• 대한전기학회
Mobile QR Code QR CODE : The Transactions of the Korean Institute of Electrical Engineers
  • COPE
  • kcse
  • 한국과학기술단체총연합회
  • 한국학술지인용색인
  • Scopus
  • crossref
  • orcid

  1. (Dept. of Control & Robot Engineering, Gyeongsang National University, Korea.)



terminal sliding mode control, variable structure system, proof of Ukin's Theorem, diagonalization method, transformation method

1. Introduction

The sliding mode control(SMC) can be divided into the two parts:. linear sliding mode control(LSMC)(1)(2) and terminal sliding mode control(TSMC)(3)(4).

In the LSMC, the variable structure system(VSS) with the SMC can provide the effective means to control of uncertain systems under parameter variations and external disturbances(5). One of its essential advantages is the robustness of the controlled system against matched parameter uncertainties and external disturbances in the sliding mode on the predetermined sliding surface. In order to fully utilize the advantages of the sliding mode on the predetermined sliding surface, the precise existence condition of the sliding mode should be satisfied and proved completely for the complete formulation of the design of the LSMC(11)(12) as well as the TSMC. Utkin presented the two methodologies to prove the existence condition of the sliding mode on the sliding surface(1). These are known as the invariance theorem, where the equation of the sliding mode is invariant with respect to the two nonlinear transformations: the control input transformation and sliding surface transformation. These methods are also called the diagonalization methods. Those were reviewed in (2). DeCarlo, Zak, and Matthews tried to prove Utkin's invariance theorem. But, the proof is not clear. In (5), Su, Drakunov, and Ozguner mentioned the sliding surface transformation, which would diagonalize the control coefficient matrix to the dynamics for the sliding surface. But they did not prove the existence condition of the sliding mode on the predetermined sliding surface. However the proofs of Utkin’s theorem are given in (6) for single input(SI) uncertain linear plants, in (10) for SI uncertain integral linear plants without the reaching phase problems, in (14) for multi input(MI) uncertain linear plants, in (15) for MI uncertain integral linear plants, in (16) for SI uncertain nonlinear plants, and in (17) for integral augmented uncertain nonlinear systems without the reaching phase problems(11)(12). The researches on the proof of Utkin’s theorem is sufficient to prove Utkin’s theorem for various uncertain plants in the LSMC. However, in the TSMC, the proof of Utkin’s theorem is rare except (32).

In the TSMC, for the first time, Haimo developed a finite time controller with a finite time stabilization in 1986(3). Haimo invented the finite time stabilization rather than the asymptotic convergence of the LSMC. The TSMC has the advantages over the LSMC, for example, convergence in finite time and high control precision. Zak presented the terminal attractors with the finite time convergence in 1988(4). After those, a lot of the researches on the TSMCs for example many theoretical developments and application examples are reported until now(18)-(32). In (18), Zhihong et. al. studied the TSMC which is applied to the control of multi-input multi-output(MIMO) robot manipulators. Zhihong and Yu reported the TSMCs for higher order single-input single-output(SISO) linear systems with the hierarchical terminal sliding surface and regular MIMO systems with the fractional order sliding surface in (19). In 1997, Yu et. al. suggested a TSMC having the fast transient performance with the recursive sliding surface for higher order SISO systems in (20). The acceleration term is added to the terminal sliding manifold in order to speed up the output response. Yu et. al suggested a nonsingular terminal sliding mode control of a class of nonlinear dynamical systems in (21). The conventional TSMCs until now have the singularity problem that the control input becomes infinity in certain domain. However, there is no singular problem in (21). Feng et. al investigated a second order TSMC for uncertain multivariable systems for chattering-free performance and nonsingularity in (22). Zong et. al proposed a higher order sliding mode control with self-tuning law based on the integral sliding mode when the uncertainty in the input matrix is not the zero that is $\triangle b\ne 0$ in (23). The method can be viewed as the finite stabilization based on the higher sliding mode with the geometric homogeneity and the integral sliding surface with no reaching phase. A derivative and integral TSMC for a class of MIMO nonlinear systems is suggested by Chiu in (24). The recursive sign or fractional integral terminal sliding manifolds are proposed to remove the reaching phase. Pen et. al in 2015 designed an integral terminal sliding surface for uncertain nonlinear systems without the singularity by means of the saturation on the singular component of the control for temporary avoiding the singularity in (25). For the noncanonical plants of the interceptors, a fast robust guidance and control is designed based on a fast fractional integral terminal sliding surface for removing the reaching phase in (26). Hu et. al analyzed a dynamic sliding mode manifold based continuous fractional order nonsingular terminal sliding mode control for a class of second order nonlinear systems when in 2020 and in (27). In the algorithm, the uncertainty term in the input matrix is treated as the total lumped uncertainty. To cope with the reaching phase, the time varying sliding hyperplanes are proposed in (28). But the real output is not predicted. A study of nonsingular fast terminal sliding mode fault tolerant control based on the nonsingular sliding surface is presented by Xu et. al in 2015 and in (30). In (31), the discontinuous and continuous control input transformation integral TSMCs by using the integral sliding surface without the reaching phase and with the output prediction performance as the one approach are presented for second order uncertain plants when $\triangle b\ne 0$. Applying the idea of (9) for the LSMC, the integral sliding surface without the reaching phase is suggested for the TSMCs. The exponent of the power function can be any positive numbers satisfying $q>p>0$ such that $0<p/q<1$. The ideal sliding dynamics of the integral sliding surface is derived and the real robust output can be predicted, predesigned, predetermined by means of the solution of the ideal sliding dynamics. Based on defining a new auxiliary nonlinear state, the closed loop exponential stability together with the existence condition of the sliding mode on the predetermined sliding surface is investigated theoretically for the complete formulation of the TSMC design for the output prediction performance. As a remedy of the singularity, a certain limit is imposed on the new auxiliary nonlinear state. In (32), the proof of Utkin’s theorem is given for second order SI non-integral uncertain linear plants when $\triangle b\ne 0$. About only the two transformations, the control input one and sliding surface one, Utkin’s theorem is proved rigorously in the TSMCs.

The researches on the proof of Utkin’s theorem in the TSMC are rare except (32). Only the design approaches of the control input transformation SMCs are in (24)-(26)(27)(29)-(31). In (31), the algorithm is first called as the control input transformation SMC. About the sliding surface part transformation, the research does not exist. The integral action is augmented to the TSMC for removing the reaching phase in (31). To the integral TSMC systems, the proof of Utkin theorem is essentially needed.

In this paper, a complete proof of Utkin's theorem is presented for the ITSMC of second-order SI uncertain linear systems when $\triangle b\ne 0$. The paper presents five approaches for designing ITSMCs for second-order uncertain linear plants, namely: control input transformation, sliding surface full transformation, and three sliding surface part transformations. The invariance theorem is demonstrated clearly and comparatively for the first two transformation methods, which are known as the two diagonalization methods, in the context of second-order SI integral uncertain linear systems when $\triangle b\ne 0$. Although the control input transformation has been previously discussed in works such as (31) and (32), it is explicitly named here for the first time. Similarly, the sliding surface transformation is introduced for the first time in the TSMC, except for (32), and the sliding surface part transformations make their debut in the TSMC through this research. This paper covers the first three transformations, and further studies will include the last two sliding surface part transformations. The output prediction, predetermination, and predesign capabilities of the first three transformations yield the same performance. The design examples and simulation studies are provided to illustrate the practical value of the main results.

2. Main Results of Proof of Utkin's Theorem for TSMCs

The invariant theorem of Utkin is as follows(1)(2):

Theorem 1: The equation of the sliding mode is invariant with respect to the two nonlinear transformations, i.e. the control input transformation and sliding surface transformation:

(1)
\begin{align*} u*(x)=H_{u}(x,\:t)· u(x)\\ s*(x)=H_{s}(x,\:t)· s(x) \end{align*}

for $\det H_{u}\ne 0{and}\det H_{s}\ne 0$.

