Abstract
We consider in this paper switched systems, a class of hybrid systems recently used with success in various domains such as automotive industry and power electronics. We propose a state-dependent control strategy which makes the trajectories of the analyzed system converge to finite cyclic sequences of points. Our method relies on a technique of decomposition of the state space into local regions where the control is uniform. We have implemented the procedure using zonotopes, and applied it successfully to several examples of the literature and industrial case studies in power electronics.
Similar content being viewed by others
Notes
We will sometimes denote such a trajectory under the form: \(x_0\rightarrow _{{\uppi }_{i_1}}x_1\rightarrow _{{\uppi }_{i_2}}\cdots \) with \(i_1,i_2,\ldots \in I\).
More formally, no point will belong to a \(W_j\) for \(j \in J \setminus J'\) if it does not belong to a \(W_{j'}\) for \(j' \in J'\).
Note that the number of simulated patterns in the table can be larger than the number of possible patterns for given \(n\) and \(k\), since many patterns will be simulated multiple times for different tiles.
For \({\uppi }=u_1\ldots u_m\), the operator \(Pre_{{\uppi }}\) is defined by: \(Pre_{\uppi }(X)=\{x' \ |\ x'\rightarrow ^{u_1}_{\tau }\cdots \rightarrow _{\tau }^{u_m} x \text{ for } \text{ some } x\in X\}\).
Note that we consider here a box \(R\), which is smaller than the original box \(R=[1.55,2.15]\times [1.0,1.4]\) used in example 3. This is because the safety zone \(S = [1.7,2.0]\times [1.10,1.30]\) considered here is itself included in the original \(R\).
The associated patterns are: \({\uppi }_1 = (1 1 2 2 1 2 2 1 2 2)\), \({\uppi }_2 = (1 2 1 2 1 2 2 2)\), \({\uppi }_3 = (1 2 1 2 2 1 2 2)\), \({\uppi }_4 = (1 2 2)\), \({\uppi }_5 = (2)\), \({\uppi }_6 = (1 2)\), \({\uppi }_7 = (1 2)\), \({\uppi }_8 = (1)\), \({\uppi }_9 = (1)\), \({\uppi }_{10} = (1)\), \({\uppi }_{11} = (1 2)\), \({\uppi }_{12} = (1 2)\), \({\uppi }_{13} = (2)\), \({\uppi }_{14} = (2)\), \({\uppi }_{15} = (1 2)\), \({\uppi }_{16} = (2 2 1)\).
The associated patterns are: \({\uppi }_1 = (-10 \times -10 \times -10\times -10\times -10\times 0\times 10)\), \({\uppi }_2 = (-10)\), \({\uppi }_3 = (0)\), \({\uppi }_4 = (-10 \times -10\times -10\times 10)\), \({\uppi }_5 = (-10)\), \({\uppi }_6 = (0)\), \({\uppi }_7 = (-10)\), \({\uppi }_8 = (10 \times 10\times 0\times 0)\), \({\uppi }_9 = (0)\), \({\uppi }_{10}= (10 \times 10\times 10\times 10\times 0\times 10\times -10)\).
The associated patterns are: \({\uppi }_1 = (1 0 1 0)\), \({\uppi }_2 = (1)\), \({\uppi }_3 = (1 0)\), \({\uppi }_4 = (0)\), \({\uppi }_5 = (0)\), \({\uppi }_6 = (1)\), \({\uppi }_7 = (0)\), \({\uppi }_8 = (0)\), \({\uppi }_9 = (0)\), \({\uppi }_{10} = (1 0)\).
