Full-state control method for the master-slave robot system with flexible joints and time-varying delays

Abstract

A full-state control method for a master-slave robot system with flexible joints and time-varying delays is provided. In a teleoperation system formed by connecting a master robot and a slave robot through network, a proportional damping controller based on a position error and velocities, and a full-state feedback controller based on backstepping are designed for the master robot and the slave robot, respectively. High-dimension uniform accurate differentiators are designed to realize an exact difference to the virtual controllers. Delay-dependent stability criteria are established by constructing Lyapunov functions. Therefore, the criteria for selecting controller parameters are presented such that the global stability of the master-slave robot system with flexible joints and time-varying delays is realized. For the master-slave robot system with flexible joints, the global precise position tracking performance is realized by adopting a full-state feedback controller based on the backstepping method and the high-dimensional uniform accurate differentiators. Moreover, the global asymptotic convergence of the system is guaranteed and the robustness of the system is improved.

Description

TECHNICAL FIELD

This present invention relates to the control field of the flexible master-slave robot system, in particular to a full-state control method for a master-slave robot system with flexible joints and time-varying delays.

BACKGROUND

The master-slave robot system with flexible joints is a very complex dynamic system with high-nonlinearity, strong-coupling, and time-varying delays. To complete some complex and difficult work, flexible joint manipulator has the advantages of small volume, high flexibility, high deadweight ratio of load, low energy consumption, and wider working space compared with rigid joint manipulators. In most cases, the teleoperation system is a kind of remote operating system to complete more complex operations. It mainly consists of an operator, a master robot system, a network information transmission channel, a slave robot system, and external working environment. Its workflow can be roughly described as follows: the operator manipulates the master robot directly, so the control instructions will be sent to the slave robot through the network information transmission channel, then the slave robot will act on the remote environment according to the instructions. Meanwhile, the relevant information will also be feedback to the operator timely. Therefore, a closed-loop teleoperation system can be constituted and the operating tasks can be completed effectively. For slave robot systems, n-DOF (Degree Of Freedom) robots with flexible joints are usually used to execute commands from the master side.

With the development and progress of science and technology, the exploration of human beings has been constantly expanding. Since flexible joints are characterised by small size and high flexibility, a flexible master-slave robot system is more widely used. However, vibration of the flexible joint may affect the control accuracy and system stability. Therefore, the proposed control method can not only enable the robot to complete the work objectives but also eliminate the vibration of the flexible joint to ensure the stability of the system.

For the master-slave robot system with flexible joints, a full-state feedback control method based on the backstepping technique is proposed. The position error between the master and slave robots with time delays are used to analyze the global tracking performance. The globally accurate position tracking is realized and the robustness of the closed-loop system is improved. The high-dimension uniform accurate differentiators are employed to realize the precise difference to the virtual controllers and the convergence of the system is improved.

SUMMARY

The purpose of the invention is to provide a full-state control method for the master-slave robot system with flexible joints and time-varying delays to solve the problem of position tracking and stability existing in the master-slave robot system with flexible joints. To achieve these above objectives, the following technical schemes are adopted.

A full-state control method for the master-slave robot system with flexible joints and time-varying delays includes the following steps:

Step 1: connecting a master robot and a slave robot through the network to form a teleoperation system.

Step 2: measuring system parameters of the master robot and the slave robot, among which the position and velocity information of joints and motors are measured in real-time.

Step 3: designing the first virtual controller and the second virtual controller, respectively.

The first virtual controller is X^*_s3=X_s1+S_s⁻¹(k_s(X_m1(t−T_m(t))−X_s1)−α_sX_s2), the second virtual controller is

$X_{s4}^{*}={\stackrel{.}{X}}_{s3}^{*}-2\frac{k_m}{k_s}{S}_s^T{X}_{s2}-k_1\left(X_{s3}-X_{s3}^{*}\right),$ the subscripts m and s denotes the master robot and the slave robot, respectively, X*_s3 and X*_s4 are virtual controllers, {dot over (X)}*_s3, is the first derivative of the virtual controller X*_s3, T_m(t) and T_s(t) are forward time delay (from the master robot to the slave robot) and feedback time delay (from the slave robot to the master robot), respectively, α_s is a damping coefficient which is a positive constant, k_m, k_s>0 are proportional coefficients, S_s⁻¹ and S_s^T are the inverse matrix and the transpose matrix of a diagonal positive-definite constant matrix S_s which contains the joint stiffness of the slave robot, respectively, and k₁ is selected to be a positive constant.

Step 4: designing high-dimension uniform accurate differentiators to carry out a precise difference to the first virtual controller and the second virtual controller.

Step 41:

letting X₁=X*_s3, X₂={dot over (X)}*_s3, σ₁=X₁−Y₁, σ₂=X₂−Y₂,

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_1=X_2\\ {\stackrel{.}{X}}_2={\ddot{X}}_{s3}^{*}\end{array}& \left(1\right)\\ \{\begin{array}{c}{\stackrel{.}{Y}}_1={\lambda }_1\frac{{\sigma }_1}{{{\sigma }_1}^{1/2}}+{\lambda }_2{\sigma }_1{{\sigma }_1}^{P-1}+Y_2\\ {\stackrel{.}{Y}}_2={\alpha }_1\frac{{\sigma }_1}{{\sigma }_1}\end{array}& \left(2\right)\end{array}$ with the Equations (1) and (2),

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{\sigma }}_1=-{\lambda }_1\frac{{\sigma }_1}{{{\sigma }_1}^{1/2}}-{\lambda }_2{\sigma }_1{{\sigma }_1}^{P-1}+{\sigma }_2\\ {\stackrel{.}{\sigma }}_2=-{\alpha }_1\frac{{\sigma }_1}{{\sigma }_1}+{\ddot{X}}_{s3}^{*}\end{array}& \left(3\right)\end{array}$ is derived, where X*_s3 denotes the virtual controller, {dot over (X)}*_s3 denotes the first derivative of the virtual controller X*_s3, {umlaut over (X)}_s3 denotes the second derivative of the virtual controller X*_s3, Y₁ is an estimate of the virtual controller X*_s3, Y₂ is an estimate of {dot over (X)}*_s3, σ₁ and σ₂ are estimation errors, λ₁, λ₂, α₁>0 are system control gains, P>1 is a constant, {umlaut over (X)}*_s3 supposed to be bounded and satisfies ∥{umlaut over (X)}*_s3∥≤L₃ with a known constant L₃>0, and if the parameters are selected to satisfy the conditions α₁>4L₃, λ₁>√{square root over (2α₁)}, the estimation errors σ₁, σ₂ will globally converge to the origin quickly, thus a precise difference value {dot over (X)}*_s3 of X*_s3 is obtained.

Step 42:

letting X₃=X*_s4, X₄={dot over (X)}*_s4, σ₃=X₃−Y₃, σ_z, =X₄−Y₄

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_3=X_4\\ {\stackrel{.}{X}}_3={\ddot{X}}_{s4}^{*}\end{array}& \left(4\right)\\ \{\begin{array}{c}{\stackrel{.}{Y}}_3={\lambda }_3\frac{{\sigma }_3}{{{\sigma }_3}^{1/2}}+{\lambda }_4{\sigma }_3{{\sigma }_3}^{P-1}+Y_4\\ {\stackrel{.}{Y}}_4={\alpha }_2\frac{{\sigma }_3}{{\sigma }_3}\end{array}& \left(5\right)\end{array}$ with the Equations (4) and (5),

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{\sigma }}_3=-{\lambda }_3\frac{{\sigma }_3}{{{\sigma }_3}^{1/2}}-{\lambda }_4{\sigma }_3{{\sigma }_3}^{P-1}+{\sigma }_4\\ {\stackrel{.}{\sigma }}_4=-{\alpha }_2\frac{{\sigma }_3}{{\sigma }_3}+{\ddot{X}}_{s4}^{*}\end{array}& \left(6\right)\end{array}$ is derived, where X*_s4 denotes a virtual controller, {dot over (X)}*_s4 denotes the first derivative of the virtual controller X*_s4, {umlaut over (X)}*_s4 denotes the second derivative of the virtual controller X*_s4, Y₃ is an estimate of the virtual controller X*_s4, Y₄ is an estimate of {dot over (X)}*_s4, σ₃ and σ₄ are estimation errors, λ₃, λ₄, α₂>0 are system control gains, P>1 is a constant, {umlaut over (X)}*_s4 is supposed to be bounded ∥{umlaut over (X)}*_s4∥≤L₄ with a known positive constant L₄>0, and if parameters are selected to satisfy the conditions α₂>4L₄, λ₃>√{square root over (2α₂)}, the estimation errors σ₃, σ₄ will globally converge to the origin quickly, thus a precise difference value {dot over (X)}*_s4 of X*_s4 is obtained.

