for compliant control of a joint motor of an assistive robot, the technical content of which is the same as any one of Embodiments 2-3, further, in step 1), the step of obtaining human motion data comprises:

[0150] 1.1) Use IMU units attached to the hip joint, knee joint and ankle joint of the left and right legs of the human body to collect the angular velocity and acceleration of each joint; use the plantar pressure sensor to collect plantar pressure information;

[0151] 1.2) Processing the angular velocity and acceleration to obtain the posture angle of the human body joint; Example 5:

[0152] A method for compliant control of joint motors of power-assisted robots, the technical content of which is the same as any one of embodiments 2-4, and further, a gait cycle refers to the difference between two consecutive time points when the hip joint extension angles of the same side of the lower limb reach the maximum.

[0153] Embodiment 6:

[0154] A method for compliant control of a joint motor of an assistive robot, the technical content of which is the same as any one of embodiments 2-5, further, in step 2), the step of predicting the current gait cycle of a human body comprises:

[0155] 2.1) Calculate the gait period T i-j ,Right now:

[0156] T i-j =S i-j -S i-j-1 (1)

[0157] In the formula, S i-j S is the time for the ij step to reach the maximum hip extension angle; i-j-1 The time for ij-1 step to reach the maximum hip extension angle; the initial value of j is 1;

[0158] 2.2) Determine whether j≥m is true. If not, set j=j+1 and return to step 2.1). If yes, proceed to step 2.3);

[0159] 2.3) Predict the current gait cycle Ti of the human body, that is:

[0160] T i =ω1T i-1 +ω2T i-2 +...+ω n T i-n (2)

[0161] In the formula, ω j is the weight coefficient; j=1,2,…,m.

[0162] Embodiment 7:

[0163] A method for compliant control of a joint motor of an assistive robot, the technical content of which is the same as any one of Embodiments 2-6, and further, the expected assist curve is as follows:

[0164]

[0165] Where Fn is the maximum value of the expected power assist curve; t o is the starting time of the assist, t e is the end time of the assistance curve, and x is the time variable.

[0166] Embodiment 8:

[0167] A method for compliant control of a joint motor of an assistive robot, the technical content of which is the same as any one of embodiments 2-7, further, in step 4), the step of calculating the fuzzy incremental entropy of the electromyographic signal comprises:

[0168] 4.1) Treat the i-th action signal segment as a separate time series {u1(i):1≤i≤N1};

[0169] A vector sequence in time series is represented as ,Right now:

[0170]

[0171] in,

[0172] 4.2) Calculate the vector sequence in the time series With vector sequence The distance between Right now:

[0173]

[0174] 4.4) Calculate vector sequence With vector sequence Similarity Right now:

[0175]

[0176] In the formula, r = k·std(T) represents the constant tolerance value; k is a constant greater than 0; the function std(·) represents the standard deviation of the time series;

[0177] 4.5) Calculate the function φ m (r), that is:

[0178]

[0179] In the formula, for and The similarity between

[0180] 4.6) Construct the approximate entropy fApEn(m,r) expression of the time series, that is:

[0181] fApEn(m,r,N)=φ m (r)-φ m+1 (r) (9)

[0182] 4.7) Using Gaussian function as the fuzzy membership function to solve formula (9), we can get the approximate entropy f corresponding to the i-th action signal segment. i ,Right now:

[0183]

[0184] 4.8) Construct a time series X1 with n-1 zero values, that is:

[0185] X1={f1,0,…,0} (11)

[0186] 4.9) Estimate the fApEn value for the time series X1 and obtain the fuzzy incremental entropy sequence h1;

[0187] 4.10) Repeat steps 4.1)-4.9) to obtain a fuzzy incremental entropy sequence h1,h2,…,h consisting of n fuzzy incremental entropies. n .