For a second order SI uncertain canonical linear system(31):

(2)
\begin{align*} \dot x_{1}=x_{2}\\ \dot x_{2}=(a_{10}+\triangle a_{1})x_{1}+(a_{20}+\triangle a_{2})x_{2}\\ +(b_{0}+\triangle b)u+\triangle d(x,\:t) \end{align*}

where $x_{1}\in R \quad{and}\quad x_{2}\in R$ are the state variables, $u\in R$ is the control input, $a_{10},\:a_{20},\: {and} \quad b_{0}\in R$ are the nominal values, $\triangle a_{1},\: \triangle a_{2},\: {and} \quad \triangle b$ are the uncertainties, those are assumed to be matched and bounded, and $\Delta d(x,\:t)$ is the external disturbance which is also assumed to be matched and bounded.

Assumption 1:

$(b_{0})^{-1}\triangle b=\triangle b(b_{0})^{-1}=\triangle I$, and $\vert\triangle I |\le\rho <1$ where $\rho$ is a positive constant less than 1.

An integral state $x_{0}\in R$ with a special initial condition is augmented for use later in the integral terminal sliding surface as follows:

(3)

$x_{0}(t)=\int_{0}^{t}x_{1}(\tau)d\tau +\int_{-\infty}^{0}x_{1}(\tau)d\tau$

$=\int_{0}^{t}x_{1}(\tau)d\tau +x_{0}(0)$

where $x_{0}(0)$ is the special initial condition for the integral state which is determined later.

Based on the idea of (9) of the LSMC, for removing the reaching phase completely, the integral terminal sliding surface for the TSMC $s\in R$ is proposed as follows(31):

(4)
$s=C_{0}· x_{0}+C_{1}· x_{1}^{p/q}+x_{2}(=0)$

where $p \quad {and} \quad q$ are any positive numbers satisfying $q>p>0$ such that $p/q$ is real fractional that is $0<p/q<1$, in which any positive number such that $0<p/q<1$ are first mentioned except (31) and (32). The $C_{0}$ and $C_{1}$ are designed such that the polynomial $r^{2}+C_{1}r+C_{0}=0$ should be Hurwitz. The special initial condition $x_{0}(0)$ in Eq. (3) for the integral state is determined so that the integral terminal sliding surface Eq. (4) is the zero at $t=0$ for any initial condition $x_{1}(0) \quad {and} \quad x_{2}(0)$ as

(5)
$x_{0}(0)=-C_{0}^{-1}\left[C_{1}x_{1}^{p/q}(0)+x_{2}(0)\right]$

With the initial condition Eq. (5) for the integral state, the integral terminal sliding surface is zero at the initial time $t=0$ that is $s(t)_{t=0}=0$. Hence, the integral sliding surface Eq. (4) can define the surface from any given initial condition finally to the origin in the state space, and the controlled system slides from the initial time $t=0$. The first condition of removing reaching phase problems is satisfied(11)(12). In the sliding mode, the equation $s=0=\dot s$ is satisfied. Then from Eq. (2), Eq. (3), and Eq. (4) the ideal sliding dynamics is derived as

(6)
\begin{align*} \dot x_{0}=x_{1}\\ \dot x_{1}=x_{2}=-C_{0}x_{0}-C_{1}x_{1}^{p/q},\: \end{align*}

which is a dynamic representation of the integral terminal sliding surface Eq. (4). The solution of Eq. (6) is identical to the ideal integral terminal sliding surface and the real robust controlled output itself(11)(12). Therefore, the output can be pre-designed, pre-determined, and predicted.

Now, the suggested discontinuous ITSMC input for the uncertain plant Eq. (2) and the integral terminal sliding surface Eq. (4) is taken as follows:

(7)
\begin{align*} u_{1}=-(k_{1}x_{1}+k_{2}x_{2}+k_{3}x_{3}+\Delta k_{1}x_{1}+\triangle k_{2}x_{2}\\ +\triangle k_{3}x_{3}+k_{4}s+\triangle k_{5}sign(s)) \end{align*}

where an auxiliary nonlinear state $x_{3}$ is defined as

(8)
$x_{3}\equiv x_{1}^{(p/q -1)}x_{2}$

which is first defined in (31). Based on defining the auxiliary state $x_{3}$ in Eq. (8), the discontinuous input is chattering according to the condition of $sx_{3}$ in Eq. (15). Since that, it is easily shown that the existence condition of the sliding mode is clearly satisfied when $\triangle b\ne 0$. One takes the constant gains as

(9)
$k_{1}=a_{10}+C_{0}$

(10)
$k_{2}=a_{20}$

(11)
$k_{3}=C_{1}\dfrac{p}{q}$

(12)
$k_{4}>0$

and takes the discontinuously switching gains as follows:

(13)
$\triangle k_{1}=\begin{cases} \ge\dfrac{\max\left\{\triangle a_{1}-\triangle Ik_{1}\right\}}{(1-\rho)}sign(sx_{1})>0\\ \le\dfrac{\min\left\{\triangle a_{1}-\triangle Ik_{1}\right\}}{(1-\rho)}sign(sx_{1})<0 \end{cases}$

(14)
$\triangle k_{2}=\begin{cases} \ge\dfrac{\max\left\{\triangle a_{2}-\triangle Ik_{2}\right\}}{(1-\rho)}sign(sx_{2})>0\\ \le\dfrac{\min\left\{\triangle a_{2}-\triangle Ik_{2}\right\}}{(1-\rho)}sign(sx_{2})<0 \end{cases}$

(15)
$\triangle k_{3}=\begin{cases} \ge\dfrac{\max\left\{-\triangle Ik_{3}\right\}}{(1-\rho)}sign(sx_{3})>0\\ \le\dfrac{\min\left\{-\triangle Ik_{3}\right\}}{(1-\rho)}sign(sx_{3})<0 \end{cases}$

(16)
$\Delta k_{5}=\begin{cases} \ge\dfrac{\max\{\Delta d(t)\}}{(1-\rho)}sign(s)>0\\ \le\dfrac{\min\{\Delta d(t)\}}{(1-\rho)}sign(s)<0 \end{cases}$

where $sign(s)$ is the $sig\nu m(s)$ function as

(17)
$sign(s)=\begin{cases} +1 \quad {for} \quad s>0\\ 0 \quad {for} \quad s=0\\ -1 \quad {for} \quad s<0 \end{cases}$

2.1 Control input transformation

(18)
\begin{align*} u*=(b_{0})^{-1}u_{1},\:H_{u}=(b_{0})^{-1}\\ =(b_{0})^{-1}[-k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}-\triangle k_{1}x_{1}\\ -\triangle k_{2}x_{2}-\triangle k_{3}x_{3}-k_{4}s-\triangle k_{5}sign(s)] \end{align*}

where for the second order SI uncertain integral terminal linear case, the control input transformation is selected as $H_{u}=(b_{0})^{-1}$ while it is chosen as $H_{u}=(CB_{0})^{-1}$ in the LSMC. Most of the TSMCs commonly use this transformation(24)-(26)(27)(29)-(30) without explicitly naming it. However, the name of the control input transformation is first introduced in (31) and (32). $b_{0}^{-1}$ is multiplied to all the components of the discontinuous input Eq. (7) because the transformation of the control input for easy proving that the existence condition of the sliding mode is satisfied as one approach among the five approaches of the transformation (diagonalization)s(1)(2)(31). Since $\triangle b\ne 0$, the effect of $\triangle b\ne 0$ is considered in the selection of the discontinuous chattering gains Eq. (13)-Eq. (16). The results of $\triangle b\ne 0$ is the increase of the magnitude of the discontinuous chattering gains compared with the case when $\triangle b=0$. In the discontinuous input Eq. (18), the integral terminal sliding surface itself is one of the feedback elements and that brings the controlled system closer to the ideal predetermined terminal sliding surface(9).

