CN113478462A - Method and system for controlling intention assimilation of upper limb exoskeleton robot based on surface electromyogram signal - Google Patents

Abstract

The present invention provides an intention assimilation control method and system for an upper limb exoskeleton robot based on surface EMG signals, including: step 1: using the Kane method to establish a dynamic model of the upper limb exoskeleton robot; step 2: based on the dynamic model, through The surface EMG signal is used for intention recognition; Step 3: Intention assimilation control is carried out through the virtual target. The present invention proposes an intent assimilation control method, which covers continuous interaction behaviors from cooperation to competition, with less force guidance, safer obstacle avoidance and wider interaction behaviors.

Description

Intention assimilation control method and system for upper limb exoskeleton robot based on surface electromyographic signals

technical field

The invention relates to the technical fields of human-computer interaction, artificial intelligence and interactive control, and in particular, to a method and a system for controlling the intention assimilation of an upper limb exoskeleton robot based on surface electromyography signals.

Background technique

Robot technology has developed rapidly in recent years, especially human-computer interaction robots. Human-computer interface is the most important part of human-computer interaction research. The quality of human-computer interface signals directly affects the control effect and experimental results. Compared with the human-machine interface of the motion intention signal, the surface EMG signal has great advantages in terms of accuracy and delay, and is more accurate in the estimation of human motion and force.

In terms of control strategies, the diversification of interactive robot control strategies is an important factor for its promotion and application. The basic control strategy is PID control, which is simple and convenient to apply, but it can only be controlled according to a fixed trajectory and cannot introduce human intentions; in order to reflect Human body intentions and surface EMG signals are also introduced into the robot control strategy. At the same time, artificial intelligence algorithms are used to connect EMG signals with human joints, and a certain control effect has been achieved. Starting from the concept of master-slave role homotopy switching, the Human-computer interaction behaviors are divided into: assistance, cooperation, coordination and confrontation, etc. Intent assimilation control covers continuous interaction behaviors from cooperation to competition, less force guidance, safer obstacle avoidance and wider interaction behaviors.

The patent document CN108283569A (application number: CN201711449077.7) discloses an exoskeleton robot control system and control method, so as to solve the problem that the existing rehabilitation exoskeleton robot has poor versatility and cannot correctly judge the motion intention of the human body, and cannot realize human-machine collaboration function effect. Exoskeleton robot control system, including attitude sensor, angle sensor, pressure sensor, surface electromyographic signal sensor, processor, exoskeleton robot wearable parts and human-computer interaction module.

SUMMARY OF THE INVENTION

In view of the defects in the prior art, the purpose of the present invention is to provide a method and system for the intention assimilation control of an upper limb exoskeleton robot based on surface EMG signals.

The intention assimilation control method for an upper limb exoskeleton robot based on surface EMG signals provided according to the present invention includes:

Step 1: Use the Kane method to establish a dynamic model of the upper limb exoskeleton robot;

Step 2: Based on the dynamic model, the intent is identified by the surface EMG signal;

Step 3: Intention assimilation control is carried out through the virtual target.

Preferably, the step 1 includes:

Step 1.1: There is no relative motion between the robot, the human and the object, and the robot and the human manipulate the object together, and the object satisfies the dynamic equation:

为物体的位置坐标对时间的二阶导数，f和u_h是机器人和人作用在物体上的力，M_o为物体的质量矩阵，G_o为物体的重力；in,

is the second derivative of the position coordinate of the object to time, f and u _h are the forces acting on the object by the robot and the human, M _o is the mass matrix of the object, and G _o is the gravity of the object;

Step 1.2: Use the Kane method to establish the dynamic model of the upper limb exoskeleton robot, and obtain the joint space dynamic equation when the n-degree-of-freedom upper limb exoskeleton robot is in contact with the environment:

是科里奥利和离心扭矩，G_q(q)是重力矩；Among them, q is the joint coordinates of the robot, τ _q is the control input, J ^T (q) is the Jacobian matrix, M _q (q) is the robot inertia matrix,

are the Coriolis and centrifugal torques, and G _q (q) is the gravitational moment;

Transform to the robot operation space to get the dynamic equation:

where u represents the control input of the upper limb exoskeleton robot,

表示矩阵的伪逆；M _r , C _r , G _r represent the inertia matrix, Coriolis force and centrifugal force matrix and gravity matrix of the upper limb exoskeleton robot in the Cartesian space coordinate system, respectively, the symbol

represents the pseudo-inverse of a matrix;

Step 1.3: Simultaneously combine formulas (1) and (3) to obtain the combined dynamic equation of the object and robot:

M≡M _o +M _r ,G≡G _o +G _r ,C≡C _r …………(5)

M, C, and G represent the inertial matrix, Coriolis force, centrifugal force matrix and gravity matrix of the upper-limb exoskeleton robot and human interaction system in the Cartesian space coordinate system, respectively;

