Abstract
Under a regularity assumption we prove that reachability in fixed time for nonlinear control systems is robust under control sampling.
Similar content being viewed by others
Notes
This regularity assumption on f can be relaxed at several places in the paper (see Remark 3.7 for details).
With respect to the existing literature, we add the word “strongly", in contrast to the notion of “weakly" regular control defined in Sect. 2.3.
In the literature, usually the \(\mathrm {U}\)-Pontryagin cone of a control \(u \in {\mathcal {U}}\cap \mathrm {L}^\infty ([0,T],\mathrm {U})\) is defined as the smallest closed convex cone containing all strong \(\mathrm {U}\)-variation vectors associated with u (see, e.g., [21]). As explained in Remark 2.3, considering the closure (or not) has no impact on the notions and results presented in this paper. Nevertheless we emphasize that the multiple needle-like variations of the control u (see Sect. 4.5) generate (only) the \(\mathrm {U}\)-Pontryagin cone of u as defined in Definition 2.5 (i.e., without closure).
This fact can also be derived from the Hamiltonian characterizations.
This fact is obvious when u belongs to \(\mathrm {Int}(\mathrm {L}^\infty ([0,T],\mathrm {U}))\) since then \({\mathcal {T}}_{\mathrm {L}^\infty _\mathrm {U}}[u] = \mathrm {L}^\infty ([0,T],{\mathbb {R}}^m)\). However, note that the inclusion \(\mathrm {Int}(\mathrm {L}^\infty ([0,T],\mathrm {U})) \subset \mathrm {L}^\infty ([0,T],\mathrm {Int}(\mathrm {U}))\) may be strict (take \(T=m=1\), \(\mathrm {U}=[0,1]\) and \(u(t)=t\) for a.e. \(t \in [0,T]\)).
For example, take \(n=m=1\), \(s=2\) and \(f(x,u,t)=u^4\) for all \((x,u,t) \in {\mathbb {R}}\times {\mathbb {R}}\times [0,T]\). Then considering \(\mathrm {L}^2\)-controls makes no sense.
For example, when \(m=1\), the set \((-\infty ,0] \cup [1,+\infty )\) is strongly nonconnected, while the set \((-\infty ,0) \cup (0,+\infty )\) is nonconnected (but not strongly).
Up to the nullity of the maximized Hamiltonian function due to the free final time and autonomous context considered in [33].
References
Agrachev A, Sachkov YL (2004) Control theory from the geometric viewpoint, volume 87 of Encyclopaedia of mathematical sciences. Springer-Verlag, Berlin, Control Theory and Optimization, II
Agrachev AA, Caponigro M (2010) Dynamics control by a time-varying feedback. J Dyn Control Syst 16(2):149–162
Athans M, Falb PL (1966) Optimal control. McGraw-Hill Book Co., New York-Toronto, Ont.-London, An introduction to the theory and its applications
Bian W, Webb JRL (1999) Constrained open mapping theorems and applications. J Lond Math Soc 60(3):897–911
Bonnard B, Chyba M (2003) Singular trajectories and their role in control theory, vol 40. Mathématiques & Applications (Berlin) [Mathematics & Applications]. Springer-Verlag, Berlin
Boscain U, Piccoli B (2004) Optimal syntheses for control systems on 2-D manifolds, vol 43. Mathématiques & Applications (Berlin) [Mathematics & Applications]. Springer-Verlag, Berlin
Bourdin L, Trélat E (2017) Linear-quadratic optimal sampled-data control problems: convergence result and Riccati theory. Automatica J IFAC 79:273–281
Brammer RF (1972) Controllability in linear autonomous systems with positive controllers. SIAM J Control 10:339–353
Chabi EH, Zouaki H (2000) Existence of a continuous solution of parametric nonlinear equation with constraints. J Convex Anal 7(2):413–426
Coron J-M (2007) Control and nonlinearity, vol 136. American Mathematical Society, Providence, RI, Mathematical Surveys and Monographs
Fuller AT (1960) Relay control systems optimized for various performance criteria. IFAC Proceedings Volumes, 1(1):520–529, 1st International IFAC Congress on Automatic and Remote Control. USSR, Moscow, p 1960
Gamkrelidze RV (1978) Principles of optimal control theory. Plenum Press, New York-London, revised edition, 1978. Translated from the Russian by Karol Malowski, Translation edited by and with a foreword by Leonard D. Berkovitz, Mathematical Concepts and Methods in Science and Engineering, Vol. 7