The corresponding patterns are: \({\uppi }_1 = (1 1 2 2 1 2 1 2 2 2)\), \({\uppi }_2 = (1 1 2 2 1 2 2 1 2 1 2 2 2)\), \({\uppi }_3 = (2 1 1 2 1 2 2 2)\), \({\uppi }_4 = (1 2 1 2 1 2 2 2)\), \({\uppi }_5 = (1 2 2)\), \({\uppi }_6 = (2)\), \({\uppi }_7 = (1 2)\), \({\uppi }_8 = (1 2)\), \({\uppi }_9 = (1 2)\), \({\uppi }_{10} = (1 2)\), \({\uppi }_{11} = (1)\), \({\uppi }_{12} = (1)\), \({\uppi }_{13} = (1)\), \({\uppi }_{14} = (1 2)\), \({\uppi }_{15} = (1 2)\), \({\uppi }_{16} = (2 1 2 2 1)\).
References
Asarin E, Bournez O, Dang T, Maler O, Pnueli A (2000) Effective synthesis of switching controllers for linear systems. Proc IEEE, Spec Issue Hybrid Syst 88(7):1011–1025
Beccuti A, Papafotiou G, Morari M (2005) Optimal control of the boost dc-dc converter. In: Proceedings of 44th IEEE conference on decision and control European control conference (CDC-ECC ’05), 4457–4462, Dec 2005
Buisson J, Richard P-Y, Cormerais H (2005) On the stabilisation of switching electrical power converters. In: HSCC, vol. 3414 LNCS, pp 184–197
Cámara J, Girard A, Gößler G (2011) Safety controller synthesis for switched systems using multi-scale symbolic models. In: Proceedings of 50th IEEE conference on decision and control (CDC-ECC ’11), 520–525
Corronc EL, Girard A, Goessler G (2013) Mode sequences as symbolic states in abstractions of incrementally stable switched systems. In: CDC, IEEE, pp 3225–3230
Ding J, Li E, Huang H, Tomlin CJ (2011) Reachability-based synthesis of feedback policies for motion planning under bounded disturbances. In: IEEE international conference on robotics and automation (ICRA’11), pp 2160–2165
Feld G, Fribourg L, Labrousse D, Revol B, Soulat R (2012) Correct by design control of 5-level and 7-level converters. Research Report LSV-12-25, LSV, ENS Cachan, France, Dec 2012
Girard A (2012) Low-complexity switching controllers for safety using symbolic models. In: Heemels M, De Schutter B, Lazar M (eds) 4th IFAC conference on analysis and design of hybrid systems, June, 2012. Eindhoven, Pays-Bas, 2012
Girard A, Pola G, Tabuada P (2010) Approximately bisimilar symbolic models for incrementally stable switched systems. IEEE Trans Autom Control 55:116–126
Henzinger TA (1996) The theory of hybrid automata. In Proceedings of the 11th Annual IEEE Symposium on Logic in Computer Science, LICS ’96, IEEE Computer Society, Washington, DC, USA pp 278–292, 1996
Hiskens IA (2001) Stability of limit cycles in hybrid systems. In: Proceedings of the 34th Hawaii International conference on system sciences, pp 163–328
Kühn W (1998) Zonotope dynamics in numerical quality control. In: Hege H-C, Polthier K (eds) Mathematical visualization. Springer Verlag, Berlin, pp 125–134
Liberzon D, Morse AS (1999) Basic problems in stability and design of switched systems. IEEE Control Syst Mag 19:59–70
Lygeros J, Tomlin C, Sastry S (1999) Controllers for reachability specifications for hybrid systems. Automatica 35(3):349–370
Majumdar R, Zamani M (2012) Approximately bisimilar symbolic models for digital control systems. In: Madhusudan P, Seshia SA (eds) CAV, Lecture Notes in Computer Science, vol 7358. Springer, Berlin, pp 362–377, 2012
Ocaml Team. OCaml Web page. http://caml.inria.fr/ocaml/index.fr.html, 2013
Octave Web page. http://www.gnu.org/software/octave/
PPL Team. PPL Web page. http://bugseng.com/products/ppl/, 2013
Ramadge PJ, Wonham WM (1989) The control of discrete event systems. Proc IEEE 77(1):81–98
Reißig G (2009) Computation of discrete abstractions of arbitrary memory span for nonlinear sampled systems. In: Majumdar R, Tabuada P (eds) HSCC, Lecture Notes in Computer Science vol 5469. Springer, Berlin pp 306–320, 2009
Senesky M, Eirea G, Koo T-J (2003) Hybrid modelling and control of power electronics. In HSCC, Lecture Notes in Computer Science, vol 2623, pp 450–465
Acknowledgments
We are grateful to the anonymous referees for their numerous helpful remarks.