(19)

\begin{align*} \dot s =C_{0}x_{1}+C_{1}\dfrac{p}{q}x_{3}+\dot x_{2}\\ =C_{0}x_{1}+C_{1}\dfrac{p}{q}x_{3}+(a_{10}+\triangle a_{1})x_{1}\\ +(a_{20}+\triangle a_{2})x_{2}+(b_{0}+\triangle b)u+\triangle d(x,\:t) \end{align*}

\begin{align*} =(a_{10}+C_{0})x_{1}+a_{20}x_{2}+C_{1}\dfrac{p}{q}x_{3}-k_{1}x_{1}-k_{2}x_{2}\\ -k_{3}x_{3}+\triangle a_{1}x_{1}-\triangle Ik_{1}x_{1}-(1+\triangle I)\triangle k_{1}x_{1}\\ +\triangle a_{2}x_{2}-\triangle Ik_{2}x_{2}-(1+\triangle I)\triangle k_{2}x_{2}-\triangle Ik_{3}x_{3}\\ -(1+\triangle I)k_{3}x_{3}-(1+\triangle I)k_{4}s+\triangle d(x,\:t)\\ -(1+\triangle I)\triangle k_{5}sign(s) \end{align*}

From Eq. (9)-(11), the real dynamics of $s$ becomes finally

(20)
\begin{align*} \dot s =\triangle a_{1}x_{1}-\triangle Ik_{1}x_{1}-(1+\triangle I)\triangle k_{1}x_{1}+\triangle a_{2}x_{2}\\ -\triangle Ik_{2}x_{2}-(1+\triangle I)\triangle k_{2}x_{2}-\triangle Ik_{3}x_{3}\\ -(1+\triangle I)k_{3}x_{3}-(1+\triangle I)k_{4}s+\triangle d(x,\:t)\\ -(1+\triangle I)\triangle k_{5}sign(s) \end{align*}

The original design problem of the ITSMC is finally converted to the stabilization problem against uncertainties and external disturbances by means of the discontinuously chattering input components and the feedback of the integral sliding surface. The performance(output) designed in the integral sliding surface becomes the real performance(output) for the output prediction, predetermination, and predesign(11)(12), and hence is completely separated with the performance robustness problem. The total closed loop stability with the transformed discontinuous control input Eq. (18) and the integral terminal sliding surface Eq. (4), along with the precise existence condition of the sliding mode, will be investigated in Theorem 1.

Theorem 1: If the integral terminal sliding surface Eq. (4) is designed to be stable, the transformed discontinuous control input Eq. (18) with the integral terminal sliding surface Eq. (4) satisfies the existence condition of the sliding mode on the pre-designed integral terminal sliding surface and closed loop exponential stability to the integral terminal sliding surface $s=0$ including the origin.

Proof: Take a Lyapunov function candidate as

(21)
$V(x)=\dfrac{1}{2}s^{2}$

Differentiating Eq. (21) with time leads to

(22)
$\dot V(x)=s·\dot s$

Substituting Eq. (20) into Eq. (22) leads to

(23)
\begin{align*} \dot V(x)=s(\triangle a_{1}-\triangle Ik_{1})x_{1}-s(1+\triangle I)\triangle k_{1}x_{1}\\ +s(\triangle a_{2}-\triangle Ik_{2})x_{2}-s(1+\triangle I)\triangle k_{2}x_{2}\\ -s\triangle Ik_{3}x_{3}-s(1+\triangle I)k_{3}x_{3}-(1+\triangle I)k_{4}s^{2}\\ +s\triangle d(x,\:t)-(1+\triangle I)\triangle k_{5}vert s vert \end{align*}

Since the uncertainty and external disturbance terms in Eq. (23) are canceled out due to the chattering discontinuous input terms by means of the switching gains in Eq. (13)-Eq. (16), one can obtain the following equation(11)(12)

(24)

$\dot V(x)=s·\dot s\le -(1-\rho)k_{4}s^{2}\le 0$

and

$s_{i}·\dot s_{i}< -(1-\rho)k_{4i}s_{i}^{2}\le 0,\:i=1,\:2,\:...,\:m$

The existence condition of the sliding mode on the predetermined integral terminal sliding surface by the transformed discontinuous control input is proved theoretically for the complete formulation of the TSMC design for the output prediction. By only through the proof of the existence condition of the sliding mode, the strong robustness of every point on the whole trajectory of the predetermined integral terminal sliding surface from a given initial condition to the origin is guaranteed. Hence, the controlled robust output can be predicted, predesigned, and predetermined. The second condition of removing reaching phase problems is satisfied(11)(12). From Eq. (24) the following equation is obtained.

(25)
$\dot V(x)\le -2(1-\rho)k_{4}V(x)$

From Eq. (25), the following equation is obtained

(26)

$\dot V(x)+2(1-\rho)k_{4}V(x)\le 0$

$V(x(t))\le V(x(0))e^{-2(1-\rho)k_{4}t}$

which completes the proof of Theorem 1.

The existence condition of the sliding mode is proved with respect to the control input transformation for second order uncertain linear systems. The equation of the sliding mode, i.e. the sliding surface is invariant with respect to the control input transformation.

2.2 Sliding surface (full) transformation

(27)
\begin{align*} s*=(b_{0})^{-1}· s,\:H_{s}(x,\:t)=(b_{0})^{-1}\\ =(b_{0})^{-1}(C_{0}x_{0}+C_{1}x_{1}^{p/q}+x_{2}) \quad (=0) \end{align*}

The transformation is selected as $H_{s}(x,\:t)=(b_{0})^{-1}$ as the same way as the control input transformation described above, while in the LSMC it is chosen as $H_{s}(x,\:t)=(CB_{0})^{-1}$. Except (32), this transformation appears for the first time in the TSMC. Most TSMCs typically use only the well known control input transformation without giving it a specific name. The special initial condition $x_{0}(0)$ in Eq. (2) for the integral state is determined to ensure the integral sliding surface Eq. (27) is the zero at $t=0$ for any initial condition $x_{1}(0) \quad {and} \quad x_{2}(0)$ as

(28)
$x_{0}(0)=-C_{0}^{-1}\left[C_{1}x_{1}^{p/q}(0)+x_{2}(0)\right]$

With the initial condition Eq. (28) for the integral state, the integral terminal sliding surface is zero at the initial time $t=0$ that is $s(t)_{t=0}=0$. Hence, the transformed integral terminal sliding surface Eq. (27) can define the surface from any given initial condition finally to the origin in the state space, as a result, the controlled system slides from the initial time $t=0$. The first condition of removing reaching phase problems is satisfied(11)(12). In the sliding mode, the equation $s=0=\dot s$ is satisfied. Then from Eq. (2), Eq. (3), and Eq. (27), the ideal sliding dynamics is derived as

(29)
\begin{align*} \dot x_{0}=x_{1}\\ \dot x_{1}=x_{2}=-C_{0}x_{0}-C_{1}x_{1}^{p/q},\: \end{align*}

which is a dynamic representation of the transformed integral terminal sliding surface Eq. (27) and the same as that of the control input transformation since $s=0=s*$ when $b_{0}\ne 0or\infty$. The solution of Eq. (29) is identical to the set of the ideal integral terminal sliding surface and the real robust controlled output itself(11)(12). Therefore, the output can be pre-designed, predetermined, and predicted. The prediction of the controlled output is possible by means of the solution of Eq. (29).

Now, the suggested discontinuous ITSMC input for uncertain plant Eq. (2) and the transformed integral terminal sliding surface Eq. (27) is taken as follows:

(30)
\begin{align*} u_{2}=-k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}-\Delta k_{1}x_{1}\\ -\triangle k_{2}x_{2}-\triangle k_{3}x_{3}-k_{4}s*-\triangle k_{5}sign(s*) \end{align*}

where one takes the constant gains as

(31)
$k_{1}=b_{0}^{-1}\left\{a_{10}+C_{0}\right\}$

(32)
$k_{2}=b_{0}^{-1}a_{20}$

(33)
$k_{3}=b_{0}^{-1}C_{1}\dfrac{p}{q}$

(34)
$k_{4}>0$

and takes the discontinuously switching gains as follows:

(35)
$\triangle k_{1}=\begin{cases} \ge\dfrac{\max\left\{b_{0}^{-1}\triangle a_{1}-\triangle Ik_{1}\right\}}{(1-\rho)}sign(s*x_{1})>0\\ \le\dfrac{\min\left\{b_{0}^{-1}\triangle a_{1}-\triangle Ik_{1}\right\}}{(1-\rho)}sign(s*x_{1})<0 \end{cases}$

(36)
$\triangle k_{2}=\begin{cases} \ge\dfrac{\max\left\{b_{0}^{-1}\triangle a_{2}-\triangle Ik_{2}\right\}}{(1-\rho)}sign(s*x_{2})>0\\ \le\dfrac{\min\left\{b_{0}^{-1}\triangle a_{2}-\triangle Ik_{2}\right\}}{(1-\rho)}sign(s*x_{2})<0 \end{cases}$