Grasse KA (1981) Perturbations of nonlinear controllable systems. SIAM J Control Optim 19(2):203–220
Grasse KA (1983) Some topological covering theorems with applications to control theory. J Math Anal Appl 91(2):305–318
Grasse KA (1984) On accessibility and normal accessibility: the openness of controllability in the fine \(C^{0}\) topology. J Differ Equ 53(3):387–414
Grasse KA (1992) On the relation between small-time local controllability and normal self-reachability. Math Control Signals Syst 5(1):41–66
Grasse KA (1995) Reachability of interior states by piecewise constant controls. Forum Math 7(5):607–628
Grasse KA, Sussmann HJ (1990) Global controllability by nice controls. Nonlinear controllability and optimal control, vol 133. Monogr. Textbooks Pure Appl. Math. Dekker, New York, pp 33–79
Klamka J (1996) Constrained controllability of nonlinear systems. J Math Anal Appl 201(2):365–374
Krener AJ, Schättler H (1989) The structure of small-time reachable sets in low dimensions. SIAM J Control Optim 27(1):120–147
Lee EB, Markus L (1967) Foundations of optimal control theory. John Wiley & Sons Inc, New York
Liberzon D (2012) Calculus of variations and optimal control theory. A concise introduction. Princeton University Press, Princeton
Lusin N (1912) Sur les propriétés des fonctions mesurables. C R Acad Sci Paris 154:1688–1690
Nesić D, Teel AR (2001) Sampled-data control of nonlinear systems: an overview of recent results. Perspectives in robust control (Newcastle, 2000), vol 268. Lect. Notes Control Inf. Sci. Springer, London, pp 221–239
Pontryagin LS, Boltyanskii VG, Gamkrelidze RV, Mishchenko EF (1962) The mathematical theory of optimal processes. Translated from the Russian by K. N. Trirogoff; edited by L. W. Neustadt. Interscience Publishers John Wiley & Sons, Inc. New York-London
Sharon Y, Margaliot M (2007) Third-order nilpotency, nice reachability and asymptotic stability. J Differ Equ 233(1):136–150
Sontag ED (1983) Remarks on the preservation of various controllability properties under sampling. In: Mathematical tools and models for control, systems analysis and signal processing, Vol. 3 (Toulouse/Paris, 1981/1982), Travaux Rech. Coop. Programme 567, CNRS, Paris, pp 623–637
Sontag ED (1984) An approximation theorem in nonlinear sampling. Mathematical theory of networks and systems (Beer Sheva, 1983), vol 58. Lect. Notes Control Inf. Sci. Springer, London, pp 806–812
Sontag ED (1988) Finite-dimensional open-loop control generators for nonlinear systems. Int J Control 47(2):537–556
Sontag ED, Sussmann HJ (1982) Accessibility under sampling. In: 1982 21st IEEE conference on decision and control, pp 727–732
Sontag ED (1998) Mathematical control theory, volume 6 of Texts in applied mathematics. Springer-Verlag, New York, second edition, Deterministic finite-dimensional systems
Sussmann HJ (1976) Some properties of vector field systems that are not altered by small perturbations. J Differ Equ 20(2):292–315
Sussmann HJ (1987) Reachability by means of nice controls. In: Proceedings of the 26th IEEE conference on decision and control, pp 1368–1373
Sussmann HJ, Jurdjevic V (1972) Controllability of nonlinear systems. J Differ Equ 12:95–116
Trélat E Contrôle optimal. Mathématiques Concrètes. [Concrete Mathematics]. Vuibert, Paris, 2005. Théorie & applications. [Theory and applications]
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
An example
We develop here an example inspired from [17, Section II], showing that the converse of the geometric Pontryagin maximum principle is not true in general and that, given a control \(u \in {\mathcal {U}}\cap \mathrm {L}^\infty ([0,T],\mathrm {U})\), the condition that \(x_u(T)\) belongs to the interior of the \(\mathrm {L}^\infty _\mathrm {U}\)-accessible set is not a sufficient condition for Property (\( {\mathcal {R}}^{\mathrm{{unif}}}_u \)), even if \(\mathrm {U}\) is convex.