Step 1.4: Measure the position, velocity and human force of the upper limb exoskeleton robot end, using a robot controller with gravity compensation and linear feedback, the expression is:

where τ is the target position _of the robot, and L1 and L2 are the gains corresponding to the position error and velocity _;

Model the force a person acts on an object as:

Among them, L _h,1 and L _h,2 are the control gains of the human, τ _h is the target position of the human, and the equations (6) and (7) are brought into the equation (5) to obtain the upper-limb exoskeleton robot-human interaction closed-loop system dynamics equation:

Preferably, the step 2 includes:

Step 2.1: Collect the electromyographic signals of the human wrist, forearm and elbow by means of electromyography;

Step 2.2: Perform filtering, data segmentation and feature extraction on the collected EMG signals, and perform feature extraction according to the waveform type, so that the extracted features correspond to different intent categories;

Step 2.3: Use the multi-criteria linear programming in the database combined with the online random forest classification method for training prediction;

Step 2.4: When the model predicts, compare the predicted category of each base classifier with the corresponding confidence level and the preset threshold to decide whether the base classifier will vote. Finally, use the Boost algorithm to collect the voting results of all base classifiers and weight them. and, find the predicted category with the most votes, and output the activity intent when the votes are greater than the mean.

Preferably, the step 3 includes:

评估人类对上肢外骨骼机器人与人交互系统动力学的影响，公式为：Step 3.1: Through the human virtual target

To evaluate the impact of humans on the dynamics of the upper-limb exoskeleton robot-human interaction system, the formula is:

和

使用测量平均值，或者与机器人控制器增益相同的值，即

上标符号v表示估计值；Among them, the human control gain

and

Use the measured average, or the same value as the robot controller gain, i.e.

The superscript symbol v represents the estimated value;

进行估计，表达式为：Step 3.2: Use a surface EMG-based intent recognition method, or parameterize the pair via an internal model.

To estimate, the expression is:

的参数向量，

t表示时间，m为预先设定的参数，因此

为由内模型参数决定并随时间变化的量；Among them, the superscript symbol T represents the transposition, and θ is the virtual target position of the calculation person

The parameter vector of ,

t represents time, m is a preset parameter, so

is the quantity determined by the internal model parameters and varies with time;

将其带入式(5)后得到扩展模型：State Vectors of Robot-Human Interaction Systems Using Upper Limb Exoskeletons

After taking it into formula (5), the extended model is obtained:

Among them, φ represents: the state vector of the upper limb exoskeleton robot and human interaction system, v∈N(0, E[v, v ^T ]) is the system noise, that is, the mean value is 0, and the variance is E[v, v ^T ] Gaussian noise;

Step 3.3: Measure the position and speed of the robot's endpoints and the interaction force with the human through the sensor, and obtain the measurement vector of the upper limb exoskeleton robot-human interaction system:

Among them, μ∈N(0, E[μ, μ ^T ]) is the environmental measurement noise, that is, the Gaussian noise with mean value 0 and variance E[μ, μ ^T ];

Step 3.4: Compute the extended state estimate of the robot using the system observer:

Among them, ∧ represents the estimated value; z represents the measurement vector of the upper limb exoskeleton robot and human interaction system;

The linear quadratic estimation gain K=PH ^T R ^-1 , where P is a positive definite matrix, is obtained by solving the Riccati differential equation:

Among them, the noise covariance matrix Q≡E[v,v ^T ],R≡E[μ,μ ^T ], A represents the system matrix, substituting into equation (11), expressed as the following form:

Preferably, the interaction between the human and the robot is determined by the relationship τ and τ _h between the human and the robot:

When τ=τ _h , it means that the robot uses the assistance of the human virtual target, and the robot follows its original target τ _r ;

When τ = 2τ _r -τ _h , the robot imposes its own goal by eliminating the human goal from the upper limb exoskeleton robot-human interaction system;

Use the following formula to design the target position of the robot based on the estimated human target for interactive behavior assimilation:

τ _r represents the original target position of the upper limb exoskeleton robot; λ represents the hyperparameter for adjusting the original target position of the upper limb exoskeleton robot and the human target position, which is dynamically adjusted according to the end position x.

The intention assimilation control system for an upper limb exoskeleton robot based on surface EMG signals provided according to the present invention includes:

Module M1: Use the Kane method to establish a dynamic model of an upper limb exoskeleton robot;

Module M2: Intention recognition through surface EMG signals based on dynamic model;

Module M3: Control of intention assimilation through virtual targets.