Step 5: designing the controllers of the master-slave robot system with flexible joints by using the backstepping method as follows

${\tau }_m=-k_m\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-{\alpha }_m{X}_{m2}{\tau }_s=S_s\left(X_{s3}-X_{s1}\right)+J_s\left({\stackrel{.}{X}}_{s4}^{*}-\left(X_{s3}-X_{s3}^{*}\right)-k_2\left(X_{s4}-X_{s4}^{*}\right)\right)$ $X_{s3}^{*}=X_{s1}+S_s^{-1}\left(k_s\left(X_{m1}\left(t-T_m\left(t\right)\right)-X_{s1}\right)-{\alpha }_s{X}_{s2}\right)$ $X_{s4}^{*}={\stackrel{.}{X}}_{s3}^{*}-2\frac{k_m}{k_s}{S}_s^T{X}_{s2}-k_1\left(X_{s3}-X_{s3}^{*}\right)$ where τ_m and τ_s are control torques provided by the controllers, X*_s3 and X*_s4 are virtual controllers, {dot over (X)}*_s3 and {dot over (X)}*_s4 are the first derivatives of the virtual controllers X*_s3 and X*_s4, respectively, T_m(t) and T_s(t) are forward time delay (from the master robot to the slave robot) and backward time delay (from the slave robot to the master robot), respectively, α_m and α_s are damping coefficients which are positive constants, k_m, k_s>0 are proportional coefficients, S_s⁻¹ and S_s^T are the inverse matrix and the transpose matrix of a diagonal positive-definite constant matrix S_s which contains the joint stiffness of the slave robot, respectively, J_s is a diagonal constant matrix of the moment of actuator interia, and k₁ and k₂ are selected to be positive constants.

Step 6: establishing the delay-dependent system stability criteria by constructing Lyapunov Equations, providing the criteria for selecting controller parameters, and realizing the global stability of the master-slave robot system with flexible joints and time-varying delays.

When the controller parameters are selected such that the following inequalities hold,

$-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}<0,-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}<0$ the joint and motor velocities {dot over (q)}_m, {dot over (q)}_s, {dot over (θ)}_s and position error q_m−q_s of the master-slave robot system with flexible joints are all bounded.

Furthermore, when a force F_h exerted by an operator to the master robot and a force F_e exerted by the external environment to the slave robot are both zero, the controllers are designed as follows:

$$\begin{gathered} \begin{array}{cc}{\tau }_m=-k_m\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-{\alpha }_m{X}_{m2}+G_m\left(X_{m1}\right)\text{ \\ τs=Ss(Xs⁢3-Xs⁢1)+Js(X.s⁢4*-(Xs⁢3-Xs⁢3*)-k2(Xs⁢4-Xs⁢4*))⁢ \\ Xs⁢3*=Xs⁢1+Ss-1(ks(Xm⁢1(t-Tm(t))-Xs⁢1)-αs⁢Xs⁢2+Gs(Xs⁢1))⁢ \\ Xs⁢4*=X.s⁢3*-2⁢kmks⁢SsT⁢Xs⁢2-k1(Xs⁢3-Xs⁢3*)}& \left(7\right)\end{array} \end{gathered}$$ where τ_m and τ_s are control torques provided by the controllers, X*_s3 and X*_s4 are virtual controllers, {dot over (X)}*_s3 and {dot over (X)}*_s4 are the first derivatives of the virtual controllers X*_s3 and X*_s4, respectively, T_m(t) and T_s(t) are forward time delay (from the master robot to the slave robot) and backward time delay (from the slave robot to the master robot), respectively, α_m and α_s are damping coefficients which are positive constants, k_m, k_s>0 are proportional coefficients, S_s⁻¹ and S_s^T are the inverse matrix and the transpose matrix of a diagonal positive-definite constant matrix S_s which contains the joint stiffness of the slave robot, respectively, G_m(X_m1), G_s(X_s1) are the gravitational torques of the master robot and the slave robot, J_s is a diagonal constant matrix of the moment of actuator inertia, and k₁ and k₂ are selected to be positive constants.

When the controller parameters are selected such that the following inequalities hold,

$-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}<0,-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}<0$ it can be guaranteed that the joint and motor velocities {dot over (q)}_m, {dot over (q)}_s, {dot over (θ)}_s and the position error q_m−q_s of the master-slave robot system with flexible joints will converge to zero asymptotically and that the global master-slave robot system with flexible joints will asymptotically become stable, where I is the identity matrix, Z⁻¹ and P⁻¹ are the inverse matrices of positive definite matrices Z and P, respectively, α_m and α_s are damping coefficients which are positive constants, k_m, k_s>0 are proportional coefficients, and it is supposed that the time delay T_m(t) and T_s(t) are bounded, i.e. there are positive scalars T_m and T_s, such that the inequalities T_m(t)≤T_m, T_s(t)≤T_s hold.

Preferably, the system parameters of the master robot and the slave robot include the length and the mass of the robot manipulators of the master robot and the slave robot, the positions and the velocities of the robot joints and motor of the master robot and the slave robot, and the force exerted by the operator and the force exerted by the external environment measured by using force sensors.

Preferably, the design process of the first virtual controller and the second virtual controller is as follows:

As for the first virtual controller, the first Lyapunov Equation is selected as follows,

$$\begin{gathered} \begin{array}{cc}{V}_1=V_{11}+V_{12}+V_{13}\text{ \\ V1⁢1=Xm⁢2T⁢Mm(Xm⁢1)⁢Xm⁢2+kmks⁢Xs⁢2T⁢Ms(Xs⁢1)⁢Xs⁢2+2⁢(Um(Xm⁢1)-βm)+2⁢kmks⁢(Us(Xs⁢1)-βs)+2⁢∫0t(-Xm⁢2T(σ)⁢Fh(σ)+kmks⁢Xs⁢2T(σ)⁢Fe(σ))⁢d⁢σ⁢ \\ V1⁢2=km(Xm⁢1-Xs⁢1)T⁢(Xm⁢1-Xs⁢1)⁢ \\ V1⁢3=∫-T\_m0∫t+θtXm⁢2T(ξ)⁢Z⁢Xm⁢2(ξ)⁢d⁢ξ⁢d⁢θ+∫-T\_s0∫t+θtXs⁢2T(ξ)⁢P⁢Xs⁢2(ξ)⁢d⁢ξ⁢d⁢θ}& \left(8\right)\end{array} \end{gathered}$$ where the integral terms satisfy ∫₀^t−X_m2^T(σ)F_h(σ)dσ≥0, ∫₀^tX_s2^T(σ)F_e(σ)dσ≥0, Z and P are positive definite matrices, it is supposed that the time delay T_m(t) and T_s(t) are bounded, i.e. there are positive scalars T_m and T_s, such that T_m(t)≤T_m, T_s(t)≤T_s, M_m(X_m1) and M_s(X_s1) are the positive-definite inertia matrices of the master robot and the slave robot, respectively, U_m(X_m1) and U_s(X_s1) are the potential energy of the master robot and the slave robot satisfying

$G_m\left(X_{m1}\right)=\frac{\partial U_m\left(X_{m1}\right)}{\partial X_{m1}},G_s\left(X_{s1}\right)=\frac{\partial U_s\left(X_{s1}\right)}{\partial X_{s1}},$ there are positive scalars β_m and β_s such that U_m(X_m1)≥β_m, U_s(X_s1)≥β_s, and k_m, k_s>0 are the proportional coefficients.