Take \(T = n = m = 2\) and \(\mathrm {U}= {\mathbb {R}}^2\). Take \(g_1 \in \mathrm {C}([0,2],{\mathbb {R}})\) be a continuous function that is positive on the interval [0, 1) and vanishing on the interval [1, 2]. Take \(g_2 \in \mathrm {L}^\infty ([0,2],{\mathbb {R}})\) be arbitrarily fixed and \(g_3 \in \mathrm {AC}([0,2],{\mathbb {R}})\) be defined by \(g_3(t) = \int _0^t g_1(\xi )g_2(\xi ) \, \mathrm{d}\xi \) for all \(t \in [0,2]\). Note that \(g_3\) is constant on the interval [1, 2]. We denote by G the corresponding constant values. We set \(x^0 = 0_{{\mathbb {R}}^2}\) and
for all \(((x_1,x_2),(u_1,u_2),t) \in {\mathbb {R}}^2 \times {\mathbb {R}}^2 \times [0,2]\).
Claim 1
The point (G, 0) is an equilibrium of the control system on the interval [1, 2], independently of the control.
Proof
Since \(g_1(t) = 0\) and \(g_3(t) = G\) for all \(t \in [1,2]\), we have \(f((G,0),u,t) = 0_{{\mathbb {R}}^2}\) for all \((u,t) \in {\mathbb {R}}^2 \times [1,2]\). \(\square \)
Claim 2
Let \(u \in \mathrm {L}^\infty ([0,2],{\mathbb {R}}^2)\) satisfying \(u_1(t)=g_2(t)\) for a.e. \(t \in [0,1]\). Then \(u \in {\mathcal {U}}\) and \(x_u = (g_3,0)\). In particular \(x_u(2) = (G,0)\).
Proof
Since \(g_1(2-\xi ) = 0\) for all \(\xi \in [0,1]\), \(x_{u,1}(t) = \int _0^t g_1(\xi )u_1(\xi )\, \mathrm{d}\xi = \int _0^t g_1(\xi )g_2(\xi )\, \mathrm{d}\xi = g_3(t)\) for all \(t \in [0,1]\). From the second coordinate, we obtain that \(x_{u,2}(t) = 0\) for all \(t \in [0,1]\). Since \(x_u(1) = (g_3(1),0) = (G,0)\), we get from Claim 1 that \(x_u(t) = (G,0) = (g_3(t),0) \) for all \(t \in [1,2]\). \(\square \)
Claim 3
Let \(u \in {\mathcal {U}}\) such that \(x_u(T') = (G,0)\) for some \(T' \in [1,2]\). Then \(u_1(t) = g_2(t)\) for a.e. \(t \in [0,1]\).
Proof
By Claim 1, \(x_u(1) = (G,0)\). Since \(g_1(2-\xi ) = 0\) for all \(\xi \in [0,1]\), we get that \(0 = x_{u,2}(1) = \int _0^1 (x_{u,1}(\xi ) - g_3(\xi ) )^2 \, \mathrm{d}\xi \) and thus \(x_{u,1}(t) = g_3(t)\) for all \(t \in [0,1]\). Deriving this equality leads to \(g_1(t)u_1(t) = g_1(t)g_2(t)\) for a.e. \(t \in [0,1]\). Since \(g_1\) is positive on the interval [0, 1), we get that \(u_1(t) = g_2(t)\) for a.e. \(t \in [0,1]\). \(\square \)
Claim 4
The end-point mapping is surjective.
Proof
Let \(x^1 \in {\mathbb {R}}^2\). Let us prove that there exists \(u \in {\mathcal {U}}\) such that \(\mathrm {E}(u) = x_u(2) = x^1\). If \(x^1 = (G,0)\), from Claim 2, it is sufficient to take any control \(u \in \mathrm {L}^\infty ([0,2],{\mathbb {R}}^2)\) which satisfies \(u_1(t) = g_2(t)\) for a.e. \(t \in [0,1]\). In the rest of this proof, we focus on the case \(x^1 \ne (G,0)\).