(37)
$\triangle k_{3}=\begin{cases} \ge\dfrac{\max\left\{-\triangle Ik_{3}\right\}}{(1-\rho)}sign(s*x_{3})>0\\ \le\dfrac{\min\left\{-\triangle Ik_{3}\right\}}{(1-\rho)}sign(s*x_{3})<0 \end{cases}$

(38)
$\Delta k_{5}=\begin{cases} \ge\dfrac{\max\left\{b_{0}^{-1}\Delta d(t)\right\}}{(1-\rho)}sign(s*)>0\\ \le\dfrac{\min\left\{b_{0}^{-1}\Delta d(t)\right\}}{(1-\rho)}sign(s*)<0 \end{cases}$

Then the real dynamics of the transformed integral sliding surface by the discontinuous control input, i.e. the time derivative of $s*$ becomes

(39)

\begin{align*} \dot s *=b_{0}^{-1}\left\{C_{0}x_{1}+C_{1}\dfrac{p}{q}x_{3}+\dot x_{2}\right\}\\ =b_{0}^{-1}C_{0}x_{1}+b_{0}^{-1}C_{1}\dfrac{p}{q}x_{3}+b_{0}^{-1}(a_{10}+\triangle a_{1})x_{1}\\ +b_{0}^{-1}(a_{20}+\triangle a_{2})x_{2}+b_{0}^{-1}(b_{0}+\triangle b)u\\ +b_{0}^{-1}\triangle d(x,\:t) \end{align*}

\begin{align*} =b_{0}^{-1}(a_{10}+C_{0})x_{1}+b_{0}^{-1}a_{20}x_{2}+b_{0}^{-1}C_{1}\dfrac{p}{q}x_{3}\\ -k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}+b_{0}^{-1}\triangle a_{1}x_{1}-\triangle Ik_{1}x_{1}\\ -(1+\triangle I)\triangle k_{1}x_{1}+b_{0}^{-1}\triangle a_{2}x_{2}-\triangle Ik_{2}x_{2}\\ -(1+\triangle I)\triangle k_{2}x_{2}-\triangle Ik_{3}x_{3}-(1+\triangle I)k_{3}x_{3}\\ -k_{4}s*-\triangle Ik_{4}s+b_{0}^{-1}\triangle d(x,\:t)\\ -(1+\triangle I)\triangle k_{5}sign(s*) \end{align*}

From Eq. (31)-(33), the real dynamics of $s*$ becomes finally

(40)
\begin{align*} \dot s* =b_{0}^{-1}\triangle a_{1}x_{1}-\triangle Ik_{1}x_{1}-(1+\triangle I)\triangle k_{1}x_{1}\\ +b_{0}^{-1}\triangle a_{2}x_{2}-\triangle Ik_{2}x_{2}-(1+\triangle I)\triangle k_{2}x_{2}\\ -\triangle Ik_{3}x_{3}-(1+\triangle I)k_{3}x_{3}-(1+\triangle I)k_{4}s*\\ +b_{0}^{-1}\triangle d(x,\:t)-(1+\triangle I)\triangle k_{5}sign(s*) \end{align*}

From Eq. (40), the original design problem of the ITSMC is finally converted to the stabilization problem against the uncertainties and external disturbances by means of the discontinuously chattering input and the feedback of the transformed integral terminal sliding surface. The total closed loop stability with the discontinuous control input Eq. (30) and the transformed integral terminal sliding surface Eq. (27) together with the precise existence condition of the sliding mode will be investigated in Theorem 2.

Theorem 2: If the transformed integral terminal sliding surface Eq. (27) is stably designed, the discontinuous control input Eq. (30) with the stable transformed integral terminal sliding surface Eq. (27) satisfies the existence condition of the sliding mode on the pre-designed integral terminal sliding surface and closed loop exponential stability to the integral terminal sliding surface $s=0$ including the origin.

Proof: Take a Lyapunov function candidate as

(41)
$V(x)=\dfrac{1}{2}s*^{2}$

Differentiating Eq. (41) with time leads to

(42)
$\dot V(x)=s *·\dot s*$

Substituting Eq. (40) into Eq. (42) leads to

(43)
\begin{align*} \dot V(x)=s*(b_{0}^{-1}\triangle a_{1}-\triangle Ik_{1})x_{1}-s*(1+\triangle I)\triangle k_{1}x_{1}\\ +s*(b_{0}^{-1}\triangle a_{2}-\triangle Ik_{2})x_{2}-s*(1+\triangle I)\triangle k_{2}x_{2}\\ -s*\triangle Ik_{3}x_{3}-s*(1+\triangle I)k_{3}x_{3}-(1+\triangle I)k_{4}s*^{2}\\ +s*b_{0}^{-1}\triangle d(x,\:t)-(1+\triangle I)\triangle k_{5}vert s* vert \end{align*}

Since the uncertainty and external disturbance terms in Eq. (43) are canceled out due to the chattering control input terms by means of the switching gains in Eq. (35)-Eq. (38), one can obtain the following equation(11)(12)

(44)

$\dot V(x)=s*·\dot s *\le -(1-\rho)k_{4}s*^{2}\le 0$

and

$s_{i}*·\dot s_{i}*\le -(1-\rho)k_{4i}s_{i}*^{2}\le 0,\:i=1,\:2,\:...,\:m$

The theoretical proof of the existence condition for sliding mode on a predetermined transformed integral terminal sliding surface, implemented by a discontinuous control input, is crucial for establishing a complete formulation of TSMC design for output prediction performance. This proof ensures strong robustness at every point along the entire trajectory of the predetermined integral sliding surface, from a given initial condition to the origin without the reaching phase, by satisfying the existence condition of the sliding mode. As a result, we can predict, pre-design, and pre-determine controlled robust output. Moreover, this proof also satisfies the second condition of removing reaching phase problems, as demonstrated in previous works (11)(12). From Eq. (44), we can derive the following equation.

(45)
$\dot V(x)\le -2(1-\rho)k_{4}V(x)$

From Eq. (45), the following equation is obtained

(46)

$\dot V(x)+2(1-\rho)k_{4}V(x)\le 0$

$V(x(t))\le V(x(0))e^{-2(1-\rho)k_{4}t}$

which completes the proof of Theorem 2.

If the sliding mode equation $s*=0$, then $s=0$ since $b_{0}\ne 0$ or $b_{0}\ne\infty$. The inverse augment also holds, therefore the both ideal sliding surfaces are equal i.e. $s=0=s*$, which completes the proof of Theorem 2.

Until the first two transformations, Utkin's invariant theorem can be applied. However, for the next transformation, the theorem cannot be used. Nevertheless, this transformation can serve as an alternative design method for the ITSMC and LSMC. It is possible to prove the existence condition of the sliding mode and stabilization using this method.

2.3 Sliding surface part transformations

(47)
$s^{+1}=C_{0}x_{0}+C_{1}x_{1}^{p/q}+(b_{0})^{-1}x_{2}$

The part transformation is selected as $(b_{0})^{-1}$ that is multiplied to only $x_{2}$ term in the integral terminal sliding surface. This part transformation appears for the first time in the TSMC. The property of the Utkin’s invariant theorem can not be applicable since $s=0$ is not equal to $s^{+1}=0$ except $b_{0}=1$. Besides this part transformation, there are the two more part transformation sliding surfaces as

(48)
$s^{+2}=C_{0}x_{0}+(b_{0})^{-1}C_{1}x_{1}^{p/q}+(b_{0})^{-1}x_{2}$

(49)
$s^{+3}=(b_{0})^{-1}C_{0}x_{0}+C_{1}x_{1}^{p/q}+(b_{0})^{-1}x_{2}$

The research on these two remaining part transformation sliding surfaces, $s^{+2}$ and $s^{+3}$, is dropped in this paper and will be included in the further study because of the limited allowed space.