The time derivative of V₁₁ is given by

$\begin{array}{cc}\begin{array}{c}{\stackrel{.}{V}}_{11}=-2k_m{X}_{m2}^T\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-\\ 2{\alpha }_m{X}_{m2}^T{X}_{m2}+2\frac{k_m}{k_s}{X}_{s2}^T{S}_s\left(X_{s3}-X_{s1}\right)\\ =-2k_m{X}_{m2}^T\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-2{\alpha }_m{X}_{m2}^T{X}_{m2}+\\ 2\frac{k_m}{k_s}{X}_{s2}^T\left(S_s\left(X_{s3}-X_{s3}^{*}\right)+S_s{X}_{s3}^{*}-S_s{X}_{s1}\right)\end{array}& \left(9\right)\end{array}$ where S_s is a diagonal positive-definite constant matrix which contains the joint stiffness of the slave robot, S_s⁻¹ is the inverse matrix of S_s.

With Equation (9), the first virtual controllerX*_s3=X_s1+S_s⁻¹(k_s(X_m1(t−T_m(t))−X_s1)−α_sX_s2) is derived.

As for the second virtual controller, the second Lyapunov equation is selected as follows,V₂=V₁+½(X_s3−X*_s3)^T(X_s3−X*_s3)  (10)

The time derivative of V₂ is given by{dot over (V)}₂={dot over (V)}₁+(X_s3−X*_s3)^T(X_s4−{dot over (X)}*_s3)={dot over (V)}₁+(X_s3−X*_s3)^T(X_s4−X*_s4+X*_s4−{dot over (X)}*_s3)  (11)

With Equation (11), the second virtual controller

$X_{s4}^{*}={\stackrel{.}{X}}_{s3}^{*}-2\frac{k_m}{k_s}{S}_s^T{X}_{s2}-k_1\left(X_{s3}-X_{s3}^{*}\right)$ is derived.

Preferably, a high-dimensional uniform accurate differentiators is designed as follows.

Letting X₁=X*_s3, X₂={dot over (X)}*_s3, σ₁=X₁−Y₁, σ₂=X₂−Y₂,

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_1=X_2\\ {\stackrel{.}{X}}_2={\ddot{X}}_{s3}^{*}\end{array}& \left(12\right)\end{array}$ $\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{Y}}_1={\lambda }_1\frac{{\sigma }_1}{{{\sigma }_1}^{1/2}}+{\lambda }_2{\sigma }_1{{\sigma }_1}^{P-1}+Y_2\\ {\stackrel{.}{Y}}_2={\alpha }_1\frac{{\sigma }_1}{{\sigma }_1}\end{array}& \left(13\right)\end{array}$

With Equations (12) and (13),

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{\sigma }}_1=-{\lambda }_1\frac{{\sigma }_1}{{{\sigma }_1}^{1/2}}-{\lambda }_2{\sigma }_1{{\sigma }_1}^{P-1}+{\sigma }_2\\ {\stackrel{.}{\sigma }}_2=-{\alpha }_1\frac{{\sigma }_1}{{\sigma }_1}+{\ddot{X}}_{s3}^{*}\end{array}& \left(14\right)\end{array}$ is derived, where X*_s3 denotes the virtual controller, {dot over (X)}*_s3 denotes the first derivative of the virtual controller X*_s3, {umlaut over (X)}*_s3 denotes the second derivative of the virtual controller X*_s3, Y₁ is an estimate of the virtual controller X*_s3, Y₂ is an estimate of {dot over (X)}*₃, σ₁ and σ₂ are estimation errors, λ₁, λ₂, α₁>0 are system control gains, P>1 is a constant, {umlaut over (X)}*_s3 is supposed to be bounded and satisfies ∥{umlaut over (X)}*_s3∥≤L₃ with a known constant L₃>0, and if the parameters are selected to satisfy the conditions α₁>4L₃, λ₁>√{square root over (2α₁)}, the estimation errors σ₁, σ₂ will globally converge to the origin quickly, thus a precise difference value {dot over (X)}*_s3 of X*_s3 is obtained,

Letting X₃=X*_s4, X₄={dot over (X)}*_s4, σ₃=X₃−Y₃, σ₄=X₄−Y₄,

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_3=X_4\\ {\stackrel{.}{X}}_4={\ddot{X}}_{s4}^{*}\end{array}& \left(15\right)\end{array}$ $\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{Y}}_3={\lambda }_3\frac{{\sigma }_3}{{{\sigma }_3}^{1/2}}+{\lambda }_4{\sigma }_3{{\sigma }_3}^{P-1}+Y_4\\ {\stackrel{.}{Y}}_4={\alpha }_2\frac{{\sigma }_3}{{\sigma }_3}\end{array}& \left(16\right)\end{array}$ With Equations (15) and (16),

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{\sigma }}_3=-{\lambda }_3\frac{{\sigma }_3}{{{\sigma }_3}^{1/2}}-{\lambda }_4{\sigma }_3{{\sigma }_3}^{P-1}+{\sigma }_4\\ {\stackrel{.}{\sigma }}_4=-{\alpha }_2\frac{{\sigma }_3}{{\sigma }_3}+{\ddot{X}}_{s4}^{*}\end{array}& \left(17\right)\end{array}$

is derived, where X*_s4 denotes a virtual controller, {dot over (X)}*_s4 is the first derivative of the virtual controller X*_s4, {umlaut over (X)}*_s4 is the second derivative of the virtual controller X*_s4, Y₃ is an estimate of the virtual controller X*_s4, Y₄ is an estimate of {dot over (X)}*_s4, σ₃ and σ₄ are estimation errors, λ₃, λ₄, α₂>0 are system control gains, {umlaut over (X)}*_s4 supposed to be bounded and satisfies ∥{umlaut over (X)}*_s4∥≤L₄ with a known positive is constant L₄>0, and, if the parameters are selected to satisfy the conditions α₂>4L₄, λ₃>√{square root over (2α₂)}, the estimation errors σ₃, σ₄ will globally converge to the origin quickly, thus a precise difference value {dot over (X)}*_s4 of X*_s4 is obtained.

Preferably, a delay-dependent system stability criteria is established by constructing the Lyapunov Equations, thus a criteria for selecting controller parameters are provided. The detailed steps are as follows.

S1 selecting the first Lyapunov Equation as follows,

$$\begin{gathered} \begin{array}{cc}{V}_1=V_{11}+V_{12}+V_{13}\text{ \\ V1⁢1=Xm⁢2T⁢Mm(Xm⁢1)⁢Xm⁢2+kmks⁢Xs⁢2T⁢Ms(Xs⁢1)⁢Xs⁢2+2⁢(Um(Xm⁢1)-βm)+2⁢kmks⁢(Us(Xs⁢1)-βs)+2⁢∫0t(-Xm⁢2T(σ)⁢Fh(σ)+kmks⁢Xs⁢2T(σ)⁢Fe(σ))⁢d⁢σ⁢ \\ V1⁢2=km(Xm⁢1-Xs⁢1)T⁢(Xm⁢1-Xs⁢1)⁢ \\ V1⁢3=∫-T\_m0∫t+θtXm⁢2T(ξ)⁢Z⁢Xm⁢2(ξ)⁢d⁢ξ⁢d⁢θ+∫-T\_s0∫t+θtXs⁢2T(ξ)⁢P⁢Xs⁢2(ξ)⁢d⁢ξ⁢d⁢θ}& \left(18\right)\end{array} \end{gathered}$$ where the integral terms satisfy ∫₀^t−X_m2^T(σ)F_h(σ)dσ≥0, ∫₀^tX_s2^T(σ)F_e(σ)dσ≥0, Z and P are positive definite matrices, it is supposed that the time delay T_m(t) and T_s(t) are bounded, i.e. there are positive scalars T_m and T_s, such that T_m(t)≤T_m, T_s(t)≤T_s, M_m(X_m1) and M_s(X_s1) are the positive-definite inertia matrices of the master robot and the slave robot, respectively, U_m(X_m1) and U_s(X_s1) are the potential energy of the master robot and the slave robot satisfying

$G_m\left(X_{m1}\right)=\frac{\partial U_m\left(X_{m1}\right)}{\partial X_{m1}},G_s\left(X_{s1}\right)=\frac{\partial U_s\left(X_{s1}\right)}{\partial X_{s1}},$ there are positive scalars β_m and β_s such that U_m(X_m1)≤β_m, U_s(X_s1)≥β_s, and k_m, k_s>0 are proportional coefficients.