[0188] Embodiment 9:

[0189] A method for compliant control of a joint motor of an assistive robot, the technical content of which is the same as any one of embodiments 2-8, further, the human fatigue stage includes a slight fatigue stage, a moderate fatigue stage and an exhaustion stage;

[0190] When the fuzzy incremental entropy satisfies α≤h i When ≤β1, the human body is in the stage of mild fatigue;

[0191] When the fuzzy incremental entropy satisfies β1≤h i When ≤β2, the human body’s fatigue stage is the moderate fatigue stage;

[0192] When the fuzzy incremental entropy satisfies β2≤h i When ≤γ, the human body’s fatigue stage is the exhaustion stage.

[0193] Embodiment 10:

[0194] A method for compliant control of a joint motor of an assistive robot, the technical content of which is the same as any one of embodiments 2-9, and further, when the fatigue stage of the human body is a slight fatigue stage, the expected assist value is Fn1=Fn+10%Fn;

[0195] When the human body is in the moderate fatigue stage, the expected power value is Fn1 = Fn + 15% Fn;

[0196] When the human body's fatigue stage is the exhaustion stage, the expected assistance value is Fn1 = Fn + 30% Fn.

[0197] Embodiment 11:

[0198] A method for controlling the compliant motor of a joint of an assistive robot, the technical content of which is the same as any one of embodiments 2-10, further, in step 7), the step of generating the control amount of the joint motor of an exoskeleton robot comprises:

[0199] 1) Obtain the error e between the expected power and the actual power of the exoskeleton f ;

[0200] 2) Using the admittance controller to solve the relationship equation between the error and the displacement adjustment amount, the displacement adjustment amount Δx is obtained, and the displacement adjustment amount is used as the control amount of the exoskeleton robot joint motor;

[0201] The relationship between error and displacement adjustment is shown below:

[0202]

[0203] Among them, M is the inertia parameter; B is the damping parameter; K is the stiffness parameter.

[0204] Embodiment 12:

[0205] A method for compliant control of a joint motor of an assistive robot, the steps comprising:

[0206] Step 1: Obtain human motion data through IMU and plantar pressure sensor

[0207] In this embodiment, the motion data includes angular velocity, acceleration collected by six IMUs placed at the hip joint, knee joint and ankle joint of the left and right legs, and plantar pressure information collected by plantar pressure sensors. The angular velocity and acceleration can be used to obtain the posture angle of the human joints through data processing and algorithm conversion.

[0208] Step 2: Predict the gait cycle by the time difference between the maximum hip extension angles of the left and right limbs to obtain the current gait cycle of the human body

[0209] Specifically include the following:

[0210] Gait event recognition: Thin-film pressure sensors are placed on the heel, inside of the sole, outside of the sole and under the big toe. The pressure sensors on the soles of the feet in different parts can easily tell the time when the heel touches the ground, the forefoot touches the ground, the toes touch the ground and leave the ground, and then the phase of the human gait can be divided. The phase of the human gait is divided into the pre-swing phase, the swing phase and the support phase. The pre-swing phase is defined as the time when the toes are about to leave the ground, the pressure on the inside and outside of the sole decreases before the toes, and the heel is off the ground at this time, and the pressure value is the minimum; the swing phase is defined as the pressure value of each part of the sole is very small until the pressure value of the heel increases; the support phase is when the heel begins to touch the ground and contacts the ground, so that the pressure of the heel continues to increase, and then begins to transition to the mid-support phase, the sole and toes begin to contact the ground, and the pressure begins to increase. When the pressure on the inside and outside of the sole begins to exceed the pressure on the heel, it begins to transition to the late support phase, the heel leaves the ground, the heel pressure decreases rapidly, and the pressure on the sole and toes increases until the pressure on the big toe decreases again.

[0211] Gait cycle division: The time between two toe-offs is a gait cycle. The first double support period (0% to 10%), the single support period (10% to 50%), the second double support period (50% to 60%), the initial swing period (60% to 85%), and the middle and late swing period (85% to 100%). The maximum hip extension is about 50% of the gait cycle, and the maximum hip flexion is about 85% of the gait cycle. The moment of toe-off corresponds to the moment of maximum hip extension angle. Therefore, this paper defines the gait cycle as the event segment between two consecutive maximum hip extension angles of the same side of the lower limb. The relationship between the previous gait cycle Ti-1 and the hip joint angle can be obtained by the difference between the maximum hip extension angle time of the i-1 step and the maximum hip extension angle time of the i-2 step.