Preferably, the module M1 includes:

Module M1.1: There is no relative motion between the robot, the human and the object, and the robot and the human manipulate the object together, and the object satisfies the dynamic equation:

为物体的位置坐标对时间的二阶导数，f和u_h是机器人和人作用在物体上的力，M_o为物体的质量矩阵，G_o为物体的重力；in,

is the second derivative of the position coordinate of the object to time, f and u _h are the forces acting on the object by the robot and the human, M _o is the mass matrix of the object, and G _o is the gravity of the object;

Module M1.2: Use the Kane method to establish the dynamic model of the upper limb exoskeleton robot, and obtain the joint space dynamic equation when the n-degree-of-freedom upper limb exoskeleton robot is in contact with the environment:

是科里奥利和离心扭矩，G_q(q)是重力矩；Among them, q is the joint coordinates of the robot, τ _q is the control input, J ^T (q) is the Jacobian matrix, M _q (q) is the robot inertia matrix,

are the Coriolis and centrifugal torques, and G _q (q) is the gravitational moment;

Transform to the robot operation space to get the dynamic equation:

where u represents the control input of the upper limb exoskeleton robot,

表示矩阵的伪逆；M _r , C _r , G _r represent the inertia matrix, Coriolis force and centrifugal force matrix and gravity matrix of the upper limb exoskeleton robot in the Cartesian space coordinate system, respectively, the symbol

represents the pseudo-inverse of a matrix;

Module M1.3: Simultaneous formulas (1) and (3) to obtain the combined dynamic equation of the object and robot:

M≡M _o +M _r ,G≡G _o +G _r ,C≡C _r …………(5)

M, C, and G represent the inertial matrix, Coriolis force, centrifugal force matrix and gravity matrix of the upper-limb exoskeleton robot and human interaction system in the Cartesian space coordinate system, respectively;

Module M1.4: Measure the position, velocity and human force at the end of the upper limb exoskeleton robot, using a robot controller with gravity compensation and linear feedback, the expression is:

where τ is the target position _of the robot, and L1 and L2 are the gains corresponding to the position error and velocity _;

Model the force a person acts on an object as:

Among them, L _h,1 and L _h,2 are the control gains of the human, τ _h is the target position of the human, and the equations (6) and (7) are brought into the equation (5) to obtain the upper-limb exoskeleton robot-human interaction closed-loop system dynamics equation:

Preferably, the module M2 includes:

Module M2.1: collect electromyographic signals from human wrist, forearm and elbow through electromyography;

Module M2.2: Perform filtering, data segmentation and feature extraction on the collected EMG signals, and perform feature extraction according to the waveform type, so that the extracted features correspond to different intent categories;

Module M2.3: Use the multi-criteria linear programming in the database combined with the online random forest classification method for training prediction;

Module M2.4: During model prediction, compare the predicted category of each base classifier with the corresponding confidence level and a preset threshold to decide whether the base classifier will vote, and finally use the Boost algorithm to collect the voting results of all base classifiers. Weighted summation to find the predicted category with the most votes, and output the activity intent when the number of votes is greater than the mean.

Preferably, the module M3 includes:

评估人类对上肢外骨骼机器人与人交互系统动力学的影响，公式为：Module M3.1: Virtual Goals Through Humans

To evaluate the impact of humans on the dynamics of the upper-limb exoskeleton robot-human interaction system, the formula is:

和

使用测量平均值，或者与机器人控制器增益相同的值，即

上标符号v表示估计值；Among them, the human control gain

and

Use the measured average, or the same value as the robot controller gain, i.e.

The superscript symbol v represents the estimated value;

进行估计，表达式为：Module M3.2: Use surface EMG-based intent recognition methods, or parameterize the

To estimate, the expression is:

的参数向量，

t表示时间，m为预先设定的参数，因此

为由内模型参数决定并随时间变化的量；Among them, the superscript symbol T represents the transposition, and θ is the virtual target position of the calculation person

The parameter vector of ,

t represents time, m is a preset parameter, so

is the quantity determined by the internal model parameters and varies with time;

将其带入式(5)后得到扩展模型：State Vectors of Robot-Human Interaction Systems Using Upper Limb Exoskeletons

After taking it into formula (5), the extended model is obtained:

Among them, φ represents: the state vector of the upper limb exoskeleton robot and human interaction system, v∈N(0, E[v, v ^T ]) is the system noise, that is, the mean value is 0, and the variance is E[v, v ^T ] Gaussian noise;

Module M3.3: The sensor measures the position and velocity of the robot's endpoints and the interaction force with the human, and obtains the measurement vector of the upper limb exoskeleton robot-human interaction system:

Among them, μ∈N(0, E[μ, μ ^T ]) is the environmental measurement noise, that is, the Gaussian noise with mean value 0 and variance E[μ, μ ^T ];

Module M3.4: Compute Extended State Estimates for Robots Using System Observers:

Among them, ∧ represents the estimated value; z represents the measurement vector of the upper limb exoskeleton robot and human interaction system;

The linear quadratic estimation gain K=PH ^T R ^-1 , where P is a positive definite matrix, is obtained by solving the Riccati differential equation:

Among them, the noise covariance matrix Q≡E[v,v ^T ],R≡E[μ,μ ^T ], A represents the system matrix, substituting into equation (11), expressed as the following form:

Preferably, the interaction between the human and the robot is determined by the relationship τ and τ _h between the human and the robot:

When τ=τ _h , it means that the robot uses the assistance of the human virtual target, and the robot follows its original target τ _r ;

When τ = 2τ _r -τ _h , the robot imposes its own goal by eliminating the human goal from the upper limb exoskeleton robot-human interaction system;

Use the following formula to design the target position of the robot based on the estimated human target for interactive behavior assimilation:

τ _r represents the original target position of the upper limb exoskeleton robot; λ represents the hyperparameter for adjusting the original target position of the upper limb exoskeleton robot and the human target position, which is dynamically adjusted according to the end position x.

