CN110262256B - Multilateral self-adaptive sliding mode control method of nonlinear teleoperation system - Google Patents

Abstract

The invention discloses a multilateral self-adaptive sliding mode control method of a nonlinear teleoperation system. The method is based on a radial basis function, estimates the non-power environment parameters of the slave-end environment dynamics, and transmits the non-power environment parameters back to the main end through a communication channel to reconstruct the environment force of the main end; aiming at the problems of nonlinearity and various uncertainties of a master robot and a slave robot, the invention designs a track generator and a nonlinear adaptive sliding mode controller based on a radial basis function network at the master end and the slave end respectively, designs an adaptive rate for training a nonlinear function containing system modeling information on line, and ensures the stability and the accurate position tracking performance of the system; aiming at the problem of signal communication among multiple robots, the control force distribution of the multiple slave robots is realized by designing a cooperative force distribution algorithm, so that the cooperative operation performance of the multiple slave robots on the operation task is improved.

Description

A Multilateral Adaptive Sliding Mode Control Method for Nonlinear Teleoperating Systems

technical field

The invention belongs to the field of teleoperation control, in particular to a multilateral adaptive sliding mode control method of a nonlinear teleoperating system, which can simultaneously ensure the stability and transparency of the nonlinear teleoperating system and the cooperative operation of multiple master-slave robots. .

Background technique

With the development of complex and delicate tasks, the teleoperation technology that relies on human-computer interaction has been continuously used in industrial environments, especially the development of multilateral teleoperation technology that relies on the collaborative operation of multi-master-slave robots, that is, through multiple operations. It is widely used in space exploration, deep-sea sampling, telemedicine, safety inspection and other fields, and is used as a robot. It has been widely studied as an important supporting technology in the application field.

However, the transmission of the signal in the communication channel will inevitably cause communication delay, which will affect the accuracy of the command signal received from the master robot from the robot, and deteriorate the stability and transparency of the teleoperating system. In addition, due to complex or delicate slave task requirements, multiple robots with multiple degrees of freedom are often required to work together, and such robots often have nonlinearity and various uncertainties, and the signal communication between multiple robots will It makes the signal transmission in the communication channel more complicated, and the traditional linear teleoperating system structure based on passive theory and four-channel structure cannot achieve better control performance. Therefore, it is aimed at the trade-off between the stability and transparency of the teleoperating system caused by the communication delay, as well as the nonlinearity, various uncertainties and signal communication between multiple robots with multiple degrees of freedom.

SUMMARY OF THE INVENTION

The invention proposes a multilateral adaptive sliding mode control method of a nonlinear teleoperating system, so as to solve the problems of stability, transparency, nonlinearity, various uncertainties and multi-robot cooperative operation existing in the traditional multilateral teleoperating system. question.

In order to achieve the above purpose, the specific content of the technical solution of the invention is as follows:

A multilateral adaptive sliding mode control method of a nonlinear teleoperating system, comprising the following steps:

(1) Establish the dynamic model of nonlinear multilateral teleoperating system.

(2) Design of adaptive sliding mode controller for slave robot based on radial basis neural network.

(3) Working environment estimation and master-end environment reconstruction based on radial basis neural network function.

(4) The adaptive sliding mode controller of the main robot is designed based on radial basis neural network.

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

1. Based on the radial basis neural network function, the non-power environment parameters of the environmental dynamics of the slave end are estimated, and transmitted back to the master end through the communication channel to reconstruct the environment force of the master end, which not only avoids the transmission of power signals in the communication channel , and provide accurate force feedback information for the operator.

2. Based on the radial basis neural network function, the nonlinear function containing the system modeling information is trained online by designing the adaptive rate, so as to solve the various uncertainties of the master-slave robot of the multilateral teleoperating system.

3. Through the nonlinear adaptive sliding mode control method based on radial basis neural network, the slave robot can track the trajectory signal of the master robot in real time and accurately. When the system has communication delay, nonlinearity and various uncertainties, It can improve the location tracking performance of the system.

4. By designing the Lyapunov function, the boundedness of all signals in the nonlinear multilateral teleoperating system is guaranteed, thereby maintaining the stability of the system.

5. By designing a collaborative force distribution algorithm, the control force distribution of multiple slave robots is realized, thereby improving the collaborative operation performance of multiple slave robots for job tasks.