Consider a function \(g_4 \in \mathrm {L}^\infty ([0,\frac{3}{2}],{\mathbb {R}})\) such that the measure of \(\{ t \in [0,1] \mid g_4(t) \ne g_2 (t) \}\) is positive and such that the \(\mathrm {L}^\infty \)-norm of \(g_4 - g_2\) on [0, 1] is small enough to guarantee that any control \(u \in \mathrm {L}^\infty ([0,2],{\mathbb {R}}^2)\) which satisfies \(u_1(t) = g_4(t)\) for a.e. \(t \in [0,\frac{3}{2}]\) is admissible, i.e., \(u \in {\mathcal {U}}\). This is possible by Claim 2, since \({\mathcal {U}}\) is an open subset of \(\mathrm {L}^\infty ([0,2],{\mathbb {R}}^2)\). Take such a control u (which is only determined on the interval \([0,\frac{3}{2}]\) at this step). By Claim 3, \(x_u( \frac{3}{2} ) \ne (G,0)\). Consider now a \(\mathrm {C}^1\) function \(\varrho : [\frac{3}{2},2] \rightarrow {\mathbb {R}}^2\) which satisfies \(\varrho (\frac{3}{2}) = x_u( \frac{3}{2} )\), \(\varrho (2) = x^1\) and \(\varrho (t) \ne (G,0)\) for all \(t \in [\frac{3}{2},2]\). We determine the control u on \([0,\frac{3}{2}]\) as
where \(\overline{\varrho }(t)=g_1(2-t) ( (\varrho _1(t) - G)^2 + \varrho _2(t)^2 )\) for a.e. \(t \in [\frac{3}{2},2]\). The control u belongs to \(\mathrm {L}^\infty ([0,2],{\mathbb {R}}^2)\) and \(x_u = \varrho \) along \([\frac{3}{2},2]\). Thus \(\mathrm {E}(u) = x_u(2) = \varrho (2) = x^1\). \(\square \)
Let us prove that the converse of the geometric Pontryagin maximum principle is not true in general. Take a control \(u \in \mathrm {L}^\infty ([0,T],{\mathbb {R}}^2)\) which satisfies \(u_1(t)=g_2(t)\) for a.e. \(t \in [0,1]\). By Claims 2 and 4, we have \(u \in {\mathcal {U}}\) and \(x_u(2)\) belongs to the interior of the \(\mathrm {L}^\infty _{{\mathbb {R}}^2}\)-accessible set. Consider the constant function \(p : [0,2] \rightarrow {\mathbb {R}}^2\) defined by \(p(t)=(0,1) \ne 0_{{\mathbb {R}}^2}\) for all \(t \in [0,2]\). One can easily check that \((x_u,u,p)\) is a nontrivial strong \({\mathbb {R}}^2\)-extremal lift of \((x_u,u)\) and thus u is strongly \({\mathbb {R}}^2\)-singular by Proposition 2.6.
We now prove that, given a control \(u \in {\mathcal {U}}\cap \mathrm {L}^\infty ([0,T],\mathrm {U})\), the condition that \(x_u(T)\) belongs to the interior of the \(\mathrm {L}^\infty _\mathrm {U}\)-accessible set is not a sufficient condition for Property (\( {\mathcal {R}}^{\mathrm{{unif}}}_u \)), even if \(\mathrm {U}\) is convex. Take \(g_2(t)=t\) for a.e. \(t \in [0,1]\) (which is not piecewise constant). Even if (G, 0) belongs to the interior of the \(\mathrm {L}^\infty _{\mathrm {U}}\)-accessible set (Claim 4), we easily infer from Claim 3 that (G, 0) is not \(\mathrm {PC}^{\mathbb {T}}_{\mathrm {U}}\)-reachable in time T from \(x^0\) for any partition \({\mathbb {T}}\) of [0, T]. Hence Property (\( {\mathcal {R}}_u \)) is not satisfied, and neither is the stronger Property (\( {\mathcal {R}}^{\mathrm{{unif}}}_u \)).