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

(1) The present invention introduces the surface EMG signal into the robot control strategy, and has advantages in terms of accuracy and time delay;

(2) The present invention proposes an intention assimilation control method, which covers continuous interactive behaviors from cooperation to competition, with less force guidance, safer obstacle avoidance and wider interactive behaviors;

(3) The present invention is simple and easy to implement, and is a compliance control method with high robustness.

Description of drawings

Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments with reference to the following drawings:

1 is a schematic diagram of an intention assimilation control method for an upper limb exoskeleton robot based on surface EMG signals of the present invention;

2 is a schematic diagram of obstacle avoidance and auxiliary task scenarios of the present invention;

FIG. 3 is a schematic flowchart of an intention identification method based on surface EMG signals of the present invention;

4 is a schematic diagram of the MCLP Boost algorithm of the present invention;

FIG. 5 is a schematic diagram of the change of the human-computer interaction strategy corresponding to the parameter λ adjustment of the present invention.

Detailed ways

The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the inventive concept. These all belong to the protection scope of the present invention.

Example:

Fig. 1 is a schematic diagram of an intention assimilation control method of an upper limb exoskeleton robot based on surface EMG signals according to the present invention, including the upper limb exoskeleton robot dynamics model established by the Kane method, and the intention recognition based on surface EMG signals. Method and intent assimilation control method, different task scenarios are shown in Figure 2, the intent assimilation control method of the present invention can unify different human-computer interaction strategies and perform continuous control;

Further, the specific process of using the Kane method to establish the dynamics model of the upper limb exoskeleton robot is as follows:

1) It is assumed that there is no relative motion between the robot gripper, the human hand and the object, and the robot gripper and the human hand manipulate a rigid object together, and the object is a mass point. The general object operation only considers linear motion, and the object satisfies the dynamic equation:

Among them, x(t) is the position coordinate of the object, f and u _h are the forces acting on the object by the robot and human, M _o is the mass matrix of the object, and G _o is the gravity of the object.

2) Using the dynamic model of the upper limb exoskeleton robot established by the Kane method, the joint space dynamic equation of the n-degree-of-freedom upper limb exoskeleton robot in contact with the environment is obtained:

是科里奥利和离心扭矩，G_q(q)是重力矩；Among them, q is the joint coordinates of the robot, τ _q is the control input, J ^T (q) is the Jacobian matrix, M _q (q) is the robot inertia matrix,

are the Coriolis and centrifugal torques, and G _q (q) is the gravitational moment;

Transform to the robot operation space to get the dynamic equation:

where u represents the control input of the upper limb exoskeleton robot,

表示矩阵的伪逆；The meanings of M _r , _Cr , and Gr are respectively the inertia matrix, Coriolis force, centrifugal force matrix and gravity matrix of the upper limb exoskeleton robot in the Cartesian space coordinate system _; the symbol

represents the pseudo-inverse of a matrix;

3) Simultaneous formulas (1) and (3) can obtain the combined dynamic equation of the object and the robot:

M≡M _o +M _r ,G≡G _o +G _r ,C≡C _r …………(5)

The meanings of M, C, and G are the inertial matrix, Coriolis force and centrifugal force matrix and gravity matrix of the upper limb exoskeleton robot and human interaction system in the Cartesian space coordinate system respectively;

4) Consider that the robot has information about its local environment, and measure the position, velocity and human force at the end of the upper limb exoskeleton robot interaction system, which are all affected by measurement noise. Using a robot controller with gravity compensation and linear feedback:

where τ is the target position _of the robot, and L1 and _L2 are the gains corresponding to the position error and velocity.

Model the force of a human hand on an object as:

Among them, L _h,1 and L _h,2 are the control gains of the human, and τ _h is the target position of the human. Bringing equations (6) and (7) into equation (5) can obtain a closed-loop interaction between the upper limb exoskeleton robot and the human System Dynamics Equation:

Further, the intent recognition method based on the surface EMG signal is shown in Figure 3, and the specific process is as follows:

1) The electromyography signal is collected from the human wrist, forearm and elbow by electromyography;

2) Filter the collected EMG signals, and then perform data segmentation and feature extraction. When long-sequence waveforms are used for feature extraction, overlap is not applicable, and when the waveform is short, the overlap operation can be considered, so that the extracted features can be extracted. Corresponding to different intent categories;

3) Use MCLPBoost combined with the classification method of Online Random Forest for training prediction. The MCLPBoost algorithm is shown in Figure 4. This method has good generalization performance, and it is thought that the method is based on comparison. Therefore, when making predictions, it is better to use the computational model It has the characteristics and advantages of small time overhead;

4) When the model predicts, compare the predicted category of each base classifier with the corresponding confidence (probability) and the preset threshold to decide whether the base classifier will vote. Finally, only the Boost algorithm is used to collect the voting results of all base classifiers. Weighted summation to find the prediction category with the most votes, and output the activity intent when the votes are greater than the mean;

5) Intent recognition results based on surface EMG signals can be used to generate "virtual" targets.