Description of drawings

Fig. 1 is the multilateral adaptive sliding mode control block diagram of the nonlinear teleoperating system based on radial basis neural network proposed by the present invention;

Fig. 2 is a functional block diagram of the radial basis neural network proposed by the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

The present invention will now be further described in conjunction with the embodiments and accompanying drawings:

The technical implementation scheme of the present invention is:

1) Establish the dynamic model of the nonlinear multilateral teleoperating system, which is as follows:

1-1) Establish the dynamic model of multi-master robot, multi-slave robot and environment interaction

和

表示第i个主从机器人位置、速度和加速度信号，

表示第i个主机器人的末端位置，

表示作业任务中目标物体的质心位置，D_m,i和D_s表示质量惯性矩阵，C_m,i和C_s表示科氏力/向心力矩阵，G_m,i和G_s表示重力矩阵，d_m,i和d_s表示外干扰和建模误差，u_m,i和u_s表示控制输入，F_h,i表示第i个操作者的操作力，F_e表示从机器人与作业任务中的环境力，i＝1,2,....,n。in,

and

represents the position, velocity and acceleration signals of the i-th master-slave robot,

represents the end position of the i-th master robot,

Indicates the position of the center of mass of the target object in the task, D _{m, i} and D _s represent the mass inertia matrix, C _{m, i} and C _s represent the Coriolis force/centripetal force matrix, G _{m, i} and G _s represent the gravity matrix, d _{m , i} and d _s represent external disturbance and modeling error, _{um, i} and u _s represent the control input, F _{h, i represent} the operating force of the ith operator, and Fe represent the environmental force from the robot and the task , i=1,2,....,n.

The above system has the following characteristics:

和

为斜对称矩阵；①

and

is an obliquely symmetric matrix;

②Part of the kinetic equations in equations (1) and (2) can be written in the form of the following linear equations:

Among them, W _m,i and W _s represent the uncertain parameters of the master-slave robot, and H represents the neural network matrix.

1-2) Establish a dynamic model of the working environment

Among them, We represent unknown _slave environment parameters.

2) Design the adaptive sliding mode controller of the slave robot based on the radial basis neural network, specifically:

2-1) Since the transmission of the signal in the communication channel will inevitably generate a communication delay, the position signal x _m,i (t) of the master robot is transmitted to the slave through the communication channel to obtain the delayed position signal x _m,i ( tT(t)), the trajectory generator from the robot is designed as follows:

输出用于从机器人跟踪的理想轨迹信号

其中，l_o,i表示目标物体与机器人末端位置间的关系转换，T(t)为系统的通信时延。Average position signal via input delay

Output ideal trajectory signal for tracking from robot

Among them, l _o,i represents the relationship conversion between the target object and the end position of the robot, and T(t) is the communication delay of the system.

2-2) Define the sliding surface _s of the slave robot as follows:

Among them, e _s =x _sd,o -x _s,o represents the tracking error between the slave robot and the target object,

因此，2-3) Substitute the tracking error into (5) to get

therefore,

in,

2-4) Design the slave controller according to (6) to ensure the stability of the slave terminal system. The designed controller u _s is:

u _s =σ _s +k _sv s _s -F _e -κ _sN sat(s _s ) (7)

where k _sv >0, k _sN >0.

In the slave controller (7), sat(s _s ) represents a sliding mode saturation function that avoids chattering and can be defined as:

where μ represents the boundary layer;

σ _s represents a radial basis neural network function for estimating the nonlinear function z _s , which can be defined as:

为自适应参数，

in,

is an adaptive parameter,

2-5) Design the Lyapunov function V _s from the terminal system as:

表示径向基神经网络函数的估计误差。in,

represents the estimation error of the radial basis neural network function.

The adaptive rate for designing W _s based on the Lyapunov function V _s is:

where k _s >0, Γ _s >0.

2-6) According to the slave controller (7), in order to obtain the control input u _s,i of each slave robot, the synergistic force distribution algorithm is designed as follows:

Q表示不同作业需求的权重系数，F_s ^*表示各个从机器人与目标物体的内部力，且N_sF_s ^*＝0。in,

Q represents the weight coefficient of different job requirements, F _s ^* represents the internal force between each slave robot and the target object, and N _s F _s ^* =0.

3) Working environment estimation and master-end environment reconstruction based on radial basis neural network function, specifically:

3-1) Write the dynamic model (3) of the slave operating environment in the form of a radial basis neural network function, then:

相关。where x _ew and

related.

为环境的最优估计参数，Ω_e和Ω_e0分别表示x_ew和W_e的有界集，通过MATLAB的神经网络工具箱能够实现从端作业环境的在线估计。3-2) Definition

For the optimal estimation parameters of the environment, Ω _e and Ω _e0 represent the bounded sets of x _ew and We respectively _, and the online estimation of the slave operating environment can be realized through the neural network toolbox of MATLAB.