The time derivative of V₁ is given by

$\begin{array}{cc}{\stackrel{.}{V}}_1={\stackrel{.}{V}}_{11}+{\stackrel{.}{V}}_{12}+{\stackrel{.}{V}}_{13}\le X_{m2}^T\left(-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}\right)X_{m2}+X_{s2}^T\left(-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}\right)X_{s2}+2\frac{k_m}{k_s}{X}_{s2}^T{S}_s\left(X_{s3}-X_{s3}^{*}\right)& \left(19\right)\end{array}$ where I is the identity matrix, Z⁻¹ and P⁻¹ are the inverse matrices of positive definite matrices Z and P, respectively, S_s is a diagonal positive-definite constant matrix which contains the joint stiffness of the slave robot, α_m and α_s are damping coefficients which are positive constants, k_m, k_s>0 are proportional coefficients, it is supposed that the time delay T_m(t) and T_s(t) are bounded, i.e. there are positive scalars T_m and T_s, such that T_m(t)≤T_m, T_s(t)≤T_s, and X*_s3 is the first virtual controller.

S2 selecting the second Lyapunov Equation as follows:V₂=V₁+½(X_s3−X*_s3)^T(X_s3−X*_s3)  (20)

The time derivative of V₂ is given by

$\begin{array}{cc}{\stackrel{.}{V}}_2\le X_{m2}^T\left(-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}\right)X_{m2}+X_{s2}^T\left(-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}\right)X_{s2}+{\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s4}-X_{s4}^{*}\right)-{k_1\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s3}-X_{s3}^{*}\right)& \left(21\right)\end{array}$

S3 selecting the third Lyapunov Equation as follows:V₃=V₂+½(X_s4−X*_s4)^T(X_s4−X*_s4)  (22)

The time derivative of V₃ is given by

$\begin{array}{cc}{\stackrel{.}{V}}_3\le X_{m2}^T\left(-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}\right)X_{m2}+X_{s2}^T\left(-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}\right)X_{s2}-{k_1\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s3}-X_{s3}^{*}\right)-{k_2\left(X_{s4}-X_{s4}^{*}\right)}^T\left(X_{s4}-X_{s4}^{*}\right)& \left(23\right)\end{array}$

When the controller parameters are selected such that the following inequalities hold,

$-2{\alpha }_{m}I+{\overline{T}}_{m}Z+\overline{T_s}{k}_m^2{P}^{-1}<0,-2\frac{k_m{\alpha }_s}{k_s}I+\overline{T_s}P+{\overline{T}}_m{k}_m^2{Z}^{-1}<0$ the joint and motor velocities {dot over (q)}_m, {dot over (q)}_s, {dot over (θ)}_s and the position error q_m−q_s of the master-slave robot system with flexible joints are all bounded.

When a force F_h exerted by an operator to the master robot and a force F_e exerted by the external environment to the slave robot are both zero, the controllers are designed as follows:

$\begin{array}{cc}{\tau }_m=-k_m\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-{\alpha }_m{X}_{m2}+G_m\left(X_{m1}\right){\tau }_s=S_s\left(X_{s3}-X_{s1}\right)+J_s\left({\stackrel{.}{X}}_{s4}^{*}-\left(X_{s3}-X_{s3}^{*}\right)-k_2\left(X_{s4}-X_{s4}^{*}\right)\right)X_{s3}^{*}=X_{s1}+S_s^{-1}\left(k_s\left(X_{m1}\left(t-T_m\left(t\right)\right)-X_{s1}\right)-{\alpha }_s{X}_{s2}+G_s\left(X_{s1}\right)\right)X_{s4}^{*}={\stackrel{.}{X}}_{s3}^{*}-2\frac{k_m}{k_s}{S}_S^T{X}_{s2}-k_1\left(X_{s3}-X_{s3}^{*}\right)& \left(24\right)\end{array}$ where τ_m and τ_s are control torques provided by the controllers, X*_s3 and X*_s4 are virtual controllers, {dot over (X)}*_s3 and {dot over (X)}*_s4 are the first derivatives of the virtual controllers X*_s3 and X*_s4, respectively, T_m(t) and T_s(t) are forward time delay (from the master robot to the slave robot) and feedback time delay (from the slave robot to the master robot), respectively, α_m and α_s are damping coefficients which are positive constants, k_m, k_s>0 are proportional coefficients, S_s⁻¹ and S_s^T are the inverse matrix and the transpose matrix of a diagonal positive-definite constant matrix S_s which contains the joint stiffness of the slave robot, respectively, G_m(X_m1), G_s(X_s1) are the gravitational torques of the master robot and the slave robot, respectively, and k₁ and k₂ are selected to be positive constants.

When the controller parameters α_m, α_s, k_m, k_s, I, T_m, T_s, Z, P are selected such that the following conditions hold,

$-2{\alpha }_{m}I+{\overline{T}}_{m}Z+\overline{T_s}{k}_m^2{P}^{-1}<0,-2\frac{k_m{\alpha }_s}{k_s}I+\overline{T_s}P+{\overline{T}}_m{k}_m^2{Z}^{-1}<0$ it is guaranteed that the joint and motor velocities {dot over (q)}_m, {dot over (q)}_s, {dot over (θ)}_s and the position error q_m−q_s of the master-slave robot system with flexible joints will converge to zero asymptotically and that the global master-slave robot system with flexible joints will asymptotically become stable.

Compared with the existing methods, the present invention has the following advantages:

1. Compared with the local state control method, the proposed full-state control method can realize the global asymptotic stability of the master-slave system better.

2. The exact position tracking accuracy can be guaranteed by applying the full-state feedback controller based on the backstepping method globally and the robustness of closed-loop system also can be improved.

3. The precise difference to the virtual controllers can be realized by designing high-dimension uniform accurate differentiators. Additionally, this method is more applicable to the high-order control systems.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic control block diagram of the present invention.

DETAILED DESCRIPTION

The invention is further explained in combination with the attached FIGURE as follows.

As shown in FIG. 1, the steps of the invention method are as follows:

Step 1: connecting the master robot and the slave robot through the network to form a teleoperation system, measuring the system parameters of the master robot and the slave robot, and measuring a force exerted by an operator and a force exerted by the external environment by using force sensors.

The system parameters of the master robot and the slave robot includes the length and the mass of the manipulators, the positions and the positive-definite inertia matrices M_m(X_m1) and M_s(X_s1), the matrices of centripetal and coriolis torques C_m(q_m, {dot over (q)}_m) and C_s(q_s, {dot over (q)}_s), the gravity torques G_m(X_m1) and G_s(X_s1), the diagonal constant matrix J_s of the moment of actuator inertia of the master robot and the slave robot, and a diagonal positive-definite constant matrix S_s that contains the joint stiffness of the slave robot respectively according to the length and the mass of the manipulators. Moreover, a force F_h exerted by the operator to the master robot and a force F_e exerted by the external environment to the slave robot are measured by using force sensors.