意图同化控制方法具体过程为：Further, the human influence on the dynamics of the upper-limb exoskeleton robot-human interaction system depends entirely on u _h , no matter what internal model it is based on, to develop an alternative method that does not require estimates of human control gains, by using arbitrary assumptions of these gains "virtual" target

The specific process of the intent assimilation control method is as follows:

可有效地评估人类对上肢外骨骼机器人与人交互系统动力学的影响，如果它满足：1) "Virtual" goals

The impact of humans on the dynamics of the upper-limb exoskeleton robot-human interaction system can be effectively assessed if it satisfies:

和

可以使用一些从许多人测量的平均值，或者与机器人控制器增益相同的值，即

Among them, the virtual human control gain

and

You can use some average measured from many people, or the same value as the robot controller gain, i.e.

可使用权利要求3所述的基于表面肌电信号的意图识别方法，或者使用内部模型参数化它：2) To estimate

The surface EMG-based intent recognition method of claim 3 can be used, or it can be parameterized using an internal model:

的参数向量，

t表示时间，m为预先设定的参数，因此

为由内模型参数决定并随时间变化的量，使用状态向量

将其带入等式(5)后可得到扩展模型：Among them, θ means the virtual target position of the calculation person

The parameter vector of ,

t represents time, m is a preset parameter, so

For quantities that are determined by internal model parameters and vary over time, use the state vector

After plugging it into equation (5), the extended model can be obtained:

Among them, φ represents: the state vector of the upper limb exoskeleton robot-human interaction system; v∈N(0, E[v, v ^T ]) is the system noise, that is, the mean value is 0, and the variance is E[v, v ^T ] Gaussian noise.

3) Considering that the robot can measure its endpoint position and velocity and the interaction force with the human with suitable sensors, the measurement vector of the robot is obtained:

Among them, μ∈N(0, E[μ, μ ^T ]) is the environmental measurement noise, that is, the Gaussian noise with mean 0 and variance E[μ, μ ^T ].

与θ是未知的，因此使用以下系统观测器来计算机器人的扩展状态估计：4) However, in equation (10)

and θ are unknown, so the following system observer is used to compute the extended state estimate of the robot:

Among them, ∧ represents the estimated value, z represents: the measurement vector of the upper limb exoskeleton robot and the human interaction system; the linear quadratic estimation gain K=PH ^T R ^-1 , P is a positive definite matrix, obtained by solving the Riccati differential equation:

Among them, the noise covariance matrix Q≡E[v,v ^T ],R≡E[μ,μ ^T ], using A to represent the system matrix, equation (11) can be expressed as the following form:

值。All parameters except θ can be observed and obtained, thus obtaining

value.

5) The interaction between humans and robots can be determined by the relationship τ and τ _h between humans and robots, such as when τ = τ _h corresponds to the robot using the assistance of a human virtual target, and when τ = τ _r the robot follows it The original goal τ _r , when τ = 2τ _r −τ _h corresponds to “adversarial”, that is, the robot imposes its own goal by eliminating the human goal from the upper limb exoskeleton robot-human interaction system.

To assimilate the interaction behavior, the robot's target position is designed according to the estimated human target using the following equation:

τ _r represents: the original target position of the upper limb exoskeleton robot; λ represents: the hyperparameters for adjusting the original target position of the upper limb exoskeleton robot and the human target position, which can be dynamically adjusted according to the end position x;

从而完成人机协同；当λ＝2时，意图同化控制器将消除模拟人类对上肢外骨骼机器人与人交互系统动态的影响，上肢外骨骼机器人与人交互系统位置最终趋同到意图同化控制器的目标τ_r。The adjustment of parameter λ corresponds to the change of human-computer interaction strategy as shown in Figure 5. When λ<1, the intent assimilation controller will coordinate the human-machine goal; when λ=1, the intent assimilation controller will ignore

In this way, the human-machine collaboration is completed; when λ=2, the intention assimilation controller will eliminate the influence of the simulated human on the dynamics of the upper limb exoskeleton robot and the human interaction system, and the positions of the upper limb exoskeleton robot and the human interaction system will eventually converge to the intention assimilation controller. target τ _r .

的引入后人机交互系统的稳定性，通过等式(13)的第二个等式可估计人的目标

Further, verify

The stability of the human-computer interaction system after the introduction of , the human goal can be estimated by the second equation of Eq.