传递到主端，从而得到主端的重构环境力为：3-3) Due to the existence of the communication delay T(t), in order to avoid the transmission of the power signal in the communication channel affecting the stability of the multilateral teleoperating system, the estimated value of the non-power environment parameters of the slave

Pass it to the master, so as to obtain the reconstructed environment force of the master as:

相关。where x _emw is the same as

related.

4) Design the adaptive sliding mode controller of the main robot based on the radial basis neural network, specifically:

4-1) Define x _{md, i} as the ideal trajectory signal of the main robot, and satisfy:

D_d,C_d,G_d表示主机器人的阻抗系数。通过选取适当的阻抗系数，(15)-(16)能够生成无源的主机器人理想轨迹x_md,i。Among them, i=1,2,...,n,

D _d , C _d , and G _d represent the impedance coefficients of the master robot. By choosing appropriate impedance coefficients, (15)-(16) can generate the passive master robot ideal trajectory x _md,i .

4-2) Define the sliding surface s _m,i of the main robot as follows:

Among them, em _,i =x _md,i -x _m,i represents the tracking error of the main robot,

因此，4-3) Substitute the tracking error into (17) to get

therefore,

in,

4-4) Design the main controller according to (18) to ensure the stability of the main terminal system. The designed controller _um,i is:

u _m,i =σ _m,i +k _mv, is _m,i -F _h,i -κ _mN, isat(s _m,i ) (19)

where k _mv,i >0, and k _mN,i >0.

In the slave controller (19), sat(s _m ) represents a sliding mode saturation function that avoids chattering and can be defined as:

where μ represents the boundary layer;

σ _m,i represents a radial basis neural network function for estimating the nonlinear function z _m,i , which can be defined as:

为自适应参数，

in,

is an adaptive parameter,

4-5) Design the Lyapunov function V _m,i of the main terminal system as:

表示径向基神经网络函数的估计误差。in,

represents the estimation error of the radial basis neural network function.

Based on the Lyapunov function V _m,i , the adaptive rate of designing W _m,i is:

where k _m,i >0, Γ _m,i >0.

Claims (7)

1. a kind of multilateral adaptive sliding mode control method of nonlinear teleoperating system, is characterized in that, comprises the following steps: 1) Establish the dynamic model of the nonlinear multilateral teleoperating system, which is as follows: 1-1) Establish the dynamic model of multi-master robot, multi-slave robot and environment interaction

Among them, q _m,i ,

and q _s,i ,

Indicates the position, velocity and acceleration signals of the i-th master-slave robot, x _m,i ,

Represents the end position, end velocity and end acceleration signal of the i-th master robot, x _s,o ,

Indicates the centroid position, centroid velocity and centroid acceleration signal of the target object in the task, D _m,i and D _s indicate the mass inertia matrix, C _m,i and C _s indicate the Coriolis force/centripetal force matrix, G _m,i and G _s represents the gravity matrix, d _m,i and d _s represent the external disturbance and modeling error, _um,i and u _s represent the control input, F _{h,i represents} the operating force of the ith operator, and Fe represents the slave robot and the environmental force in the task, i=1,2,....,n; The above system has the following characteristics: ①

and

is an obliquely symmetric matrix; ②Part of the kinetic equations in equations (1) and (2) can be written in the form of the following linear equations:

Among them, W _m,i and W _s represent the uncertain parameters of the master-slave robot, and H represents the neural network matrix; 1-2) Establish a dynamic model of the working environment

Among them, We represent unknown _slave environment parameters; 2) Design the adaptive sliding mode controller of the slave robot based on the radial basis neural network, specifically: 2-1) Design the trajectory generator of the slave robot to output the ideal trajectory, ideal velocity and acceleration signals x _sd,o (t) for tracking of the slave robot,

2-2) Define the sliding surface _s of the slave robot as follows:

Among them, es = x _{sd, o} - x _{s , o} represents the tracking error between the slave robot and the target object,

Indicates the sliding surface adjustment parameters; 2-3) Substitute the tracking error into (5) to get

therefore,

in,

represents the unknown system dynamics parameters of the slave robot; 2-4) Design the slave controller according to (6) to ensure the stability of the slave terminal system. The designed controller u _s is: u _s =σ _s +k _sv s _s -F _e -κ _sN sat(s _s ) (7) where k _sv >0 and k _sN >0 represent performance tuning parameters from the controller performance, and σ _s represents a radial basis neural network function used to estimate the nonlinear function z _s ; 2-5) Design the Lyapunov function V _s from the terminal system as:

in,

represents the estimation error of the radial basis neural network function; 2-6) Based on the Lyapunov function V _s , the adaptive rate of designing W _s is:

where k _s >0 and Γ _s >0 represent the learning rate adjustment parameters of the adaptation rate,

represents the input of the radial basis neural network function σ _s ; 2-7) According to the slave controller (7), in order to obtain the control input u _s,i of each slave robot, design a synergy distribution algorithm; 3) Working environment estimation and master-end environment reconstruction based on radial basis neural network function, specifically: 3-1) Write the dynamic model (3) of the slave operating environment in the form of a radial basis neural network function, then:

Among them, x _ew represents the input of the neural network function, and is the same as x _s,o ,

related; 3-2) Definition

are the optimal estimation parameters of the environment, Ω _e and Ω _e0 represent the bounded sets of x _ew and We respectively _, and the online estimation of the slave operating environment is realized by the neural network toolbox of MATLAB; 3-3) Due to the existence of the communication delay T(t), in order to avoid the transmission of the power signal in the communication channel affecting the stability of the multilateral teleoperating system, the estimated value of the non-power environment parameters of the slave

Pass it to the master, so as to obtain the reconstructed environment force of the master as:

Among them, x _emw represents the input of the neural network function, and is the same as x _md,i ,

related; 4) Design the adaptive sliding mode controller of the main robot based on the radial basis neural network, specifically: 4-1) Define x _{md, i} as the ideal trajectory signal of the main robot, and satisfy:

Among them, i=1,2,...,n,

D _d , C _d , G _d represent the impedance coefficients of the main robot; by selecting the impedance coefficients, (15)-(16) can generate the passive ideal trajectory x _md,i of the main robot; 4-2) Define the sliding surface s _m,i of the main robot as follows:

Among them, em _,i =x _md,i -x _m,i represents the tracking error of the main robot,

Indicates the sliding surface adjustment parameters; 4-3) Substitute the tracking error into (17) to get

therefore,

in,

represents the unknown system dynamics parameters of the main robot; 4-4) Design the main controller according to (18) to ensure the stability of the main terminal system. The designed controller _um,i is: u _m,i =σ _m,i +k _mv, is _m,i -F _h,i -κ _mN, isat(s _m,i ) (19) where k _mv,i >0 and k _mN,i >0 represent the performance adjustment parameters of the main controller performance, and σ _m,i represents a radial basis neural network function for estimating the nonlinear function z _m,i ; 4-5) Design the Lyapunov function V _m,i of the main terminal system as:

in,

represents the estimation error of the radial basis neural network function; 4-6) Based on the Lyapunov function V _m,i to design the adaptive rate of W _m,i :

Among them, k _m,i >0 and Γ _m,i >0 represent the learning speed adjustment parameter of the adaptation rate,

represents the input of the radial basis neural network function σ _m,i . 2. The multilateral adaptive sliding mode control method according to claim 1, wherein in the step 2-1), because the transmission of the signal in the communication channel will inevitably generate a communication time delay, the position of the main robot The signal x _m,i (t) is transmitted to the slave through the communication channel to obtain the delayed position signal x _m,i (tT(t)). The trajectory generator of the slave robot is designed as follows:

where t _f represents the time constant; the average position signal through the input delay

Output ideal trajectory, ideal velocity and ideal acceleration signals x _sd,o (t), ideal for tracking from the robot

Among them, l _{o, i} represent the relationship conversion between the target object and the end position of the robot, and T(t) is the communication delay of the system. 3. Multilateral adaptive sliding mode control method according to claim 1, is characterized in that, in described step 2-4), sat (s _s ) represents a kind of sliding mode saturation function that avoids chattering, is defined as:

where μ represents the boundary layer, and sgn(s _s ) represents the sign function. 4. The multilateral adaptive sliding mode control method according to claim 1, wherein in the step 2-4), σ _s is defined as:

in,

is an adaptive parameter. 5. The multi-side adaptive sliding mode control method according to claim 1, wherein, in the step 2-7), according to the slave controller (7), in order to obtain the control input u _{s of each slave robot, i} , the design synergy distribution algorithm is as follows:

in,

is the distribution coefficient, and

Θ represents the weight coefficient of different job requirements,

represents the internal force of each slave robot and the target object, and

6. The multilateral adaptive sliding mode control method according to claim 1, wherein in the step 4-4), sat(s _m,i ) represents a sliding mode saturation function that avoids chattering, and defines for:

where μ represents the boundary layer, and sgn(s _m,i ) represents the sign function. 7. The multilateral adaptive sliding mode control method according to claim 1, wherein in the step 4-4), σ _m,i is defined as:

in,

is an adaptive parameter.

Images