The system dynamics equation can be described asM_m(q_m){umlaut over (q)}_m+C_m(q_m,{dot over (q)}_m){dot over (q)}_m+G_m(q_m)=τ_m+F_h M_s(q_s){umlaut over (q)}_s+C_s(q_s,{dot over (q)}_s){dot over (q)}_s+G_s(q_s)=S_s(θ_s−q_s)−F_e J_s{umlaut over (θ)}_s+S_s(θ_s−q_s)=τ_s  (1)where the subscripts m and s denotes the master robot and the slave robot, respectively. q_m, q_s∈Rⁿ are the vectors of joint displacements. {dot over (q)}_m, {dot over (q)}_s∈Rⁿ are the vectors of joint velocities. {umlaut over (q)}_m, {umlaut over (q)}_s∈Rⁿ are the vectors of joint accelerations. θ_s∈Rⁿ is the vector of motor displacements, {umlaut over (θ)}_s∈Rⁿ is the vector of motor accelerations. M_m(q_m), M_s(q_s)∈R^n×n are the positive-definite inertia matrices of the system, C_m(q_m, {dot over (q)}_m), C_s(q_s, {umlaut over (q)}_s)∈Rⁿ are the matrices of centripetal and coriolis torques, G_m(q_m), G_s(q_s)∈Rⁿ are the gravitational torques. J_s∈R^n×n is the diagonal constant matrix of the moments of actuator inertia. S_s∈Rⁿ is a diagonal positive-definite constant matrix that contains the joint stiffness of the slave robot. F_h, F_e∈Rⁿ are the force exerted by the operator to the master robot and the force exerted by the external environment to the slave robot, respectively. τ_m, τ_s∈Rⁿ are control torques provided by the controllers.

Step 2: measuring the position and velocity information of joints and motors in real-time, designing a proportional damping controller for the master robot based on position error and velocities, and designing a full-state feedback controller for the slave robot based on backstepping recursive technology in combination with Lyapunov equation.

For the master robot with rigid joints, letting X_m1=q_m, X_m2={dot over (q)}_m, the state space expression of the system is obtained as follows:

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_{m1}=X_{m2}\\ {\stackrel{.}{X}}_{m2}=M_m^{-1}\left(X_{m1}\right)\left({\tau }_m+F_h-C_m\left(X_{m1},X_{m2}\right)X_{m2}-G_m\left(X_{m1}\right)\right)\end{array}& \left(2\right)\end{array}$ where the subscripts m denotes the master robot, q_m∈Rⁿ is the vector of joint displacements, {dot over (q)}_m∈Rⁿ is the vector of joint velocities, M_m⁻¹(X_m1) is the inverse matrix of the positive-definite inertia matrix M_m(X_m1), τ_m∈Rⁿ is a control torque provided by the controller, C_m(q_m, {dot over (q)}_m)∈Rⁿ is the matrix of centripetal and coriolis torque, G_m(q_m)∈Rⁿ is the gravitational torque, and F_h∈Rⁿ is a force exerted by an operator to the master robot.

For the slave robot with flexible joints, letting X_s1=q_s, X_s2={dot over (q)}_s, X_s3=θ_s, X_s4={dot over (θ)}_s, the state space expression of the system is obtained as follows:

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_{s1}=X_{s2}\\ {\stackrel{.}{X}}_{s2}=M_s^{-1}\left(X_{s1}\right)\left(S_s\left(X_{s3}-X_{s1}\right)-F_e-C_s\left(X_{s1},X_{s2}\right)X_{s2}-G_s\left(X_{s1}\right)\right)\\ {\stackrel{.}{X}}_{s3}=X_{s4}\\ {\stackrel{.}{X}}_{s4}=J_s^{-1}\left({\tau }_s-S_s\left(X_{s3}-X_{s1}\right)\right)\end{array}& \left(3\right)\end{array}$ where the subscripts s denotes the master robot, q_s∈Rⁿ is the vector of joint displacements, {dot over (q)}_s∈Rⁿ is the vector of joint velocities, θ_s∈Rⁿ is the vector of motor displacements, {dot over (θ)}_s∈Rⁿ is the vector of motor velocities. M_s⁻¹(X_s1) is the inverse matrix of the positive-definite inertia matrix M_s(X_s1), J_s⁻¹ is the inverse matrix of the diagonal constant matrix J_s of the moments of actuator interia, τ_s∈Rⁿ is the control torque of the controller, C_s(q_s, {dot over (q)}_s)∈Rⁿ is the matrix of centripetal and coriolis torque, G_s(q_s)∈Rⁿ is the gravitational torque, F_e is a force exerted by the external environment to the slave robot, and, S_s∈R^n×n is a diagonal positive-definite constant matrix that contains the joint stiffness of the slave robot.

The first Lyapunov Equation is selected as follows,

$\begin{array}{cc}{V}_1=V_{11}+V_{12}+V_{13}{V}_{11}=X_{m2}^T{M}_m\left(X_{m1}\right)X_{m2}+\frac{k_m}{k_s}{X}_{s2}^T{M}_S\left(X_{s1}\right)X_{s2}+2\left(U_m\left(X_{m1}\right)-{\beta }_m\right)+\frac{2k_m}{k_s}\left(U_s\left(X_{s1}\right)-{\beta }_s\right)+2{\int }_0^t\left(-X_{m2}^T\left(\sigma \right)F_h\left(\sigma \right)+\frac{k_m}{k_s}{X}_{s2}^T\left(\sigma \right)F_e\left(\sigma \right)\right)d\sigma V_{12}={k_m\left(X_{m1}-X_{s1}\right)}^T\left(X_{m1}-X_{s1}\right)V_{13}={\int }_{-{\overline{T}}_m}^0{\int }_{t+\theta }^t{X}_{m2}^T\left(\xi \right)Z{X}_{m2}\left(\xi \right)d\xi d\theta +{\int }_{-\overline{T_s}}^0{\int }_{t+\theta }^t{X}_{s2}^T\left(\xi \right)P{X}_{s2}\left(\xi \right)d\xi d\theta & \left(4\right)\end{array}$ where the integral terms satisfy ∫₀^t−X_m2^T(σ)F_h(σ)dσ≥0, ∫₀^tX_s2^T(σ)F_e(σ)dσ≥0. Z and P are positive definite matrices. It is supposed that the time delay T_m(t) and T_s(t) are bounded, i.e. there are positive scalars T_m and T_s, such that T_m(t)≤T_m, T_s(t)≤T_s. M_m(M_m1) and M_s(X_s1) are positive-definite inertia matrices of the master robot and the slave robot, respectively.U_m(X_m1) and U_s(X_s1) are the potential energy of the master robot and the slave robot satisfying

$G_m\left(X_{m1}\right)=\frac{\partial U_m\left(X_{m1}\right)}{\partial X_{m1}},G_s\left(X_{s1}\right)=\frac{\partial U_s\left(X_{s1}\right)}{\partial X_{s1}}.$ There are positive scalars β_m and β_s such that U_m(X_m1)≥β_m, U_s(X_s1)≥β_s, and k_m, k_s>0 are proportional coefficients.

The time derivatives of V₁, V₁₁, V₁₂, V₁₃ are given by

$\begin{array}{cc}\begin{array}{c}{\stackrel{.}{V}}_{11}=2X_{m2}^T\left(F_h-k_m\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-{\alpha }_m{X}_{m2}-C_m\left(X_{m1},X_{m2}\right)X_{m2}-G_m\left(X_{m1}\right)\right)\\ +2\frac{k_m}{k_s}{X}_{s2}^T\left(s,\left(X_{s3}-X_{s1}\right)-F_e-C_s\left(X_{s1},X_{s2}\right)X_{s2}-G_s\left(X_{s1}\right)\right)+\frac{2k_m}{k_s}{G}_s\left(X_{s1}\right)X_{s2}^T\\ +X_{m2}^T{\stackrel{.}{M}}_m\left(X_{m1}\right)X_{m2}+\frac{k_m}{k_s}{X}_{s2}^T{\stackrel{.}{M}}_s\left(X_{s1}\right)X_{s2}+2G_m\left(X_{m1}\right)X_{m2}^T-2X_{m2}^T{F}_h+2\frac{k_m}{k_s}{X}_{s2}^T{F}_e\\ =-2k_m{X}_{m2}^T\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-2{\alpha }_m{X}_{m2}^T{X}_{m2}+2\frac{k_m}{k_s}{X}_{s2}^T{S}_s\left(X_{s3}-X_{s1}\right)\\ =-2k_m{X}_{m2}^T\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-2{\alpha }_m{X}_{m2}^T{X}_{m2}+2\frac{k_m}{k_s}{X}_{s2}^T\left(S_s\left(X_{s3}-X_{s3}^{*}\right)+S_s{X}_{s3}^{*}-S_s{X}_{s1}\right)\end{array}& \left(5\right)\end{array}$

