Dynamic anti-windup method for a class of time-delay control systems with input saturation (Q4487008)

The aim of this paper is to give an extension of the anti-windup concept to time-delay systems. First, one formulates the problem related to time-delay control systems with input saturation and presents a dynamic anti-windup scheme. The plant and controller are expressed by the following minimal state space representations: \[ e^{-s\tau_1} P(s):\begin{cases} \dot x_p(t)= Ax_p(t)+ Bu(t-\tau_1)\\ y(t)=Cx_p(t)+ Du(t-\tau_1)\\ u(t)=\sigma \bigl(v(t)),\;\sigma\bigl(v(t)\bigr)\in {\mathcal B}_{\mathbb{R}^m}\end{cases} \] \[ K(s):\begin{cases} \dot x_c(t)=Fx_c(t)+Ge(t)\\ v(t)=Hx_c(t)+Le(t)\\ e(t)=r(t)-y(t-\tau_2)-d(t-\tau_2),\end{cases} \] where \(r(t)\in\mathbb{R}^n\) is the reference input, \(u(t)\in \mathbb{R}^m\) is the control input, \(y(t)\in\mathbb{R}^n\) is the output, \(v(t)\in\mathbb{R}^m\) is the controller output, \(d(t)\in \mathbb{R}^n\) is the disturbance, \(x_p(t)\in \mathbb{R}^p\) is the state of the plant and \(x_c(t)\in\mathbb{R}^2\) is the state of the controller. The constant matrices \(A,B,C,D,F,G,H\) and \(L\) have appropriate dimension, \({\mathcal B}_{\mathbb{R}^m}\) is the bounded subspace of \(\mathbb{R}^m\) whose boundary is shaped by the saturation limit, and the continuous mapping \(\sigma (\cdot): \mathbb{R}^m\to {\mathcal B}_{\mathbb{R}^m}\) denotes the control input saturation. The main objective is to provide an additional dynamic compensator \(M(s,\tau_1, \tau_2)\) which gives less performance degradation due to control input saturation, having the following state-space representation \[ M(s,\tau_1, \tau_2):\begin{cases} \dot x_d(t)=\Phi (t,\tau_1,\tau_2) x_d(t)+\Gamma(t, \tau_2, \tau_2) \zeta(t)\\ \xi(t)=\Sigma(t, \tau_2,\tau_2) x_d(t)+\Lambda(t, \tau_1, \tau_2)\zeta(t)\\ \zeta(t)=v(t)-u(t), \end{cases} \] where the matrices \(\Phi (\cdot, \cdot,\cdot)\), \(\Gamma(\cdot, \cdot,\cdot)\), \(\Sigma(\cdot,\cdot,\cdot)\) and \(\Lambda(\cdot,\cdot,\cdot)\) are design parameters. Then, the dynamics of the compensated controller are as follows: \[ K_c(s):\begin{cases} \dot x_c(t)= Fx_c(t)+ Ge(t)-\xi(t)\\ v(t)=Hx_d(t)+Le(t)\\ e(t)=r(t)-y(t-\tau_2)-d(t-\tau_2), \end{cases} \] where \(\xi(t)\in\mathbb{R}^2\) is the compensation signal. The authors find the dynamic compensator \(M^*(s,\tau_1,\tau_2)\) which minimizes a performance index \(J\), namely: \[ \min_{M(s, \tau_1,\tau_2)} J =\int^\infty_0 \bigl\|x_c(t)- \widetilde x_c(t)\bigr \|^2dt, \] where \(x_c(t)\) and \(\widetilde x_c(t)\) are the controller states in the absence and presence of input saturation, respectively. The role of the anti-windup scheme with this dynamic compensator is to maintain the controller states exactly to be the same as those without input saturation. On the other hand, the proposed anti-windup scheme guarantees the total stability of the resulting system. Next, the authors extend the proposed anti-windup method to the time-delay control systems with input saturation in which controllers contain the delay elements. Finally, some illustrative examples are presented.