The special initial condition $x_{0}(0)$ in Eq. (47) for the integral state is determined so that the part transformation integral sliding surface Eq. (47) is the zero at $t=0$ for any given initial condition $x_{1}(0){and}x_{2}(0)$ as

(50)
$x_{0}(0)=-C_{0}^{-1}\left[C_{1}x_{1}^{p/q}(0)+b_{0}^{-1}x_{2}(0)\right]$

With the initial condition Eq. (50) for the integral state, the integral terminal sliding surface is zero at the initial time $t=0$ that is $s^{+}(t)_{t=0}=0$. Hence, the part transformation integral sliding surface Eq. (47) can define the surface from any given initial condition finally to the origin in the state space, and the controlled system can slide from initial time $t=0$. The first condition of removing reaching phase problems is satisfied(11)(12). In the sliding mode, the equation $s^{+1}=0=\dot s^{+1}$ is satisfied. Then from Eq. (2) and Eq. (47) the ideal sliding dynamics is derived as

(51)
\begin{align*} \dot x_{0}=x_{1}\\ \dot x_{1}=x_{2}=-b_{0}\left\{C_{0}x_{0}+C_{1}x_{1}^{p/q}\right\} \end{align*}

which is the dynamic representation of the part transformation integral terminal sliding surface Eq. (47). The solution of Eq. (51) is identical to the set of the ideal part transformed integral terminal sliding surface and the real robust controlled output itself(11)(12). Therefore, the output can be pre-designed, predetermined, and predicted.

Now, the suggested discontinuous ITSMC input for uncertain plant Eq. (2) and the partially transformed integral terminal sliding surface Eq. (47) is taken as follows:

(52)
\begin{align*} u_{2}=-k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}-\Delta k_{1}x_{1}-\triangle k_{2}x_{2}\\ -\triangle k_{3}x_{3}-k_{4}s^{+1}-\triangle k_{5}sign(s^{+1}) \end{align*}

where one takes the constant gains as

(53)
$k_{1}=b_{0}^{-1}(a_{10})+C_{0}$

(54)
$k_{2}=b_{0}^{-1}a_{20}$

(55)
$k_{3}=C_{1}\dfrac{p}{q}$

(56)
$k_{4}>0$

and takes the discontinuously switching gains as follows:

(57)
$\triangle k_{1}=\begin{cases} \ge\dfrac{\max\left\{b_{0}^{-1}\triangle a_{1}-\triangle Ik_{1}\right\}}{(1-\rho)}sign(s^{+1}x_{1})>0\\ \le\dfrac{\min\left\{b_{0}^{-1}\triangle a_{1}-\triangle Ik_{1}\right\}}{(1-\rho)}sign(s^{+1}x_{1})<0 \end{cases}$

(58)
$\triangle k_{2}=\begin{cases} \ge\dfrac{\max\left\{b_{0}^{-1}\triangle a_{2}-\triangle Ik_{2}\right\}}{(1-\rho)}sign(s^{+1}x_{2})>0\\ \le\dfrac{\min\left\{b_{0}^{-1}\triangle a_{2}-\triangle Ik_{2}\right\}}{(1-\rho)}sign(s^{+1}x_{2})<0 \end{cases}$

(59)
$\triangle k_{3}=\begin{cases} \ge\dfrac{\max\left\{-\triangle Ik_{3}\right\}}{(1-\rho)}sign(s^{+1}x_{3})>0\\ \le\dfrac{\min\left\{-\triangle Ik_{3}\right\}}{(1-\rho)}sign(s^{+1}x_{3})<0 \end{cases}$

(60)
$\Delta k_{5}=\begin{cases} \ge\dfrac{\max\left\{b_{0}^{-1}\Delta d(t)\right\}}{(1-\rho)}sign(s^{+1})>0\\ \le\dfrac{\min\left\{b_{0}^{-1}\Delta d(t)\right\}}{(1-\rho)}sign(s^{+1})<0 \end{cases}$

Then the real dynamics of the partially transformed integral sliding surface by the discontinuous control input, i.e. the time derivative of $s$ becomes

(61)

\begin{align*} \dot s^{+1}=C_{0}x_{1}+C_{1}\dfrac{p}{q}x_{3}+b_{0}^{-1}\dot x_{2}\\ =C_{0}x_{1}+C_{1}\dfrac{p}{q}x_{3}+b_{0}^{-1}(a_{10}+\triangle a_{1})x_{1}\\ +b_{0}^{-1}(a_{20}+\triangle a_{2})x_{2}+b_{0}^{-1}(b_{0}+\triangle b)u\\ +b_{0}^{-1}\triangle d(x,\:t) \end{align*}

\begin{align*} =(b_{0}^{-1}a_{10}+C_{0})x_{1}+b_{0}^{-1}a_{20}x_{2}+C_{1}\dfrac{p}{q}x_{3}\\ -k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}+b_{0}^{-1}\triangle a_{1}x_{1}-\triangle Ik_{1}x_{1}\\ -(1+\triangle I)\triangle k_{1}x_{1}+b_{0}^{-1}\triangle a_{2}x_{2}-\triangle Ik_{2}x_{2}\\ -(1+\triangle I)\triangle k_{2}x_{2}-\triangle Ik_{3}x_{3}-(1+\triangle I)k_{3}x_{3}\\ -k_{4}s^{+1}-\triangle Ik_{4}s^{+1}+b_{0}^{-1}\triangle d(x,\:t)\\ -(1+\triangle I)\triangle k_{5}sign(s^{+1}) \end{align*}

From Eq. (53)-Eq. (55), the real dynamics of $s^{+1}$ becomes finally

(62)
\begin{align*} \dot s^{+1}=b_{0}^{-1}\triangle a_{1}x_{1}-\triangle Ik_{1}x_{1}-(1+\triangle I)\triangle k_{1}x_{1}\\ +b_{0}^{-1}\triangle a_{2}x_{2}-\triangle Ik_{2}x_{2}-(1+\triangle I)\triangle k_{2}x_{2}\\ -\triangle Ik_{3}x_{3}-(1+\triangle I)k_{3}x_{3}-(1+\triangle I)k_{4}s^{+1}\\ +b_{0}^{-1}\triangle d(x,\:t)-(1+\triangle I)\triangle k_{5}sign(s^{+1}) \end{align*}

From Eq. (62), the original design problem of the TSMC is finally converted to the stabilization problem against the uncertainties and external disturbances by means of the discontinuously chattering input and the feedback of the transformed integral sliding surface. The total closed loop stability with the discontinuous control input Eq. (52) and the transformed integral sliding surface Eq. (47) together with the precise existence condition of the sliding mode will be investigated in Theorem 3.

Theorem 3: If the stable partially transformed integral terminal sliding surface Eq. (47) is designed, the discontinuous control input Eq. (52) with the stable partially transformed integral sliding surface Eq. (47) satisfies the existence condition of the sliding mode on the pre-designed integral sliding surface and closed loop exponential stability to the integral sliding surface $s=0$ including the origin.

Proof: Take a Lyapunov function candidate as

(63)
$V(x)=\dfrac{1}{2}s^{+1^{2}}$

Differentiating Eq. (63) with time leads to

(64)
$\dot V(x)=s^{+1}·\dot s^{+1}$

Substituting Eq. (62) into Eq. (63) leads to

(65)
\begin{align*} \dot V(x)=s^{+1}(b_{0}^{-1}\triangle a_{1}-\triangle Ik_{1})x_{1}-s^{+1}(1+\triangle I)\triangle k_{1}x_{1}\\ +s^{+1}(b_{0}^{-1}\triangle a_{2}-\triangle Ik_{2})x_{2}\\ -s^{+1}(1+\triangle I)\triangle k_{2}x_{2}-s^{+1}\triangle Ik_{3}x_{3}\\ -s^{+1}(1+\triangle I)k_{3}x_{3}-(1+\triangle I)k_{4}s^{+1^{2}}\\ +s^{+1}b_{0}^{-1}\triangle d(x,\:t)-(1+\triangle I)\triangle k_{5}vert s^{+1}vert \end{align*}

Since the uncertainty and external disturbance terms in Eq. (65) are canceled out due to the chattering control input terms by means of the switching gains in Eq. (57)-Eq. (59), one can obtain the following equation(11)(12)

(66)

$\dot V(x)=s^{+1}·\dot s^{+1}\le -(1-\rho)k_{4}s^{+1^{2}}\le 0$

and

$s_{i}^{+1}·\dot s_{i}^{+1}\le -(1-\rho)k_{4i}s_{i}^{+1^2}\le 0,\:i=1,\:2,\:...,\:m$

The theoretical proof of the existence condition for sliding mode on a predetermined transformed integral sliding surface, implemented by a discontinuous control input, is essential to establish a complete formulation of TSMC design for output prediction performance. Through this proof, we can guarantee strong robustness at every point along the whole trajectory of the predetermined integral sliding surface, from a given initial condition to the origin. This, in turn, enables us to predict, pre-design, and pre-determine the controlled robust output. Moreover, the second condition of removing reaching phase problems is also satisfied, as described in previous works (11)(12). From equation Eq. (22), we can derive the following result.