With Equation (5), the first virtual controller is derived as X*_s3=X_s1±S_s⁻¹(k_s(X_m1(t−T_m(t))−X_s1)−α_sX_s2). By substituting the first virtual controller X*_s3 into Equation (5),

$\begin{array}{cc}\begin{array}{c}{\stackrel{.}{V}}_{11}=-2k_m{X}_{m2}^T\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-2{\alpha }_m{X}_{m2}^T{X}_{m2}\\ +2\frac{k_m}{k_s}{X}_{s2}^T\left(S_s\left(X_{s3}-X_{s3}^{*}\right)+k_s\left(X_{m1}\left(t-T_m\left(t\right)\right)-X_{s1}\right)-{\alpha }_s{X}_{s2}\right)\\ =-2{\alpha }_m{X}_{m2}^T{X}_{m2}-2\frac{k_m{\alpha }_s}{k_s}{X}_{s2}^T{X}_{\mathrm{s2}}-2k_m{X}_{m2}^T\left(X_{m1}-X_{s1}\right)-2k_m{X}_{m2}^T\left(X_{s1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)\\ -2k_m{X}_{s2}^T\left(X_{s1}-X_{m1}\right)-2k_m{X}_{s2}^T\left(X_{m1}-X_{m1}\left(t-T_m\left(t\right)\right)\right)+2\frac{k_m}{k_s}{X}_{s2}^T{S}_s\left(X_{s3}-X_{s3}^{*}\right)\\ =-2{\alpha }_m{X}_{m2}^T{X}_{m2}-2{k_m\left(X_{m1}-X_{s1}\right)}^T\left(X_{m2}-X_{s2}\right)-2\frac{k_m{\alpha }_s}{k_s}{X}_{s2}^T{X}_{s2}-2k_m{X}_{m2}^T{\int }_{t-T_s\left(t\right)}^t{X}_{s2}\left(\xi \right)d\xi \\ -2k_m{X}_{s2}^T{\int }_{t-T_m\left(t\right)}^t{X}_{m2}\left(\xi \right)d\xi +2\frac{k_m}{k_s}{X}_{s2}^T{S}_s\left(X_{s3}-X_{s3}^{*}\right)\end{array}{\stackrel{.}{V}}_{12}=2{k_m\left(X_{m2}-X_{s2}\right)}^T\left(X_{m1}-X_{s1}\right)\begin{array}{c}{\stackrel{.}{V}}_{13}={\overline{T}}_m{X}_{m2}^{T}Z{X}_{m2}-{\int }_{t-{\overline{T}}_m}^t{X}_{m2}^T\left(\xi \right)Z{X}_{m2}\left(\xi \right)d\xi +\overline{T_s}{X}_{s2}^{T}P{X}_{s2}-{\int }_{t-{\overline{T}}_s}^t{X}_{s2}^T\left(\xi \right)P{X}_{s2}\left(\xi \right)d\xi \\ \le {\overline{T}}_m{X}_{m2}^{T}Z{X}_{m2}-{\int }_{t-T_m\left(t\right)}^t{X}_{m2}^T\left(\xi \right)Z{X}_{m2}\left(\xi \right)d\xi +\overline{T_s}{X}_{s2}^{T}P{X}_{s2}-{\int }_{t-T_s\left(t\right)}^t{X}_{s2}^T\left(\xi \right)P{X}_{s2}\left(\xi \right)d\xi \end{array}& \left(6\right)\end{array}$ is obtained. Furthermore, with the inequality, it holds that

$\begin{array}{cc}-2k_m{X}_{m2}^T{\int }_{t-T_s\left(t\right)}^t{X}_{s2}\left(\xi \right)d\xi -{\int }_{t-T_s\left(t\right)}^t{X}_{s2}^T\left(\xi \right)P{X}_{s2}\left(\xi \right)d\xi \le \overline{T_s}{k}_m^2{X}_{m2}^T\left(t\right)P^{-1}{X}_{m2}\left(t\right)-2k_m{X}_{s2}^T{\int }_{t-T_m\left(t\right)}^t{X}_{m2}\left(\xi \right)d\xi -{\int }_{t-T_m\left(t\right)}^t{X}_{m2}^T\left(\xi \right)Z{X}_{m2}\left(\xi \right)d\xi \le {\stackrel{\_}{T}}_m{k}_m^2{X}_{s2}^T\left(t\right)Z^{-1}{X}_{s2}\left(t\right)\mathrm{So}& \left(7\right)\\ \begin{array}{c}{\stackrel{.}{V}}_1={\stackrel{.}{V}}_{11}+{\stackrel{.}{V}}_{12}+{\stackrel{.}{V}}_{13}\le X_{m2}^T\left(-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+\overline{T_s}{k}_m^2{P}^{-1}\right)X_{m2}\\ +X_{s2}^T\left(-2\frac{k_m{\alpha }_s}{k_s}I+\overline{T_s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}\right)X_{s2}+2\frac{k_m}{k_s}{X}_{s2}^T{S}_s\left(X_{s3}-X_{s3}^{*}\right)\end{array}& \left(8\right)\end{array}$ is derived, where I is the identity matrix, z⁻¹ and P⁻¹ are the inverse matrices of positive definite matrices Z and P, respectively. S_s is a diagonal positive-definite constant matrix which contains the joint stiffness of the slave robot. α_m and α_s are damping coefficients which are positive constants. k_m, k_s>0 are proportional coefficients. It is supposed that the time delay T_m(t) and T_s(t) are bounded, i.e. there are positive scalars T_m and T_s, such that T_m(t)≤T_m, T_s(t)≤T_s, X*_s3 is the first virtual controller.

The second Lyapunov Equation is selected as follows,V₂=V₁+½(X_s3−X*_s3)^T(X_s3−X*_s3)  (9)

The time derivative of V₂ is given by{dot over (V)}₂={dot over (V)}₁+(X_s3−X*_s3)^T(X_s4−{dot over (X)}*_s3)={dot over (V)}₁+(X_s3−X*_s3)^T(X_s4−X*_s4+X*_s4−{dot over (X)}*_s3)  (10)With equation (10), the second virtual controller is derived as

$X_{s4}^{*}={\stackrel{.}{X}}_{s3}^{*}-2\frac{k_m}{k_s}{S}_s^T{X}_{s2}-k_1\left(X_{s3}-X_{s3}^{*}\right).$ By substituting the second virtual controller X*_s4 into Equation (10),

$\begin{array}{cc}{\stackrel{.}{V}}_2\le X_{m2}^T\left(-2{\alpha }_{m}I+{\overline{T}}_{m}Z+\overline{T_s}{k}_m^2{P}^{-1}\right)X_{m2}+X_{s2}^T\left(-2\frac{k_m{\alpha }_s}{k_s}I+\overline{T_s}P+{\overline{T}}_m{k}_m^2{Z}^{-1}\right)X_{s2}+{\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s4}-X_{s4}^{*}\right)-{k_1\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s3}-X_{s3}^{*}\right)& \left(11\right)\end{array}$ is derived.