[0212] The gait cycle of the previous gait can be accurately obtained by the difference in the maximum hip extension angle time of the left and right limbs, but due to the time lag problem, the current gait cycle cannot be obtained. Under different gait cycles, the parameters of the force curve applied by the assist are also different, so it is necessary to predict the current gait cycle to ensure that the assist force curve and the biological joint torque curve of the current gait can be correctly matched. This paper adopts the mean prediction method to realize the prediction of the gait cycle, and performs a weighted average of the first three gait cycles to predict the value of the current gait cycle Ti:

[0213] T i =0.5T i-1 +0.3T i-2 +0.2T i-3

[0214] Step 3: Generate the expected power curve

[0215] This patent selects an assisting strategy to provide assistance to hip flexion during the second double support period and the initial swing phase (50% to 85% of the gait cycle) and to provide assistance to knee flexion during the second double support period to 75% of the gait cycle (50% to 75% of the gait cycle).

[0216] This patent adopts a sine curve to form a power assist curve. The power assist formula shown below is used to fit the power assist curve of the knee joint in the support phase. In the formula, Fn is the maximum value of the power assist curve, to is the start time of the power assist, te is the end time of the power assist curve, and x is the time variable.

[0217]

[0218] Step 4: Calculate the fuzzy incremental entropy of the EMG signal

[0219] like Figure 2 As shown in the figure, the calculation method of the fuzzy incremental entropy of the electromyographic signal is as follows:

[0220] The representation of standard fuzzy entropy is:

[0221] For a time series containing N data points {u(i):1≤i≤N}, one of the vector sequences is represented as

[0222]

[0223] in

[0224] and The distance between It can be expressed as:

[0225]

[0226] Given an r, then and The similarity between It can be determined by the fuzzy membership function: r cannot be given due to different action situations, it can only be obtained based on experiments and is a positive number

[0227]

[0228] function Determined by the following formula:

[0229]

[0230] Through the vector sequence And the function φ m+1 (r) The time series fApEn(m,r) can be defined as:

[0231] fApEn(m,r,N)=φ m (r)-φ m+1 (r)

[0232] The result of using Gaussian function as fuzzy membership function to calculate fApEn is:

[0233]

[0234] The representation of fuzzy incremental entropy of EMG signal is as follows:

[0235] In the process from non-fatigue state to exhaustion, since this paper adopts isotonic contraction to induce fatigue, the electromyography of the whole fatigue process can be divided into n isotonic contraction action segments. The electromyogram shape of each action is roughly similar. The complete electromyographic signal can be divided by sliding windows with equal distances without overlapping the time of a single action.

[0236] The first action signal segment is regarded as a separate time series {u1(i):1≤i≤N1}, and the time series estimation fApEn is performed according to the above steps to obtain the f1 value, which is input into the time with n 0 values, which can be expressed as:

[0237] X1={f1,0,…,0}

[0238] At this time, the number of 0 values ​​is n-1. Estimate the fApEn value of the time series X1 and obtain the first value h1 of the fuzzy incremental entropy feature sequence. In the study of electromyographic signals during fatigue, since the sequence has only one non-zero value, the entropy can be approximated to 0, so this is used as the starting point of the fatigue process, and the state is set as a non-fatigue state.

[0239] Repeat this step for each signal segment until the time series X of the fApEn value of the input n-th action is generated. n , which can be expressed as:

[0240] X n ={f1,f2,…,f n}

[0241] For the time series X n Estimate the fApEn value and obtain the last value h of the fuzzy incremental entropy feature sequence n At this time, a new sequence h1,h2,…,h consisting of n fuzzy incremental entropies can be obtained. n , which is the fuzzy incremental entropy sequence of the signal.