(67)
$\dot V(x)\le -2(1-\rho)k_{4}V(x)$

From Eq. (67), the following equation is obtained

(68)

$\dot V(x)+2(1-\rho)k_{4}V(x)\le 0$

$V(x(t))\le V(x(0))e^{-2(1-\rho)k_{4}t}$

which completes the proof of Theorem 3.

Although the invariance property of Utkin's theorem may not be applicable to this particular transformation, it still serves as a valuable design and stabilization approach for ITSMCs. In light of the results from Theorem 1, Theorem 2, and Theorem 3, there are now five approaches or design methods for ITSMCs: control input transformation, full sliding surface transformation, and three sliding surface part transformations. While the first approach is already well-known (though not always referred to by name), having been discussed in previous works such as (31) and (32), the second transformation is introduced here for the first time in the context of TSMCs except (32). As for the last three cases, they also make their debut in TSMCs through this research.

3. Design Example and Illustrative Simulation Study

Consider a second order uncertain canonical system

(69)
$\dot x =\left[\begin{matrix} 0& 1\\+- 0.2& -3 +- 0.3\end{matrix}\right]x(t)+\left[\begin{matrix} 0\\2 +- 0.3\end{matrix}\right]u+\left[\begin{matrix} 0\\+- 9.5\end{matrix}\right]$

where the nominal parameter $a_{10}$, $a_{20}$, and $b_{0}$, matched uncertainties $\triangle a_{1}$, $\triangle a_{2}$, and $\triangle b$, and external disturbance $\triangle d(x,\:t)$ are

(70)

$a_{10}=0$, $a_{20}=-3$, $b_{0}=2.0$, $\triangle a_{1}=+-0.3$,

$\triangle a_{2}=+-0.3$, $\triangle b=+-0.3$, and $\Delta d(x,\:t)=+-9.5$

The $\triangle I$ in assumption 1 becomes

(71)
$\triangle I=\triangle b(b_{0})^{-1}= +- 0.3/2= +- 0.15$ and $\rho =0.15$

which satisfies the assumption 1

To design the proposed ITSMC with the integral sliding surface and transformed control input, first the stable coefficient in the suggested integral sliding surface is determined as

(72)
$C_{0}=9.0$ and $C_{1}=6.0$

such that the polynomial is Hurwitz

(73)
$r^{2}+C_{1}r+C_{0}=r^{2}+6r+9=(r+3)^{2}$

The $p$ and $q$ are selected as

(74)

$p=2.8$ and $q=5$

$p/q=0.56<1$

which satisfies the terminal condition. The $p$ is positive real not positive odd integer because any positive number is possible such that $0<p/q<1$. Then the integral sliding surface becomes

(75)
$s=x_{2}+6x_{1}^{0.56}+9x_{0}(=0)$

where

(76)
$x_{0}(t)=\int_{0}^{t}x_{1}(\tau)d\tau +x_{0}(0)$

and the ideal sliding dynamics becomes

(77)
\begin{align*} \dot x_{0}=x_{1}x_{0}(0)\\ \dot x_{1}=-9x_{0}-6x_{1}^{0.56}=x_{2},\: x_{1}(0) \end{align*}

which is the design goal for the same outputs with respect to the three transformations

3.1 Control input transformation

(78)
\begin{align*} u*=(b_{0})^{-1}u_{1},\:H_{u}=(b_{0})^{-1}=2^{-1}\\ =2^{-1}[-k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}-\Delta k_{1}x_{1}-\Delta k_{2}x_{2}\\ -\Delta k_{3}x_{3}-\triangle k_{4}s-\triangle k_{5}sign(s)] \end{align*}

From Eq. (9)-Eq. (11), the constant feedback gains are accordingly designed as

(79)
$k_{1}=a_{10}+C_{0}=9.0$

(80)
$k_{2}=a_{20}=-3.0$

(81)
$k_{3}=C_{1}\dfrac{p}{q}=6.0*0.56=3.36$

(82)
$k_{4}=300.0>0$

If one takes the switching gains as follows:

(83)

$\Delta k_{1}=\begin{cases} 2.5{if}sx_{1}>0\\ -2.5{if}sx_{1}<0 \end{cases}$, $\Delta k_{2}=\begin{cases} 4.5{if}sx_{2}>0\\ -4.5{if}sx_{2}<0 \end{cases}$,

$\Delta k_{3}=\begin{cases} 3.5{if}sx_{3}>0\\ -3.5{if}sx_{3}<0 \end{cases}$ $\Delta k_{5}=\begin{cases} 12.0{if}s>0\\ -12.0{if}s<0 \end{cases}$

The equation Eq. (24) becomes

(84)
$\dot V(x)=s·\dot s\le -255.0s^{2}$

The simulation is carried out under 1[msec] sampling time and with $x(0)=[3 1.5]^{T}$ initial state and by Eq. (4), the initial condition for the integral state becomes

(85)
\begin{align*} x_{0}(0)=-C_{0}^{-1}\left[C_{1}x_{1}^{p/q}(0)+x_{2}(0)\right]\\ =-9^{-1}[6*3^{0.56}+1.5]\\ =-1.4 \end{align*}

Fig. 1 shows the two finite time stabilization output responses, $x_{1}$ and $x_{2}$ for the two cases those are the ideal sliding dynamics output and the real output with uncertainty and external disturbance. Those outputs are identical, which means that the prediction, predetermination, predesign of the output, and design separation of the performance design and robustness problem are possible, by means of removing the reaching phase. Fig. 2 shows the phase trajectories with the ideal trajectory and the real trajectory. The real trajectory is identical to the ideal one as can be shown in Fig. 2. The terminal sliding surface time trajectory and the corresponding control input are depicted in Fig. 3 and Fig. 4, respectively. The abrupt change in the terminal sliding surface and control input in Fig. 3 and Fig. 4 is due to the singular problem, which requires further investigation.

3.2 Sliding surface (full) transformation

(86)
\begin{align*} s*=(b_{0})^{-1}· s,\:H_{s}(x,\:t)=(b_{0})^{-1}=\dfrac{1}{2}\\ =\dfrac{1}{2}(C_{0}x_{0}+C_{1}x_{1}^{p/q}+x_{2}) \end{align*}

For the comparison of both transformations, the same stable coefficient of the sliding surface and the same parameters of the power function are determined.

Thus the new transformed integral terminal sliding surface of the sliding surface transformation becomes

(87)
$s*=2^{-1}\left(x_{2}+6x_{1}^{0.56}+9x_{0}\right)(=0)$

그림. 1. 제어입력 변환에 의한 두가지 응답 $x_{1}$과 $x_{2}$

Fig. 1. Output responses, $x_{1}$ and $x_{2}$ by control input transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig1.png

그림. 2. 제어입력 변회에 의한 이상과 실제 플랜트의 두가지 상 궤적

Fig. 2. Phase trajectories by control input transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig2.png

그림. 3. 제어입력 변환애 의한 슬라이딩 면

Fig. 3. Sliding surface time trajectory by control input transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig3.png

그림. 4. 제어입력 변환애 의한 제어입력

Fig. 4. Control input by control input transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig4.png

The ideal sliding dynamics is the same as that of the control input transformation and the initial condition $x_{0}(0)$ is also the same as that of the above transformation.