Author information
Authors and Affiliations
Corresponding author
Additional information
A short version of this paper appears in Proc. Reachability Problems 2013, LNCS, vol. 8169, Springer, pages 135-145 under the title “Stability Controllers for Sampled Switched Systems”.
Appendices
Appendix 1: Sufficient condition of decomposition
Given a zone \(R\), an invariant decomposition does not always exist. We give hereafter some geometrical conditions on the position of \(R\) that guarantee the decomposability of \(R\) when all the modes of the system are contractive, in the sense of Definition 1: For each \(u\in U\), there exists \(0<\beta _u<1\) such that, for all \(x,y\in \mathbb {R}^n\):
For the sake of simplicity, we suppose that the switched system has only \(|U|=2\) modes, and the state space dimension is \(n=2\), but the reasoning extends to larger values of \(|U|\) and \(n\). We assume that matrix \(A_u\) associated with mode \(u\) (\(u=1,2\)) is invertible. Let \(e_u = -A_u^{-1}B_u\) be the unique attractive equilibrium point associated with mode \(u\) (\(u=1,2\)). Let us define the “pure” switching rule \(\mathcal{S}_u\) (\(u=1,2\)) which applies repeatedly mode \(u\) to any point \(x \in \mathbb {R}^2\). Let \(\mathcal{C}_1\) (resp. \(\mathcal{C}_2\)) be the \(\tau \)-sampled trajectory issued from \(e_1\) (resp. \(e_2\)) under \(\mathcal{S}_2\) (resp. \(\mathcal{S}_1\)) (i.e., \(\mathcal{C}_1 = Post^{*}_{2}(e_1)\) and \(\mathcal{C}_2 = Post^{*}_{1}(e_2)\)). Since each mode is contractive, and \(e_u\) is the unique equilibrium point associated with mode \(u\), any trajectory under control \(\mathcal{S}_u\) converges to the equilibrium point \(e_u\) (\(u=1,2\)), whatever the starting point of \(\mathbb {R}^2\). In particular, \(\mathcal{C}_1\) ends to \(e_2\) and \(\mathcal{C}_2\) to \(e_1\), as depicted in Fig. 11 for the boost converter (see Examples 1 and 2).
Theorem 4
Let \(\varSigma \) be a sampled switched affine system as defined above. Suppose that the reference point \(O\) is in \(\mathcal{C}_1\cup \mathcal{C}_2\). If \(R\subseteq \mathbb {R}^2\) is a box whose interior contains \(O\), then there exists a positive integer \(k\) such that \(R\) is \(k\)-invariant.