[0242] Step 5: Output human fatigue stage

[0243] According to the obtained fuzzy incremental entropy results, through clustering, human fatigue can be characterized in three stages: mild fatigue stage (fuzzy incremental entropy α≤h k ≤β1), moderate fatigue stage (fuzzy incremental entropy β1≤h k ≤β2) and exhaustion stage (fuzzy incremental entropy β2≤h k ≤γ).

[0244] Step 6: Select a corresponding power-assisting mode according to the fatigue stage, the power-assisting modes include: a first power-assisting mode, a second power-assisting mode and a third power-assisting mode, and set the power-assisting value of the corresponding power-assisting mode according to an experimental method.

[0245] First assist mode: The human body is in a state of slight fatigue. The assist value obtained in step 3 is increased by 10%. The maximum value of the assist curve is Fn1 = Fn + 10% Fn

[0246] Second assist mode: When the fatigue state of the human body is in a moderate fatigue state, the assist value obtained in step 3 is increased by 20%, and the maximum value of the assist curve is Fn1 = Fn + 15% Fn

[0247] The third assist mode: The human body is in a state of exhaustion. The assist value obtained in step 3 is increased by 30%. The maximum value of the assistant curve is Fn1 = Fn + 30% Fn

[0248] Step 7: Use admittance control to realize power-assisted control of the exoskeleton robot joint motors

[0249] Admittance control is a control method for force input and position output. Assume that in the lower limb exoskeleton robot power assistance system,

[0250] The displacement that needs to be adjusted is Δx, and the difference between the expected power and the actual power of the exoskeleton is e f , their relationship is:

[0251]

[0252] Among them, M is the inertia parameter; B is the damping parameter; K is the stiffness parameter.

[0253] In this way, the displacement adjustment amount Δx can be obtained by passing the difference ef between the expected assistance and the actual assistance through the admittance controller, and it is used as the motor control amount.

[0254] Step 8: Send control signals to the driver motor to control the joints of the exoskeleton robot according to different power-assist modes.

Claims

1. A method for controlling the compliance of a joint motor of an assistive robot, characterized in that: The following steps are involved: 1) Use IMU unit and plantar pressure sensor to obtain human motion data; 2) Processing human motion data to predict the current gait cycle of the human body; 3) Based on the current gait cycle of the human body, an expected power assistance curve is generated, and the maximum value of the expected power assistance curve is used as the expected power assistance base value; 4) Calculate the fuzzy incremental entropy of the electromyographic signal; 5) Determine the fatigue stage of the human body based on the fuzzy incremental entropy of the electromyographic signal; 6) Based on the fatigue stage of the human body, the expected assistance base value is adjusted to generate the expected assistance value; 7) Based on the expected power value, generate the exoskeleton robot joint motor control quantity; 8) Based on the exoskeleton robot joint motor control quantity, the motor is driven to control the exoskeleton robot joint.

2. A compliant control method for joint motors of power-assist robots according to claim 1, characterized in that: The human body motion data includes plantar pressure information, posture angles of human body joints, angular velocities and accelerations of hip joints of left and right legs of the human body.

3. A compliant control method for joint motors of power-assist robots according to claim 2, characterized in that: In step 1), the step of obtaining human motion data includes: 1.1) Use IMU units attached to the hip joints of the left and right legs of the human body to collect the angular velocity and acceleration of each joint; use the plantar pressure sensor to collect plantar pressure information; 1.2) Use Kalman filtering or complementary filtering method to process angular velocity and acceleration to obtain the posture angle of human joints.

4. The method for controlling the compliant motor of a power-assist robot joint according to claim 1, characterized in that: A gait cycle refers to the difference between two consecutive time points when the hip joint extension angle on the same side of the lower limb reaches the maximum.