Now, the integral TSMC control input is taken as follows:

(88)
\begin{align*} u_{2}=-k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}-\Delta k_{1}x_{1}-\triangle k_{2}x_{2}\\ -\triangle k_{3}x_{3}-\triangle k_{4}s*-\triangle k_{5}sign(s*) \end{align*}

From Eq. (31)-Eq. (33), by letting the constant gain

(89)
$k_{1}=(b_{0})^{-1}\left(a_{10}+C_{0}\right)=4.5$

(90)
$k_{2}=(b_{0})^{-1}a_{20}=2^{-1}(-3)=-1.5$

(91)
$k_{3}=\dfrac{1}{2}C_{1}\dfrac{p}{q}=\dfrac{1}{2}6*0.56=1.68$

(92)
$k_{4}=150.0>0$

If one take the switching gain as the design parameters

(93)

$\Delta k_{1}=\begin{cases} 1.25{if}s*x_{1}>0\\ -1.25{if}s*x_{1}<0 \end{cases}$

$\Delta k_{2}=\begin{cases} 1.75{if}s*x_{2}>0\\ -1.75{if}s*x_{2}<0 \end{cases}$

$\Delta k_{3}=\begin{cases} 1.75{if}s*x_{3}>0\\ -1.75{if}s*x_{3}<0 \end{cases}$

$\triangle k_{5}=\begin{cases} 6.0{if}s*>0\\ -6.0{if}s*<0 \end{cases}$

then one can obtain the following equation

(94)
$s*·\dot s* <- 127.5s*^{2}<0$

if $s*=0$, then $s=0$. The inverse augment is also true. The switching gains in Eq. (93) can be obtained also from Eq. (83) by multiplying $(b_{0})^{-1}=2^{-1}$.

The simulation was conducted using a sampling time of 1[msec] and with the same initial condition. Fig. 5 illustrates the two output responses, $x_{1}$ and $x_{2}$, for the ideal and real cases. These responses are almost identical to those shown in Fig. 1 because the sliding surface is identical and the continuous and discontinuous gains resulting from both control inputs, $u*$ and $u_{2}$, are the same. It is possible to predict, predesign, and

그림. 5. 슬라이딩면 변환에 의한 두가지 응답 $x_{1}$과 $x_{2}$

Fig. 5. Output responses, $x_{1}$ and $x_{2}$ by sliding surface transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig5.png

그림. 6. 슬라이딩면 변회에 의한 이상과 실제 플랜트의 두가지 상 궤적

Fig. 6. Phase trajectories by sliding surface transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig6.png

그림. 7. 슬라이딩면 변환애 의한 슬라이딩 면

Fig. 7. Sliding surface time trajectory by sliding surface transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig7.png

그림. 8. 슬라이딩면 변환애 의한 제어입력

Fig. 8. Control input by sliding surface transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig8.png

predetermine the real robust output by designing and solving the ideal sliding dynamics, just like with the above control input transformation. Fig. 6displays the two phase trajectories for the ideal and real cases, those are identical. The sliding surface time trajectory and the corresponding control input are shown in Fig. 7 and Fig. 8, respectively. The value of the sliding surface in Fig. 7 is about half of that in Fig. 3 since the new sliding surface in the sliding surface transformation is multiplied by $\dfrac{1}{2}$.

3.3 Sliding surface part transformation

(95)
$s^{+1}=C_{0}x_{0}+C_{1}x_{1}^{p/q}+\dfrac{1}{2}x_{2}(=0)$

For the comparison of the three transformations and the same ideal sliding dynamics with that of the above first two transformations for the same output (performance), the coefficients in the part transformation sliding surface are set to

(96)
$C_{0}=4.5$ and $C_{1}=3.0$

These coefficients are half of those used in the control input transformation and share the same parameters in the power function of the part transformation sliding surface as the first two transformations. As a result, the new partially transformed sliding surface of the sliding surface part transformation becomes

(97)
$s^{+1}=\dfrac{1}{2}x_{2}+3x_{1}^{0.56}+4.5x_{0}(=0)$

Now, the integral TSMC control input is taken as follows:

(98)
\begin{align*} u_{2}=-k_{1}x_{1}-k_{2}x_{2}-k_{3}x_{3}-\Delta k_{1}x_{1}-\triangle k_{2}x_{2}\\ -\triangle k_{3}x_{3}-\triangle k_{4}s^{+1}-\triangle k_{5}sign(s^{+1}) \end{align*}

From Eq. (21)-Eq. (23), by letting the constant gain

(99)
$k_{1}=(b_{0})^{-1}a_{10}+C_{0}=4.5$

(100)
$k_{2}=(b_{0})^{-1}a_{20}=2^{-1}(-3)=-1.5$

(101)
$k_{3}=C_{1}\dfrac{p}{q}=3.0*0.56=1.68$

(102)
$k_{4}=150.0>0$

그림. 9. 슬라이딩 면 부분 변환에 의한 두가지 응답 $x_{1}$과 $x_{2}$

Fig. 9. Output responses, $x_{1}$ and $x_{2}$ by sliding surface part transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig9.png

그림. 10. 슬라이딩면 부분 변회에 의한 이상과 실제 플랜트의 두가지 상 궤적

Fig. 10. Phase trajectories by sliding surface part transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig10.png

그림. 11. 슬라이딩 면 부분 변환애 의한 슬라이딩 면

Fig. 11. Sliding surface time trajectory by sliding surface part transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig11.png

그림. 12. 슬라이딩면 부분 변환애 의한 제어입력

Fig. 12. Control input by sliding surface transformation

../../Resources/kiee/KIEE.2023.72.5.628/fig12.png

If one take the switching gain as the design parameters

(103)

$\Delta k_{1}=\begin{cases} 1.25{if}s^{+1}x_{1}>0\\ -1.25{if}s^{+1}x_{1}<0 \end{cases}$

$\Delta k_{2}=\begin{cases} 2.25{if}s^{+1}x_{2}>0\\ -2.25{if}s^{+1}x_{2}<0 \end{cases}$

$\Delta k_{3}=\begin{cases} 1.75{if}s^{+1}x_{3}>0\\ -1.75{if}s^{+1}x_{3}<0 \end{cases}$

$\triangle k_{4}=\begin{cases} 6.0{if}s^{+1}>0\\ -6.0{if}s^{+1}<0 \end{cases}$

then one can obtain the following equation

(104)
$s^{+1}·\dot s^{+1}<-127.5s^{+1^{2}}<0$

The switching gains in Eq. (103) can be obtained also from Eq. (93).

Simulations are performed with a sampling time of 1[msec] and using the same initial condition as before. Fig. 9displays the output responses for the two cases: the ideal case and the real case. These responses are nearly identical to those shown in Fig. 1 and Fig. 5, respectively, as the same ideal sliding dynamics are employed. Consequently, the real robust output can be predicted, predetermined, and predesigned in the same way as those of the first two transformations. Fig. 10depicts the phase trajectories for the ideal and real cases, with the real trajectory being the same as that of the ideal case. Additionally, Fig. 11 and Fig. 12 show the sliding surface time trajectory and the corresponding control input, respectively. The outputs of the three transformations are identical to the three ideal outputs, indicating that the three transformation outputs exhibit the same level of performance because of the identical ideal dynamics employed. Therefore, the three real outputs can be predicted, predesigned, and predetermined in the same way as designed and expected.

4. Conclusion

In this paper, the invariant theorem of Utkin is proved rigorously for the ITSMC of second order SI uncertain linear systems when $\triangle b\ne 0$. The ITSMC can be designed using the five approaches:control input transformation, sliding surface transformation, and three sliding surface part transformations, which is for the first time pointed out in this paper. The Utkin’s theorem can be applicable to only the first two transformations, i.e., control input transformation and sliding surface full transformation. But to the remaining three sliding surface part transformations, the Utkin’s theorem can not applied but they have the meaning as the design approaches to the ITSMCs. The sliding surface full transformation for the first time appears in the TSMCs with the finite convergence except (32), while the three sliding surface part transformations are introduced for the first time in the TSMCs. The first three methods can result in the same performance after the design of the same ideal sliding dynamics of the integral terminal sliding surfaces since the real output exhibits exactly as designed in the integral terminal sliding surfaces. An illustrative example and simulation study are provided to demonstrate the effectiveness of the proposed approaches. The three outputs generated are identical to the ideal outputs and the three ideal outputs are also identical. Therefore, the real robust output can be predesigned, predetermined, predicted by means of removing the reaching phase. The outputs of the three transformations are consistent with the design. The algorithms in this paper can be applied to controls of higher order SISO as well as MIMO plants and motion controls of dynamical systems. Further study will be conducted to investigate the remaining two part transformation approaches among the three sliding surface part transformations.