A general result on \(\mathrm {L}^s\)-approximation by piecewise constant functions
Proposition B.1
Let \(1 \le s < +\infty \). Given any \(u \in \mathrm {L}^\infty ([0,T],\mathrm {U})\) and any \(\varepsilon > 0\), there exists a threshold \(\delta > 0\) such that, for any partition \({\mathbb {T}}\) of [0, T] satisfying \(\Vert {\mathbb {T}}\Vert \le \delta \), there exists \(v \in \mathrm {PC}^{\mathbb {T}}([0,T],\mathrm {U})\) such that \(\Vert v - u \Vert _{\mathrm {L}^s} \le \varepsilon \) and \(\Vert v \Vert _{\mathrm {L}^\infty } \le \Vert u \Vert _{\mathrm {L}^\infty }\).
Proof
Let \(u \in \mathrm {L}^\infty ([0,T],\mathrm {U})\) and \(\varepsilon > 0\). By the Lusin theorem [23], there exists a compact subset \(\mathrm {K}_\varepsilon \subset [0,T]\) such that \( ( 2 \Vert u \Vert _{\mathrm {L}^\infty } )^s \mu ( [0,T] \backslash \mathrm {K}_\varepsilon ) \le \varepsilon ^s/2\), where \(\mu \) is the Lebesgue measure, and such that u is continuous on \(\mathrm {K}_\varepsilon \). By uniform continuity of u on \(\mathrm {K}_\varepsilon \), there exists \(\delta > 0\) such that \( \Vert u(\xi _2) - u(\xi _1) \Vert _{{\mathbb {R}}^m} \le \frac{\varepsilon }{(2T)^{1/s}} \) for all \(\xi _1\), \(\xi _2 \in \mathrm {K}_\varepsilon \) satisfying \(\vert \xi _2 - \xi _1 \vert \le \delta \). Now, let \({\mathbb {T}}= \{ t_i \}_{i=0,\ldots ,N}\) be a partition of [0, T] such that \(\Vert {\mathbb {T}}\Vert \le \delta \). We set
For every \(i \in I\), we consider some \(\xi _i \in \mathrm {K}_\varepsilon \cap [t_i,t_{i+1})\) such that \(u(\xi _i) \in \mathrm {U}\) and \(\Vert u(\xi _i) \Vert _{{\mathbb {R}}^m} \le \Vert u \Vert _{\mathrm {L}^\infty }\). We also consider some \(\omega \in \mathrm {U}\) such that \(\Vert \omega \Vert _{{\mathbb {R}}^m} \le \Vert u \Vert _{\mathrm {L}^\infty }\). We now define
for every \(t \in [0,T]\). In particular we have \(v \in \mathrm {PC}^{\mathbb {T}}([0,T],\mathrm {U})\) and \(\Vert v \Vert _{\mathrm {L}^\infty } \le \Vert u \Vert _{\mathrm {L}^\infty }\). Finally we get that
which concludes the proof. \(\square \)
Note that Proposition B.1 is not true with \(s=+\infty \), as shown in the following Fuller-type example [11].
Example B.1
Take \(T = 1\), \(m=1\) and \(\mathrm {U}= {\mathbb {R}}\). Consider the oscillating function \(u \in \mathrm {L}^\infty ([0,T],\mathrm {U})\) defined by \(u(t)=1\) for a.e. \(t \in (\frac{1}{k+1},\frac{1}{k}]\) for all even \(k \in {\mathbb {N}}\backslash \{0\}\) and \(u(t)=0\) for a.e. \(t \in (\frac{1}{k+1},\frac{1}{k}]\) for all odd \(k \in {\mathbb {N}}\backslash \{0\}\). We have \(\Vert v - u \Vert _{\mathrm {L}^\infty } \ge \frac{1}{2}\) for all \(v \in \mathrm {PC}^{\mathbb {T}}([0,T],\mathrm {U})\) and all partitions \({\mathbb {T}}\) of [0, T].
Corollary B.1
Let \(1 \le s < +\infty \). Given any \(u \in \mathrm {L}^s([0,T],\mathrm {U})\) and any \(\varepsilon > 0\), there exists a threshold \(\delta > 0\) such that, for any partition \({\mathbb {T}}\) of [0, T] satisfying \(\Vert {\mathbb {T}}\Vert \le \delta \), there exists \(v \in \mathrm {PC}^{\mathbb {T}}([0,T],\mathrm {U})\) such that \(\Vert v - u \Vert _{\mathrm {L}^s} \le \varepsilon \).