Proof
Suppose \(O\in \mathcal{C}_2\). (The case \(O\in \mathcal{C}_1\) is symmetrical.) Consider a box \(R\) of interior \(\mathring{R}\) with \(O\in \mathring{R}\). There exists \(\Delta _O > 0\) such that \(\mathcal{B}(O,\Delta _O) \subseteq R\). Since \(O \in \mathcal{C}_2\), we have:
-
(a)
\(e_2 \rightarrow ^{{\uppi }_1} O\), for some pattern \({\uppi }_1\in (1)^*\). Furthermore, for all \(x\in R\), we have:
-
(b)
\(x\rightarrow ^{{\uppi }_2} x_1\) for some \(x_1\in \mathcal{B}(e_2,\Delta _O)\) and some pattern \({\uppi }_2\in (2)^*\), because \(e_2\) is an attractive equilibrium point;
-
(c)
\(x_1\rightarrow ^{{\uppi }_1} x_2\) for some \(x_2\in \mathcal{B}(O,\Delta _O)\), because of (a) and because mode \(1\) is contractive. This is depicted in Fig. 12. It follows from (b)-(c) that, for all \(x \in R\): \(x\rightarrow ^{{\uppi }_x} x_2\) for some \(x_2\in \mathcal{B}(O,\Delta _O)\) and some pattern \({\uppi }_x\in (2^*1^*)\). Hence, we have: \(\mathcal{B}(x_2,\Delta _x) \subseteq R\) for some \(\Delta _x > 0\). Since, for any \({\uppi }_x\), \(Post_{{\uppi }_x}\) is continuous, \(Pre_{{\uppi }_x}(\mathcal{B}(x_2,\Delta _{x}))\) Footnote 5 is an open subset of \(\mathbb {R}^2\) containing \(x\). Since \(R\) is a compact of \(\mathbb {R}^2\), from the set \(C=\{Pre_{{\uppi }_x}(\mathcal{B}(Post_{{\uppi }_x}(x),\Delta _{x}))\}_{x\in R}\), one can extract by Heine-Borel’s theorem, a subset \(C'=\{Pre_{{\uppi }_{x_i}}(\mathcal{B}(Post_{{\uppi }_{x_i}}(x_i),\Delta _{x_i}))\}_{i\in I}\), for some finite set of indices \(I\), such that \(C'\) covers \(R\) and \(\mathcal{B}(x_i,\Delta _{x_i})\) is \(R\)-invariant via \({\uppi }_i\). This means that \(C'\cap R\) is a \(k\)-invariant decomposition of \(R\) of the form \(\{(V_i,{\uppi }_i)\}_{i=1,\ldots ,m}\), where \(m\) is the cardinal of \(I\), \(k\) the maximum length of \({\uppi }_1,\ldots ,{\uppi }_m\), and \(V_i=\mathcal{B}(x_i,\Delta _{x_i})\cap R\) is such that \(\bigcup _{i=1}^m V_i = R\) and \(V_i\) is \(R\)-invariant via \({\uppi }_i\) (\(1\le i\le m\)). \(\square \)
The theorem gives an interesting locality condition on \(O\) (location on one of the “pure” trajectories linking the equilibrium points), for ensuring the existence of \(k\)-invariant boxes \(R\). This justifies a posteriori the existence of a decomposition for the zone \(R\) of the DC–DC converter example 2 since it overlaps trajectory \(\mathcal{C}_1\). Note also that \(R\) can be arbitrarily small as far as it intersects \(\mathcal{C}_1\) (or \(\mathcal{C}_2\)).
Appendix 2: Case studies
We now apply the decomposition procedure (enhanced for safety properties) for several examples. For each example, we will generate a \(k\)-invariant decomposition \(\Delta \) of \(R\) satisfying \( Unf _{\Delta }(R)\subseteq S\), thus proving that \( Unf _{\Delta }(R)\) is a controlled invariant, and that the system is safe.
Example 4
Let us come back to the boost DC–DC converter of example 1. For \(R\), we now take the box \([1.75,1.95]\times [1.14,1.26]\),Footnote 6 which corresponds to a medium value \(1.85\) for \(i_l\) with \(\pm 0.1\) for variability, and medium value \(1.20\) for \(v_c\) with \(\pm 0.06\) for variability. For the safety region, we take \(S = [1.7,2.0]\times [1.10,1.30]\), which corresponds to an additional variability of \(\pm 0.05\) for \(i_l\) and \(\pm 0.04\) for \(v_c\).