References

1 
V.I. Utkin, 1978, sliding Modes and Their Application in Variable Structure Systems, MoscowGoogle Search
2 
R.A. Decarlo, s.H. Zak, G.P. Mattews, 1988, Variable Structure Control of Nonlinear Multivariable Systems: A Tutorial, Proc. IEEE 76, pp. 212-232Google Search
3 
V. T. Haimo, 1986, Finite Time Controllers, sIAM Journal of Control and Optima, Vol. 24, pp. 760-770DOI
4 
M. Zak, 1988, Terminal Atractors for Addressable Memory in Neural Networks, Physics Letters A, Vol. 133, No. 1, 2, pp. 18-22DOI
5 
W. C. su, s. V. Drakunov, U. Ozguner, 1996, Constructing Discontinuity Surfaces for Variable Structure Systems: A Lyapunov Approach, Automatica, Vol. 32, No. 6, pp. 925-928DOI
6 
J. H. Lee, 2011, A Proof of Utkin’s Theorem for a SI Uncertain Linear Case, Journal of IEIE, Vol. 48, No. 6, pp. 475-481DOI
7 
J. H. Lee, 2015, A Continuous Sliding Surface Transformed VSS by Saturation Function for MIMO Uncertain Linear Plants, Journal of IEIE, Vol. 52, No. 7, pp. 1351-1358DOI
8 
J. H. Lee, 2019, A Continuous Control Input Transformed VSS by Modified Boundary Layer Function for MIMO Uncertain Linear Plants, Journal of IEIE, Vol. 56, No. 5, pp. 72-81DOI
9 
J. H. Lee, M. J. Youn, 1994, An Integral-Augmented Optimal Variable Structure control for Uncertain dynamical SISO System, KIEE(The Korean Institute of Electrical Engineers), Vol. 43, No. 8, pp. 1333-1351Google Search
10 
J. H. Lee, 2011, A Proof of Utkin’s Theorem for a SI Uncertain Integral Linear Case, Transaction of KIEE, Vol. 60, No. 4, pp. 843-847Google Search
11 
J. H. Lee, M. S. Choi, 2021, Control Input Transformed Integral SMCs with Output Prediction Performance for MIMO Uncertain Linear Plants, Transaction of KIEE, Vol. 70, No. 4, pp. 666-678DOI
12 
J. H. Lee, M. S. Choi, 2021, sliding Surface Transformed Integral SMCs with Output Prediction Performance for MIMO Uncertain Linear Plants, Transaction of KIEE, Vol. 70, No. 2, pp. 365-379DOI
13 
J. H. Lee, 2010, A Dynamic Output Feedback Variable Structure Controller for Uncertain Systems with Unmatched System Matrix Uncertainty, KIEE, Vol. 59, No. 11, pp. 2066-2072DOI
14 
J. H. Lee, 2010, A Poof of Utkin's Theorem for a MI Uncertain Linear Case, Transaction of KIEE, Vol. 59, No. 9, pp. 1680-1685DOI
15 
J. H. Lee, 2014, Poofs of Utkin's Theorem for MIMO Uncertain Integral Linear Systems, Journal of IKEEE, Vol. 18, No. 2, pp. 255-262DOI
16 
J. H. Lee, 2017, A Poof of Utkin's Theorem for Uncertain Nonlinear Plants, Transaction of KIEE, Vol. 66, No. 11, pp. 1612-1619DOI
17 
J. H. Lee, 2019, A Poof of Utkin's Theorem for ISMC of Uncertain Nonlinear Plants, Transaction of KIEE, Vol. 68, No. 3, pp. 460-470DOI
18 
M. Zhihong, A. P. Paplinski, H. R. Wu, 1994, A Robust MIMO Terminal Sliding Control Scheme for Rigid Robotic Manipulators, iEEE T. on Automatic Control, Vol. 39, No. 12, pp. 2464-2469DOI
19 
M. Zhihong, X. H. Yu, 1996, Terminal Sliding Mode Control of MIMO Linear System, Proceedigns of the 35th Conference on Decision and Control Kobe Japan, pp. 4619-4624DOI
20 
X. H. Yu, M. Zhihong, Y. Q. Wu, 1997, Terminal sliding Modes with Fast Transient Performance, Proceedigns of the 36th Conference on Decision and Control San Diego Califonia USA, pp. 962-963DOI
21 
X. Yu, M. Zhihong, Y. Feng, Z. Guan, 2002, Nonsingular Terminal Sliding Mode control of a Class of Nonlinear Dynamical Systems, iFAC 15th Triennial world Congrass Barcelona Spain, pp. 161-165DOI
22 
Y. Feng, X. Han, Y. Wang, X. Yu, 2007, second-Order Terminal Sliding Mode Control of Uncertain Multivariable Systems, int. Journal of Control, Vol. 80, No. 6, pp. 856-862DOI
23 
Q. Zong, Z. S. Zhao, J. Zhang, 2010, Higher Order Sliding Mode Control with Self-Tuning Law Based on Integral Sliding Mode, iET Control Theory Applications, Vol. 4, No. 7, pp. 1282-1289DOI
24 
C. S. Chiu, 2012, Derivative and Integral Terminal Sliding Mode Control for a Class of MIMO Nonlinear Systems, Automatica, Vol. 48, pp. 316-326DOI
25 
L. Peng, M. Jianjun, G. Lina, Z. Zhiqiang, 2015, integral Terminal Sliding Mode Control for Uncertain Nonlinear Systems, Proceedings of the 34th Chinese Control conference, pp. 824-828, Vol. hangzhou, pp. china-DOI
26 
H. T. song, T. Zhang, 2016, Fast Robust Integrated Guidance and Control Design of Interceptors, iEEE T. Control Systems Technology, Vol. 24, No. 1, pp. 349-356DOI
27 
R. Hu, H. Deng, Y. Zhang, 2020, Novel Dynamic-Sliding- Mode-Manifold-Based Continuous Fractional-Order Nonsingular Terminal Sliding Mode Control for a Class of Second-Order Nonlinear Systems, iEEE Access, Vol. 8, pp. 19820-19829DOI
28 
P. Lesniewski, 2017, sliding Mode Control with Time-Varying Sliding Hyperplanes:A Survey, 18th Int. Carpathian Control Conference(ICCC), pp. 81-86DOI
29 
Y. Feng, X. H. Yu, F. L. Han, 2013, On Nonsingular Terminal Sliding-Mode Control of Nonlinear Systems, Automatica, Vol. 49, pp. 1715-1722DOI
30 
s. S. D. Xu, C. C. Chen, Z. L. Wu, 2015, study of Nonsingular Fast Terminal Sliding Mode Fault Tolerant Control, iEEE T. Industrial Electronics, Vol. 62, No. 6, pp. 3906-3913DOI
31 
J. H. Lee, M. S. Choi, 2022, Control Input Transformed Terminal Integral SMCs with Output Prediction for Second Order Uncertain Plants, Transaction of KIEE, Vol. 71, No. 2, pp. 412-423DOI
32 
M. S. Choi, J. H. Lee, J. G. Kim, 2023, Proof of Utkin’s Theorem for TSMC of Second Order Uncertain Plants, Journal of IEIE, Vol. 59, No. 4Google Search

저자소개

최명수 (Myeong-Soo Choi)
../../Resources/kiee/KIEE.2023.72.5.628/au1.png

1982년 7월 5일생. 2007년 경상대학교 제어계측공학과 졸업(공학사),

2010년 경상국립대학교 제어계측공학과 졸업(석사).

2012년 경상국립대학교 제어로봇공학과 박사수료.

현재 경상국립대학교 공과대학 제어로봇공학과 박사과정.

Tel:+82-10-5736-6618

Fax:+82-55-772-1749

E-mail : jaya0419@gnu.ac.kr

이정훈 (Jung-Hoon Lee)
../../Resources/kiee/KIEE.2023.72.5.628/au2.png

1966년 2월 1일생.

1988년 경북대학교 전자공학과 졸업(공학사),

1990년 한국과학기술원 전기 및 전자공학과 졸업(석사).

1995년 한국과학기술원 전기 및 전자공학과 졸업(공박).

1995년~현재 경상국립대학교 공과대학 제어로봇공학과 교수.

경상국립대학교 공대 공학연구원 연구원.

1997-1999 경상국립대학교 제어계측공학과 학과장.

마르퀘스사의 Who's Who in the world 2000년 판에 등재.

American Biograhpical Institute(ABI)의 500 Leaders of Influence에 선정.

Tel:+82-55-772-1742

Fax:+82-55-772-1749

E-mail : jhleew@gnu.ac.kr