The third Lyapunov Equation is selected as follows,V₃=V₂+½(X_s4−X*_s4)^T(X_s4−X*_s4)  (12)

The time derivative of V₃ is given by

$\begin{array}{cc}\begin{array}{c}{\stackrel{.}{V}}_3={\stackrel{.}{V}}_2+{\left(X_{s4}-X_{s4}^{*}\right)}^T\left({\stackrel{.}{X}}_{s4}-{\stackrel{.}{X}}_{s4}\right)\\ ={\stackrel{.}{V}}_2+{\left(X_{s4}-X_{s4}^{*}\right)}^T\left(J_s^{-1}\left({\tau }_s-S_s\left(X_{s3}-X_{s1}\right)\right)-X_{s4}^{*}\right)\end{array}& \left(13\right)\end{array}$

The full-state feedback controller is derived from Equation (13) as

τ_s=S_s(X_s3−X_s1)+J_s({dot over (X)}*_s4−(X_s3−X*_s3)−k₂(X_s4−X*_s4)). By substituting the obtained all-state feedback controller τ_s into Equation (13),

$\begin{array}{cc}{\stackrel{.}{V}}_3\le X_{m2}^T\left(-2{\alpha }_{m}I+{\overline{T}}_{m}Z+\overline{T_s}{k}_m^2{P}^{-1}\right)X_{m2}+X_{s2}^T\left(-2\frac{k_m{\alpha }_s}{k_s}I+\overline{T_s}P+{\overline{T}}_m{k}_m^2{Z}^{-1}\right)X_{s2}-{k_1\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s3}-X_{s3}^{*}\right)-{k_2\left(X_{s4}-X_{s4}^{*}\right)}^T\left(X_{s4}-X_{s4}^{*}\right)& \left(14\right)\end{array}$ is derived.

The controllers of the master-slave robot system with flexible joints are obtained by using the backstepping method as

$\begin{array}{cc}{\tau }_m=-k_m\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-{\alpha }_m{X}_{m2}{\tau }_s=S_s\left(X_{s3}-X_{s1}\right)+J_s\left({\stackrel{.}{X}}_{s4}^{*}-\left(X_{s3}-X_{s3}^{*}\right)-k_2\left(X_{s4}-X_{s4}^{*}\right)\right)X_{s3}^{*}=X_{s1}+S_s^{-1}\left(k_s\left(X_{m1}\left(t-T_m\left(t\right)\right)-X_{s1}\right)-{\alpha }_s{X}_{s2}\right)X_{s4}^{*}={\stackrel{.}{X}}_{s3}^{*}-2\frac{k_m}{k_s}{S}_s^T{X}_{s2}-k_1\left(X_{s3}-X_{s3}^{*}\right)& \left(15\right)\end{array}$ where τ_m and τ_s are control torques provided by the controllers, X*_s3 and X*_s4 are the first virtual controller and the second virtual controller, respectively, {dot over (X)}*_s3 and {dot over (X)}*_s4 are the first derivatives of the virtual controllers X*_s3 and X*_s4, respectively, T_m(t) and T_s(t) are forward time delay (from the master robot to the slave robot) and backward time delay (from the slave robot to the master robot), respectively. α_m and α_s are damping coefficients which are positive constants. k_m, k_s>0 are proportional coefficients. S_s⁻¹ and S_s^T are the inverse matrix and the transpose matrix of a diagonal positive-definite constant matrix S_s which contains the joint stiffness of the slave robot, respectively. J_s is the diagonal constant matrix of the moments of actuator inertia. k₁ and k₂ are selected to be positive constants.

Step 3: designing high-dimensional uniform accurate differentiators to carry out a precise difference to the first virtual controller and the second virtual controller.

Letting X₁=X*_s3, X₂={dot over (X)}*_s3, σ₁=X₁−Y₁, σ₂=X₂−Y₂,

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_1=X_2\\ {\stackrel{.}{X}}_2={\ddot{X}}_{s3}^{*}\end{array}& \left(16\right)\\ \{\begin{array}{c}{\stackrel{.}{Y}}_1={\lambda }_1\frac{{\sigma }_1}{{{\sigma }_1}^{1/2}}+{\lambda }_2{\sigma }_1{{\sigma }_1}^{P-1}+Y_2\\ {\stackrel{.}{Y}}_2={\alpha }_1\frac{{\sigma }_1}{{\sigma }_1}\end{array}& \left(17\right)\end{array}$ with Equations (16) and (17),

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{\sigma }}_1=-{\lambda }_1\frac{{\sigma }_1}{{{\sigma }_1}^{1/2}}-{\lambda }_2{\sigma }_1{{\sigma }_1}^{P-1}+{\sigma }_2\\ {\stackrel{.}{\sigma }}_2=-{\alpha }_1\frac{{\sigma }_1}{{\sigma }_1}+{\ddot{X}}_{s3}^{*}\end{array}& \left(18\right)\end{array}$ is derived, where X*_s3 denotes the first virtual controller, {dot over (X)}*_s3 denotes the first derivative of the first virtual controller X*_s3, {umlaut over (X)}_s3 denotes the second derivative of the first virtual controller X*_s3, Y₁ is an estimate of the first virtual controller X*_s3, Y₂ is an estimate of {dot over (X)}*_s3, σ₁ and σ₂ are estimation errors, λ₁, λ₂, α₁>0 are system control gains, P>1 is a constant, {umlaut over (X)}*_s3 is supposed to be bounded and satisfies ∥{umlaut over (X)}*_s3∥≤L₃ with a known constant L₃>0. If the parameters are selected to satisfy the conditions α₁>4L₃, λ₁>√{square root over (2α₁)}, the estimation errors σ₁, σ₂ will converge to the origin quickly, thus a precise difference value {dot over (X)}*_s3 of X*_s3 is obtained.

Letting X₃=X*_s4, X₄={dot over (X)}*_s4, σ₃=X₃−Y₃, σ₄=X₄−Y₄,

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{X}}_3=X_4\\ {\stackrel{.}{X}}_4={\ddot{X}}_{s4}^{*}\end{array}& \left(19\right)\\ \{\begin{array}{c}{\stackrel{.}{Y}}_3={\lambda }_3\frac{{\sigma }_3}{{{\sigma }_3}^{1/2}}+{\lambda }_4{\sigma }_3{{\sigma }_3}^{P-1}+Y_4\\ {\stackrel{.}{Y}}_4={\alpha }_2\frac{{\sigma }_3}{{\sigma }_3}\end{array}& \left(20\right)\end{array}$ with Equations (19) and (20),

$\begin{array}{cc}\{\begin{array}{c}{\stackrel{.}{\sigma }}_3=-{\lambda }_3\frac{{\sigma }_3}{{{\sigma }_3}^{1/2}}-{\lambda }_4{\sigma }_3{{\sigma }_3}^{P-1}+{\sigma }_4\\ {\stackrel{.}{\sigma }}_4=-{\alpha }_2\frac{{\sigma }_3}{{\sigma }_3}+{\ddot{X}}_{s4}^{*}\end{array}& \left(21\right)\end{array}$ is derived, where X*_s4 denotes the second virtual controller, {dot over (X)}*_s4 denotes the first derivative of the second virtual controller X*_s4, {umlaut over (X)}*_s4 denotes the second derivative of the second virtual controller X*_s4, Y₃ is an estimate of the second virtual controller X*_s4, Y₄ is an estimate of {dot over (X)}*_s4, σ₃ and σ₄ are estimation errors, λ₃, λ₄, α₂>0 are system control gains, {umlaut over (X)}*_s4 is supposed to be bounded and satisfies ∥{umlaut over (X)}*_s4∥≤L₄ with a known positive constant L₄>0. If the parameters are selected to satisfy the conditions α₂>4L₄, λ₃>√{square root over (2α₂)}, the estimation errors σ₃, σ₄ will converge to the origin quickly, thus a precise difference value {dot over (X)}*_s4 of X*_s4 is obtained.

Step 4: Establishing the delay-dependent system stability criteria by constructing Lyapunov Equations, providing the criteria for selecting controller parameters, and realizing the global stability of the master-slave robot system with flexible joints and time-varying delays.