The application of the algorithm Decomposition to \(R\) and \(S\), with \(k=10\) and \(d=4\) succeeds, yielding a \(k\)-invariant decomposition \(\Delta \) of the form \(\{(V_j,{\uppi }_j)\}_{j=1,\ldots ,16}\) of \(R\) Footnote 7 satisfying \( Unf _{\Delta }(R)\subseteq S\). The \(k\)-invariant decomposition \(\Delta \) of \(R\) is depicted in Fig. 13. The \(\Delta \)-unfolding of \(R\) is depicted in Fig. 14. It is divided into regions of two different colors corresponding to the different control modes: blue (resp. red) indicates that mode 1 (resp. 2) should be applied. The safety zone \(S = [1.7,2.0]\times [1.10,1.30]\) is delimited in green.
The controlled system has been implemented using Octave [17]. The unfolded \(\Delta \)-trajectory of the system starting at point \((1.75,1.26)\), is depicted in Fig. 15.
Other applications of the decomposition procedure are described in Sect. 5.
Example 5
(Helicopter motion [6]) The problem is to control a quadrotor helicopter to some position on top of a stationary ground vehicle, while satisfying constraint on the relative velocity. By controlling the pitch and roll angles, one can modify the speed and the position of the helicopter. Let \(g\) be the gravitational constant, \(x\) (resp. \(y\)) the position according to \(x\)-axis (resp. \(y\)-axis), \({\dot{x}}\) (resp. \({\dot{y}}\)) the velocity according to \(x\)-axis (resp. \(y\)-axis), \(\phi \) the pitch command, \(\psi \) the roll command. The possible commands for the pitch and the roll are: \(\phi ,\psi \in \{-10,0,10\}\). Since each mode corresponds to a pair \((\phi ,\psi )\), there are 9 modes. The dynamics are given by the equation:
where \(X\) is \((x,{\dot{x}},y,\dot{y})^{T}\). The sampling period is \(\tau = 0.1\). The variables \(x\) and \(y\) are decoupled in the equations and follow the same equations, so we focus on the control for \(x\). We take \(R=[-0.3,0.3]\times [-0.5,0.5]\) (i.e., \(R = \{(x_1,x_2) \, |\, x_1 \in [-0.3,0.3], x_2 \in [-0.5,0.5]\}\)). This corresponds to an equilibrium zone centered at the state \((0,0)\) of the ground vehicle, and a variability of \(\pm 0.3\) for position and \(\pm 0.5\) for velocity. We take \(S = [-0.4,0.4]\times [-0.7,0.7]\) for the safety region, which corresponds to an additional variability of \(\pm 0.1\) for position and velocity. As for \(R\), we take a box guessed manually from the equilibrium zone appearing in the numerical experiments and simulations of [6]. We take the same safety zone \(S\) as the one specified in [6].
The application of algorithm Decomposition to \(R\) and \(S\) with \(k=6\) and \(d=4\) succeeds, yielding a \(k\)-invariant decomposition \(\Delta \) of \(R\) of the form \(\{(V_i,{\uppi }_i)\}_{i=1,\ldots ,10}\) Footnote 8 satisfying \( Unf _{\Delta }(R)\subseteq S\). The decomposition \(\Delta \) is represented in Fig. 16. The unfolding of \(R\) is depicted in Fig. 17. The unfolding is divided into regions of three different colors corresponding to the different control modes: color red (resp. blue, green) indicates that mode \(10\) (resp. \(-10\), \(0\)) should be applied. The safety zone \(S= [-0.4,0.4]\times [-0.6,0.6]\) in delimited in green.
The controlled system has been simulated using Octave. Figure 18 gives a simulation of the system represented in plane \((x, {\dot{x}})\), starting at point \((-0.3, 0.5)\).