Substituting the corrected equation (6) into the combined kinetic equation (5) yields:

表示估计值与实际值之间的误差，若定义

并将人手作用于物体的力(7)代入上面等式可获得：in,

represents the error between the estimated value and the actual value, if defined

Substitute the force (7) of the human hand on the object into the above equation to obtain:

The effect of τ _r and τ _h on the dynamical system can therefore be analyzed, taking into account the steady state position:

Equation (19) is simplified for stability analysis by defining:

Deduced:

This means that the position error xx _ss will disappear, if the estimated error of manpower is

According to the form (15) and the system observer (13) of the upper limb exoskeleton robot-human interaction system dynamics in the state space, there are:

By defining ξ≡[xx _ss ,x,φ ^T ] ^T , combining equations (22) and (23) we get:

where ξ is the system state vector defined in the system transient performance analysis, and the equation is a combined system including system dynamics and observers.

的特征值，进而研究式(24)的稳定性：It can be calculated by the solution of the following characteristic equation

, and then study the stability of formula (24):

[yI-(A-KH)][My ² +(C+L ₂ )y+L ₁ ]=0…………(25)

也将是稳定的：If the following two systems are stable, then

will also be stable:

The Lyapunov theory is used to test the stability of the above two systems respectively. First, the stability of the first system is proved by considering the Lyapunov candidate function:

Derivation with respect to time gives:

The stability of the second system is then proved by considering Lyapunov candidate functions:

P _v is given by the Riccati equation in equation (14):

Derivation with respect to time gives:

Combining equation (13), we get:

Substitute into (31) to get:

This proves that the two systems in equation (26) are stable, so the upper limb exoskeleton robot and human interaction system are stable.

Those skilled in the art know that, in addition to implementing the system, device and each module provided by the present invention in the form of pure computer readable program code, the system, device and each module provided by the present invention can be completely implemented by logically programming the method steps. The same program is implemented in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, and embedded microcontrollers, among others. Therefore, the system, device and each module provided by the present invention can be regarded as a kind of hardware component, and the modules used for realizing various programs included in it can also be regarded as the structure in the hardware component; A module for realizing various functions can be regarded as either a software program for realizing a method or a structure within a hardware component.

Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above-mentioned specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essential content of the present invention. The embodiments of the present application and features in the embodiments may be combined with each other arbitrarily, provided that there is no conflict.

Claims (10)

1. An intention assimilation control method of an upper limb exoskeleton robot based on a surface electromyogram signal is characterized by comprising the following steps:

step 1: establishing an upper limb exoskeleton robot dynamic model by using a Kenn method;

step 2: performing intention recognition through a surface electromyogram signal based on a dynamic model;

and step 3: the intention assimilation control is performed by the virtual object.

2. The method for controlling the intention assimilation of an upper limb exoskeleton robot based on surface electromyography according to claim 1, wherein the step 1 comprises:

step 1.1: there is no relative motion between the robot, the person and the object, and the robot and the person together manipulate the object, the object satisfying the dynamic equation:

wherein,

as the second derivative of the position coordinates of the object with respect to time, f and u_hIs the force of the robot and person on the object, M_oIs a mass matrix of the object, G_oIs the weight of the object;

step 1.2: establishing an upper limb exoskeleton robot dynamics model by using a Kenn method to obtain a joint space dynamics equation when the upper limb exoskeleton robot with n degrees of freedom is in contact with the environment:

wherein q is the joint coordinate of the robot, tau_qFor control input, J^T(q) is the Jacobian matrix, M_q(q) is the robot inertia matrix,

is the Coriolis and centrifugal torque, G_q(q) is the moment of gravity;

and converting into a robot operating space to obtain a kinetic equation:

wherein u represents the control input of the upper limb exoskeleton robot,

M_r、C_r、G_rrespectively representing an inertia matrix, a Coriolis force and centrifugal force matrix and a gravity matrix of the upper limb exoskeleton robot under a Cartesian space coordinate system, and symbols

Representing a pseudo-inverse of the matrix;

step 1.3: simultaneous equations (1) and (3) are obtained to obtain the combined kinetic equation of the object and the robot:

M≡M_o+M_r，G≡G_o+G_r，C≡C_r…………(5)

m, C, G respectively representing inertia matrix, Coriolis force and centrifugal force matrix and gravity matrix of the upper limb exoskeleton robot and human interaction system in a Cartesian space coordinate system;

step 1.4: the position, the speed and the human force of the tail end of the upper limb exoskeleton robot are measured, a robot controller with gravity compensation and linear feedback is adopted, and the expression is as follows:

where τ is the target position of the robot, L₁And L₂Is a gain corresponding to position error and velocity;

the force of a person acting on an object is modeled as:

wherein L is_h，1And L_h，2Control gain, τ, for humans_hAnd (5) bringing the formulas (6) and (7) into the formula (5) to obtain the dynamic equation of the upper limb exoskeleton robot and human interactive closed-loop system for the target position of the human:

3. the method for controlling the intention assimilation of an upper limb exoskeleton robot based on surface electromyography according to claim 1, wherein the step 2 comprises:

step 2.1: collecting electromyographic signals of wrists, forearms and elbows of a person through an electromyograph;

step 2.2: filtering, data segmentation and feature extraction are carried out on the collected electromyographic signals, and feature extraction is carried out according to waveform types, so that the extracted features correspond to different intention categories;

step 2.3: training and predicting by using a multi-criterion linear programming in a database and combining a classification method of an online random forest;

step 2.4: during model prediction, the prediction category of each base classifier is compared with the corresponding confidence coefficient and a preset threshold value to determine whether the base classifier votes, finally, a Boost algorithm is used for collecting voting results of all the base classifiers and carrying out weighted summation to find the prediction category with the largest votes, and when the votes are larger than the mean value, the activity intention is output.

4. The method for controlling the intention assimilation of an upper limb exoskeleton robot based on surface electromyography according to claim 2, wherein the step 3 comprises:

step 3.1: by a virtual target of a person

Evaluating the influence of human on the dynamics of the upper limb exoskeleton robot and the human interaction system, wherein the formula is as follows:

wherein the human controls the gain

And

using measured average values, or the same values as the robot controller gains, i.e.

The superscript symbol v represents the estimated value;

step 3.2: using an intention recognition method based on surface electromyography signals, or by internal model parameterization

And estimating, wherein the expression is as follows:

wherein the superscript symbol T represents transposition, and θ is the virtual target position of the person being calculated

The vector of parameters of (a) is,

t represents time, m is a predetermined parameter, and therefore

Is a quantity that is determined by the internal model parameters and varies with time;

state vector using upper limb exoskeleton robot and human interaction system

The extended model is obtained after substituting the formula (5):

where φ represents: state vector of upper limb exoskeleton robot and human interaction system, v ∈ N (0, E [ v, v ]^T]) Is the system noise, i.e., mean 0, variance E [ v, v^T]Gaussian noise of (2);

step 3.3: measuring the position and the speed of the end point of the robot and the interaction force with a human through a sensor to obtain a measurement vector of the upper limb exoskeleton robot and the human interaction system:

wherein, mu is N (0, E [ mu, mu ]^T]) Is the environmental measurement noise, i.e., the mean is 0 and the variance is E [ mu, mu ]^T]Gaussian noise of (2);

step 3.4: calculating an extended state estimate of the robot using the system observer:

wherein Λ represents an estimated value; z represents a measurement vector of the upper limb exoskeleton robot and the human interaction system;

linear quadratic estimation gain K-PH^TR^-1P is a positive definite matrix obtained by solving the ricatt differential equation:

wherein the noise covariance matrix Q ≡ E [ v, v ≡ E ≡ V^T]，R≡E[μ，μ^T]And a denotes a system matrix, and is substituted into equation (11) and expressed as follows:

5. the method for controlling intention assimilation of upper limb exoskeleton robot based on surface electromyography signals as claimed in claim 4, wherein interaction between human and robot is performed through relations τ and τ between human and robot_hTo determine:

when τ is τ_hRepresenting assistance of a robot using a human virtual target, the robot follows its original target τ_r；

When τ is 2 τ_r-τ_hWhile, the robot imposes its own target by eliminating the human target from the upper extremity exoskeleton robot and human interaction system;

interactive behavior assimilation from the estimated target position of the human target design robot using the following formula:

τ_rrepresenting an original target position of the upper limb exoskeleton robot; and lambda represents a hyper-parameter for adjusting the original target position and the human target position of the upper limb exoskeleton robot, and is dynamically adjusted according to the tail end position x.

6. The utility model provides an upper limbs ectoskeleton robot intention assimilation control system based on surface electromyogram signal which characterized in that includes:

module M1: establishing an upper limb exoskeleton robot dynamic model by using a Kenn method;

module M2: performing intention recognition through a surface electromyogram signal based on a dynamic model;

module M3: the intention assimilation control is performed by the virtual object.

7. The system for controlling the ideographic assimilation of an upper limb exoskeleton robot based on surface electromyography of claim 6, wherein the module M1 comprises:

module M1.1: there is no relative motion between the robot, the person and the object, and the robot and the person together manipulate the object, the object satisfying the dynamic equation:

wherein,

as the second derivative of the position coordinates of the object with respect to time, f and u_hIs the force of the robot and person on the object, M_oIs a mass matrix of the object, G_oIs the weight of the object;

module M1.2: establishing an upper limb exoskeleton robot dynamics model by using a Kenn method to obtain a joint space dynamics equation when the upper limb exoskeleton robot with n degrees of freedom is in contact with the environment:

wherein q is the joint coordinate of the robot, tau_qFor control input, J^T(q) is the Jacobian matrix, M_q(q) is the robot inertia matrix,

is the Coriolis and centrifugal torque, G_q(q) is the moment of gravity;

and converting into a robot operating space to obtain a kinetic equation:

wherein u represents the control input of the upper limb exoskeleton robot,

M_r、C_r、G_rrespectively representing an inertia matrix, a Coriolis force and centrifugal force matrix and a gravity matrix of the upper limb exoskeleton robot under a Cartesian space coordinate system, and symbols