The Lyapunov Equation is selected as

$\begin{array}{cc}{V}_{11}=X_{m2}^T{M}_m\left(X_{m1}\right)X_{m2}+\frac{k_m}{k_s}{X}_{s2}^T{M}_s\left(X_{s1}\right)X_{s2}+2\left(U_m\left(X_{m1}\right)-{\beta }_m\right)+\frac{2k_m}{k_s}\left(U_s\left(X_{s1}\right)-{\beta }_s\right)+2{\int }_0^t\left(-X_{m2}^T\left(\sigma \right)F_h\left(\sigma \right)+\frac{k_m}{k_s}{X}_{s2}^T\left(\sigma \right)F_e\left(\sigma \right)\right)d\sigma V_{12}={k_m\left(X_{m1}-X_{s1}\right)}^T\left(X_{m1}-X_{s1}\right)V_{13}={\int }_{-{\stackrel{\_}{T}}_m}^0{\int }_{t+\theta }^t{X}_{m2}^T\left(\xi \right)Z{X}_{m2}\left(\xi \right)d\xi d\theta +{\int }_{-{\stackrel{\_}{T}}_s}^0{\int }_{t+\theta }^t{X}_{s2}^T\left(\xi \right)P{X}_{s2}\left(\xi \right)d\mathrm{\xi d\theta }{V}_1=V_{11}+V_{12}+V_{13}{V}_2=V_1+\frac{1}{2}{\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s3}-X_{s3}^{*}\right)V_3=V_2+\frac{1}{2}{\left(X_{s4}-X_{s4}^{*}\right)}^T\left(X_{s4}-X_{s4}^{*}\right)& \left(22\right)\end{array}$

The time derivative of V₃ is given by

$\begin{array}{cc}{\stackrel{.}{V}}_3\le X_{m2}^T\left(-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}\right)X_{m2}+X_{s2}^T\left(-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}\right)X_{s2}-{k_1\left(X_{s3}-X_{s3}^{*}\right)}^T\left(X_{s3}-X_{s3}^{*}\right)-{k_2\left(X_{s4}-X_{s4}^{*}\right)}^T\left(X_{s4}-X_{s4}^{*}\right)& \left(23\right)\end{array}$

When the controller parameters α_m, α_s, k_m, k_s, I, T_m, T_s, Z, P are selected such that the following conditions hold,

$-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}<0,-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}<0$ the joint and motor velocities {dot over (q)}_m, {dot over (q)}_s, {dot over (θ)}_s and position error q_m−q_s of the master-slave robot system with flexible joints are all bounded.

If the force F_h exerted by an operator to the master robot and a force F_e exerted by the external environment to the slave robot are both zero, the controllers are designed as follows:

$\begin{array}{cc}{\tau }_m=-k_m\left(X_{m1}-X_{s1}\left(t-T_s\left(t\right)\right)\right)-{\alpha }_m{X}_{m2}+G_m\left(X_{m1}\right){\tau }_s=S_s\left(X_{s3}-X_{s1}\right)+J_s\left({\stackrel{.}{X}}_{s4}^{*}-\left(X_{s3}-X_{s3}^{*}\right)-k_2\left(X_{s4}-X_{s4}^{*}\right)\right)X_{s3}^{*}=X_{s1}+S_s^{-1}\left(k_s\left(X_{m1}\left(t-T_m\left(t\right)\right)-X_{s1}\right)-{\alpha }_s{X}_{s2}+G_s\left(X_{s1}\right)\right)X_{s4}^{*}={\stackrel{.}{X}}_{s3}^{*}-2\frac{k_m}{k_s}{S}_s^T{X}_{s2}-k_1\left(X_{s3}-X_{s3}^{*}\right)& \left(24\right)\end{array}$ where X*_s3 and X*_s4 are the first controller and the second virtual controller, respectively, {dot over (X)}*_s3 and {dot over (X)}*_s4 are the first derivatives of the virtual controllers X*_s3 and X*_s4, respectively, T_m(t) and T_s(t) are forward time delay (from the master robot to the slave robot) and backward time delay (from the slave robot to the master robot), respectively, α_m and α_s are damping coefficients which are positive constants, k_m, k_s>0 are proportional coefficients, S_s⁻¹ and S_s^T are the inverse matrix and the transpose matrix of a diagonal positive-definite constant matrix S_s which contains the joint stiffness of the slave robot, respectively, G_m(X_m1), G_s(X_s1) are the gravity torques of the master robot and the slave robot, J_s is a diagonal constant matrix of the moment of actuator inertia, and k₁ and k₂ are selected to be positive constants.

When the controller parameters α_m, α_s, k_m, k_s, I, T_m, T_s, Z, P are selected such that the following conditions hold,

$-2{\alpha }_{m}I+{\stackrel{\_}{T}}_{m}Z+{\stackrel{\_}{T}}_s{k}_m^2{P}^{-1}<0,-2\frac{k_m{\alpha }_s}{k_s}I+{\stackrel{\_}{T}}_{s}P+{\stackrel{\_}{T}}_m{k}_m^2{Z}^{-1}<0$ it can be guaranteed that the joint and motor velocities {dot over (q)}_m, {dot over (q)}_s, {dot over (θ)}_s and the position error q_m−q_s of the master-slave robot system with flexible joints will converge to zero asymptotically and that the global master-slave robot system with flexible joints will asymptotically become stable.

For a master-slave robot system with flexible joints, first a control command is sent by the operator to the master robot, and the master robot is controlled by the proportional damping controller τ_m and transmits the received control command to the slave robot through the network information transmission channel. Then the slave robot controlled by the full-state feedback controller τ_s executes the control command on the external environment, feedbacks the measured position and force information to the operator in time, thus a closed-loop system is formed and tasks can be completed effectively.

Based on the backstepping technique and the high-dimension uniform accurate differentiators, a full-state feedback controller is proposed for a flexible master-slave robot system. With this flexible master-slave robot system, accurate position tracking performance is achieved in the global scope. Additionally, the global asymptotic convergence is guaranteed and the robustness of the closed-loop master-slave system is improved.

The above-mentioned embodiment merely describes the preferred embodiment of the present invention. The scope of the invention should not be limited by the described embodiment. It is intended that various changes and modifications to the technical solutions of the present invention made by those skilled in the art without departing from the spirit of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A full-state control method for a master-slave robot system with flexible joints and time-varying delays, comprising: connecting a master robot and a slave robot through a network to form a teleoperation system;measuring system parameters of the master robot and the slave robot, and measuring position and velocity information of joints and motors of the master robot and the slave robot in real-time;controlling the master robot with a proportional damping controller and controlling the slave robot with a full-state feedback controller,wherein a control torque provided by the proportional damping controller and a control torque provided by the full-state feedback controller are determined according to following formulas:

2. The full-state control method of the master-slave robot system with flexible joints and time-varying delays according to claim 1, wherein the system parameters of the master robot and the slave robot comprises: length and mass of manipulators of the master robot and the slave robot, positions and positive-definite inertia matrices Mm(Xm1) and Ms(Xs1) matrices of centripetal and coriolis torques Cm(qm, {dot over (q)}m) and Cs(qs, {dot over (q)}s) gravity torques Gm(Xm1) and Gs(Xs1), diagonal constant matrix Js of moment of actuator inertia of the master robot and the slave robot, and a diagonal positive-definite constant matrix Ss that contains joint stiffness of the slave robot.

3. The full-state control method of the master-slave robot system with flexible joints and time-varying delays according to claim 2, wherein the first virtual controller X*s3 and the second virtual controller X*s4 are determined according to following formulas:

4. The full-state control method of the master-slave robot system with flexible joints and time-varying delays according to claim 3, wherein {dot over (X)}*s3 is determined according to following formulas: X1=X*s3,X2={dot over (X)}*s3,σ1=X1−Y1,σ2=X2−Y2,

5. The full-state control method of the master-slave robot system with flexible joints and time-varying delays according to claim 3, wherein {dot over (X)}*s4 is determined according to following formulas: X3=X*s4,X4={dot over (X)}*s4,σ3=X3−Y3,σ4=X4−Y4,

6. The full-state control method of the master-slave robot system with flexible joints and time-varying delays according to claim 3, wherein parameters αm, αs, km, ks, I, Tm, Ts, Z, P of the controllers are selected such that the following inequalities hold,