5. The method for controlling the compliant motor of a power-assist robot joint according to claim 1, characterized in that: In step 2), the step of predicting the current gait cycle of the human body includes: 2.1) Calculate the gait period T i-j ,Right now: T i-j =S i-j -S i-j-1 (1) In the formula, S i-j S is the time for the ij step to reach the maximum hip extension angle; i-j-1 The time for ij-1 step to reach the maximum hip extension angle; the initial value of j is 1; 2.2) Determine whether j≥m is true. If not, set j=j+1 and return to step 2.1). If yes, proceed to step 2.3); 2.3) Predict the current gait cycle Ti of the human body, that is: T i =ω1T i-1 +ω2T i-2 +...+oh n T i-n (2) In the formula, ω j is the weight coefficient; j=1,2,…,m.

6. The method for controlling the compliant motor of a power-assist robot joint according to claim 1, characterized in that: The expected power curve F is as follows: Where Fn is the maximum value of the expected power assist curve; t o is the assist start time, t e is the end time of the assistance curve, and x is the time variable.

7. The method for controlling the compliant motor of a power-assist robot joint according to claim 1, characterized in that: In step 4), the step of calculating the fuzzy incremental entropy of the electromyographic signal includes: 4.1) taking the i-th motion signal segment as a separate time series {u1(i):1≤i≤N1}; the i-th motion signal segment includes hip joint information; A vector sequence in time series is represented as Right now: in, u0 is the time series of the EMG signal amplitude; 4.2) Calculate the vector sequence in the time series With vector sequence The distance between Right now: 4.4) Calculate vector sequence With vector sequence Similarity Right now: In the formula, r = k·std(T) represents the constant tolerance value; k is a constant greater than 0; the function std(·) represents the standard deviation of the time series; 4.5) Calculate the function φ m (r), that is: In the formula, for and The similarity between 4.6) Construct the approximate entropy fApEn(m,r) expression of the time series, that is: fApEn(m,r,N)=φ m (r)-φ m+1 (r)(9) 4.7) Using Gaussian function as the fuzzy membership function to solve formula (9), we can get the approximate entropy f corresponding to the i-th action signal segment. i ,Right now: 4.8) Construct a time series X1 with n-1 zero values, that is: X1={f1,0,···,0}(11) 4.9) Estimate the fApEn value for the time series X1 and obtain the fuzzy incremental entropy sequence h1; 4.10) Repeat steps 4.1)-4.9) to obtain a fuzzy incremental entropy sequence h1,h2,···,h consisting of n fuzzy incremental entropies n .

8. The method for controlling the compliant motor of a power-assist robot joint according to claim 1, characterized in that: The human body fatigue stages include a mild fatigue stage, a moderate fatigue stage and an exhaustion stage; When the fuzzy incremental entropy satisfies α≤h i When ≤β1, the human body is in the stage of mild fatigue; When the fuzzy incremental entropy satisfies β1≤h i When ≤β2, the human body’s fatigue stage is the moderate fatigue stage; When the fuzzy incremental entropy satisfies β2≤h i When ≤γ, the human body’s fatigue stage is the exhaustion stage.

9. The method for controlling the compliance of a joint motor of an assistive robot according to claim 1, characterized in that: When the human body is in the mild fatigue stage, the expected power assist value is Fn1=Fn+10%Fn; When the human body is in the moderate fatigue stage, the expected power value is Fn1 = Fn + 15% Fn; When the human body's fatigue stage is the exhaustion stage, the expected assistance value is Fn1 = Fn + 30% Fn.

10. The method for controlling the compliance of a joint motor of an assistive robot according to claim 1, characterized in that: In step 7), the step of generating the exoskeleton robot joint motor control quantity includes: 1) Obtain the error ef between the expected power assistance and the actual power assistance of the exoskeleton; 2) Using the admittance controller to solve the relationship equation between the error and the displacement adjustment amount, the displacement adjustment amount Δx is obtained, and the displacement adjustment amount is used as the control amount of the exoskeleton robot joint motor; The relationship between error and displacement adjustment is shown below: Among them, M is the inertia parameter; B is the damping parameter; K is the stiffness parameter.

Images