Example 6
(Two-room building heating [8]) This is a simple heating model of a two-room building. Let \(T=(T_1,T_2)^{T}\) be the state variable, where \(T_i\) is the temperature of room \(i\) (\(i=1,2\)). The dynamics are given by the equation: \(\dot{T} = \begin{pmatrix} -\alpha _{21}-\alpha _{e1}-\alpha _{f}p &{} \alpha _{21}\\ \alpha _{12} &{} -\alpha _{12} - \alpha _{e2} \end{pmatrix} \times T + \begin{pmatrix} \alpha _{e1}T_e + \alpha _{f}T_fp\\ \alpha _{e2}T_e \end{pmatrix}\), where \(p\) is a mode of value \(0\) or \(1\), and the heat transfer coefficients and external temperatures are given by the values: \(\alpha _{12} = 5.10^{-2}\), \(\alpha _{21} = 5.10^{-2}\), \(\alpha _{e1} = 5.10^{-3}\), \(\alpha _{e2} = 3.3.10^{-3}\), \(\alpha _f = 8.3.10^{-3}\), \(T_e = 10\), \(T_f = 50\). The sampling period is \(\tau = 5\). We take \(R=[20.25,21.75]\times [20.25,21.75]\). This corresponds to an equilibrium zone centered at state \((21,21)\) with a variability of \(\pm 0.75\). For the safety zone, we take \(S = [20,22]\times [20,22]\), as in [8].
The application of the algorithm Decomposition to \(R\) and \(S\) with \(k=4\) and \(d=2\) succeeds, yielding a \(k\)-invariant decompositionFootnote 9 \(\Delta \) of the form \(\{(V_j,{\uppi }_j)\}_{j=1,\ldots ,10}\) of \(R\) satisfying \( Unf _{\Delta }(R)\subseteq S\). The decomposition \(\Delta \) is depicted in Fig. 19. The unfolding of \(R\) is represented on Fig. 20. The unfolding is divided into regions of two colors corresponding to the different control modes: the red (resp. blue) color indicates that control 0 (resp. 1) should be applied. The safety zone \(S\) in delimited green.
The controlled system has been simulated using Octave [17]. A simulation is depicted in Fig. 21 for starting temperature point \((20.25,21.75)\).
Example 7
(Multilevel converter) Using the decomposition procedure, we have synthesized a control for a case study originating from power electronics industry of high dimension (\(n=7\)), satisfying a safety constraint. The control has been implemented and successfully physically experimented with a prototype. See [7].
Appendix 3: Control with disturbances
Zonotopes are useful for representing and manipulating efficiently convex polytopes. They allow us to extend easily the decomposition procedure in order to allow for small perturbations of the system dynamics (see, e.g., [12]). Using zonotopes, it is possible to introduce uncertainty or disturbance in our model. All the dynamics of the system are now of the form \({\dot{x}} = A_{u}x+B_u + \varepsilon \) where \(\varepsilon \) represents disturbance under the form of a \(n\)-dimensional vector belonging to a given box \(\varOmega \) of fixed size. We will use \(\mathbf{x}(t,x,u,\varepsilon )\) to denote the point reached by the system at time \(t\) under mode \(u\) and disturbance \(\varepsilon \) from the initial condition \(x\). This defines a transition relation \(\rightarrow ^{u,\varepsilon }_{\tau }\) given by:
We define \(Post_u(X,\varOmega ) = \{x' \, | \, \exists \varepsilon \in \varOmega , \, x \rightarrow _{\tau }^{u,\varepsilon } x'\}\). This definition naturally extends to \(Post_{{\uppi }}\) where \({\uppi }\) is a pattern. In order to handle disturbance, one replaces test \(Post_{{\uppi }}(W) \subseteq R\) of algorithm \(\hbox {Find}\_\hbox {Pattern}\) by \(\square (Post_{{\uppi }}(W,\varOmega )) \subseteq R\), where ‘\(\square \)’ is the operator mapping a zonotope into the smallest enclosing box.