Representing a pseudo-inverse of the matrix;

module M1.3: simultaneous equations (1) and (3) are obtained to obtain the combined kinetic equation of the object and the robot:

M≡M_o+M_r，G≡G_o+G_r，C≡C_r…………(5)

m, C, G respectively representing inertia matrix, Coriolis force and centrifugal force matrix and gravity matrix of the upper limb exoskeleton robot and human interaction system in a Cartesian space coordinate system;

module M1.4: the position, the speed and the human force of the tail end of the upper limb exoskeleton robot are measured, a robot controller with gravity compensation and linear feedback is adopted, and the expression is as follows:

where τ is the target position of the robot, L₁And L₂Is a gain corresponding to position error and velocity;

the force of a person acting on an object is modeled as:

wherein L is_h，1And L_h，2Control gain, τ, for humans_hAnd (5) bringing the formulas (6) and (7) into the formula (5) to obtain the dynamic equation of the upper limb exoskeleton robot and human interactive closed-loop system for the target position of the human:

8. the system for controlling the ideographic assimilation of an upper limb exoskeleton robot based on surface electromyography of claim 6, wherein the module M2 comprises:

module M2.1: collecting electromyographic signals of wrists, forearms and elbows of a person through an electromyograph;

module M2.2: filtering, data segmentation and feature extraction are carried out on the collected electromyographic signals, and feature extraction is carried out according to waveform types, so that the extracted features correspond to different intention categories;

module M2.3: training and predicting by using a multi-criterion linear programming in a database and combining a classification method of an online random forest;

module M2.4: during model prediction, the prediction category of each base classifier is compared with the corresponding confidence coefficient and a preset threshold value to determine whether the base classifier votes, finally, a Boost algorithm is used for collecting voting results of all the base classifiers and carrying out weighted summation to find the prediction category with the largest votes, and when the votes are larger than the mean value, the activity intention is output.

9. The system for controlling the ideographic assimilation of an upper limb exoskeleton robot based on surface electromyography of claim 7, wherein the module M3 comprises:

module M3.1: by a virtual target of a person

Evaluating the influence of human on the dynamics of the upper limb exoskeleton robot and the human interaction system, wherein the formula is as follows:

wherein the human controls the gain

And

using measured average values, or the same values as the robot controller gains, i.e.

The superscript symbol v represents the estimated value;

module M3.2: using methods of intention recognition based on surface electromyographic signals, or byInternal model parameterization pair

And estimating, wherein the expression is as follows:

wherein the superscript symbol T represents transposition, and θ is the virtual target position of the person being calculated

The vector of parameters of (a) is,

t represents time, m is a predetermined parameter, and therefore

Is a quantity that is determined by the internal model parameters and varies with time;

state vector using upper limb exoskeleton robot and human interaction system

The extended model is obtained after substituting the formula (5):

where φ represents: state vector of upper limb exoskeleton robot and human interaction system, v ∈ N (0, E [ v, v ]^T]) Is the system noise, i.e., mean 0, variance E [ v, v^T]Gaussian noise of (2);

module M3.3: measuring the position and the speed of the end point of the robot and the interaction force with a human through a sensor to obtain a measurement vector of the upper limb exoskeleton robot and the human interaction system:

wherein, mu is N (0, E [ mu, mu ]^T]) Is the environmental measurement noise, i.e., the mean is 0 and the variance is E [ mu, mu ]^T]Gaussian noise of (2);

module M3.4: calculating an extended state estimate of the robot using the system observer:

wherein Λ represents an estimated value; z represents a measurement vector of the upper limb exoskeleton robot and the human interaction system;

linear quadratic estimation gain K-PH^TR^-1P is a positive definite matrix obtained by solving the ricatt differential equation:

wherein the noise covariance matrix Q ≡ E [ v, v ≡ E ≡ V^T]，R≡E[μ，μ^T]And a denotes a system matrix, and is substituted into equation (11) and expressed as follows:

10. the system for controlling assimilation of upper limb exoskeleton robot based on surface electromyography signals of claim 9, wherein interaction between human and robot is performed through the relations τ and τ between human and robot_hTo determine:

when τ is τ_hRepresenting assistance of a robot using a human virtual target, the robot follows its original target τ_r；

When τ is 2 τ_r-τ_hWhile, the robot imposes its own target by eliminating the human target from the upper extremity exoskeleton robot and human interaction system;

interactive behavior assimilation from the estimated target position of the human target design robot using the following formula:

τ_rrepresenting an original target position of the upper limb exoskeleton robot; and lambda represents a hyper-parameter for adjusting the original target position and the human target position of the upper limb exoskeleton robot, and is dynamically adjusted according to the tail end position x.

Images