Example 8
(Boost converter) We consider the dynamics of the boost DC–DC converter (see Examples 1 and 2) in presence of disturbances belonging to \(\varOmega = \{0\} \times \left[ \frac{-0.064}{x_l},\frac{0.064}{x_l}\right] \). These disturbances correspond to noise on the input voltage. They represent represent up to \(8 \%\) of the value of the input voltage. With such disturbances, we have not found a control preserving the safety zone \(S\) defined in example 2, and we have taken a larger (i.e., more tolerant) safety zone defined by \(S' = [1.65,2.05] \times [1.10,1.30]\). The decomposition procedure then succeeds for \(k=13\) and a \(d=5\): it generates a \(k\)-invariant decomposition \(\Delta '\) of \(R\) satisfying \( Unf _{\Delta '}(R)\subseteq S'\). The decompositionFootnote 10 \(\Delta '\) is depicted in Fig. 22. A simulation of the system starting at \((1.75,1.26)\) is depicted in Fig. 23.
Example 9
(Helicopter motion) As done in [6], we will now solve the control problem with bounded disturbances to take into account a potential real-life environment. We use the same safe zone \(S\) as in example 5 and add the following possible disturbances \(\varepsilon \in [-0.02,0.02]\times [-0.1,0.1]\). The decomposition procedure succeeds, and generates a new \(k\)-invariant decomposition \(\Delta '\) with \( Unf _{\Delta '}(R)\subseteq S\). The decomposition \(\Delta '\) is depicted in Fig. 24. A simulation of a run starting at point \((-0.3,0.5)\) is presented in Fig. 25.
Appendix 4: Example with \(|R_{\Delta }^{\infty }| = \infty \)
For this example, we use modes associated to repulsive homothetic transformation. There are \(4\) modes such that, close to each corner of the global box \(R = [-1,1] \times [-1,1]\), there exists a fixed point for one of the mode. We take for the dynamics of the modes: \(A_1 = A_2 = A_3 = A_4 = \begin{pmatrix} 1.5 &{} 0 \\ 0 &{} 1.5 \end{pmatrix}\), \(B_1 = \begin{pmatrix} 0.6 \\ 0.6 \end{pmatrix}\), \(B_2 = \begin{pmatrix} -0.6 \\ 0.6 \end{pmatrix}\), \(B_3 = \begin{pmatrix} -0.6 \\ -0.6 \end{pmatrix}\), and \(B_4 = \begin{pmatrix} 0.6 \\ -0.6 \end{pmatrix}\). The \(\Delta \)-decomposition is presented in Fig. 26. We have \(V_1 = [-1,0] \times [-1,0]\) associated to pattern \({\uppi }_1 = (1)\), \(V_2 = [0,1] \times [-1,0]\) associated to pattern \({\uppi }_2 = (2)\), \(V_3 = [0,1] \times [0,1]\) associated to pattern \({\uppi }_3 = (3)\), and \(V_4 = [-1,0] \times [0,1]\) associated to pattern \({\uppi }_4 = (4)\). A \(\Delta \)-trajectory is depicted in Fig. 27. One can see that the system appears to be chaotic and will not converge to a limit cycle.
Appendix 5: Visualizations of the iteration of \(Post_{\Delta }\)
In Fig. 28, we give the iterated images \(Post_{\Delta }^k(R)\) of the boost converter benchmark for \(k=0,10,20,40,80,100\). We see that for \(k=100\), \(Post_{\Delta }^k(R)\) is completely located inside \(V_1\). The controlled system converges to a single limit point.
Figure 29 shows the iterated images \(Post_{\Delta }^k(R)\) of the two tanks benchmark for \(k=0,5,10,15,20,25\). In the figures, it can be seens that starting with \(k=10\), the set \(Post_{\Delta }^k(R)\) does not intersect the borders of the decomposition \(\Delta \). The system converges to a limit cycle (see also Fig. 10 on page 16).
Rights and permissions
About this article
Cite this article
Fribourg, L., Kühne, U. & Soulat, R. Finite controlled invariants for sampled switched systems. Form Methods Syst Des 45, 303–329 (2014). https://doi.org/10.1007/s10703-014-0211-2
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10703-014-0211-2