Proof
Let \(u \in \mathrm {L}^s([0,T],\mathrm {U})\) and \(\varepsilon > 0\). We fix some \(\omega \in \mathrm {U}\) and we define \( C_k = \{ t \in [0,T] \mid \Vert u(t) \Vert _{{\mathbb {R}}^m} \ge k \} \) and
for a.e. \(t \in [0,T]\) and for every \(k \in {\mathbb {N}}\). In particular \(u_k \in \mathrm {L}^\infty ([0,T],\mathrm {U})\) for every \(k \in {\mathbb {N}}\). It is clear that \(( u_k(t) - u(t) )_{k \in {\mathbb {N}}}\) converges to \(0_{{\mathbb {R}}^m}\) as \(k \rightarrow +\infty \) and that \(\Vert u_k(t) - u(t) \Vert _{{\mathbb {R}}^m} \le \Vert \omega \Vert _{{\mathbb {R}}^m} + \Vert u(t) \Vert _{{\mathbb {R}}^m}\) for a.e. \(t \in [0,T]\). By the Lebesgue dominated convergence theorem, we get that \(\Vert u_k - u \Vert _{\mathrm {L}^s} \rightarrow 0\) as \(k \rightarrow +\infty \). Hence, there exists \(k \in {\mathbb {N}}\) such that \(\Vert u_k - u \Vert _{\mathrm {L}^s} \le \frac{\varepsilon }{2}\). By Proposition B.1, there exists \(\delta > 0\) such that, for any partition \({\mathbb {T}}\) of [0, T] satisfying \(\Vert {\mathbb {T}}\Vert \le \delta \), there exists \(v \in \mathrm {PC}^{\mathbb {T}}([0,T],\mathrm {U})\) such that \(\Vert v - u_k \Vert _{\mathrm {L}^s} \le \frac{\varepsilon }{2}\) and thus \( \Vert v - u \Vert _{\mathrm {L}^s} \le \Vert v - u_k \Vert _{\mathrm {L}^s} + \Vert u_k - u \Vert _{\mathrm {L}^s} \le \varepsilon \). \(\square \)
1.1 Averaging operators
For any partition \({\mathbb {T}}= \{ t_i \}_{i=0,\ldots ,N}\) of [0, T], we define the averaging operator \({\mathcal {I}}^{\mathbb {T}}: \mathrm {L}^1([0,T],{\mathbb {R}}^m) \rightarrow \mathrm {PC}^{\mathbb {T}}([0,T],{\mathbb {R}}^m)\) by
for every \(t \in [t_i,t_{i+1})\), every \(i \in \{ 0,\ldots ,N-1 \}\) and every \(u \in \mathrm {L}^1([0,T],{\mathbb {R}}^m) \). The aim of this section is to establish several useful properties of the averaging operators.
Let \({\mathbb {T}}= \{ t_i \}_{i=0,\ldots ,N}\) be a partition of [0, T]. The averaging operator \({\mathcal {I}}^{\mathbb {T}}\) is linear and projects any integrable function onto a piecewise constant function respecting the partition \({\mathbb {T}}\) (by averaging its value on each sampling interval \([t_i,t_{i+1})\)).
Lemma B.1
Let \(1 \le s \le +\infty \). For any partition \({\mathbb {T}}\) of [0, T], we have \(\Vert {\mathcal {I}}^{\mathbb {T}}(u) \Vert _{\mathrm {L}^s} \le \Vert u \Vert _{\mathrm {L}^s}\) for all \(u \in \mathrm {L}^s([0,T],{\mathbb {R}}^m)\).
Proof
Let \({\mathbb {T}}= \{ t_i \}_{i=0,\ldots ,N} \) be a partition of [0, T] and let \(u \in \mathrm {L}^s([0,T],{\mathbb {R}}^m)\). When \(s=+\infty \), the inequality \(\Vert {\mathcal {I}}^{\mathbb {T}}(u) \Vert _{\mathrm {L}^\infty } \le \Vert u \Vert _{\mathrm {L}^\infty }\) is trivial. When \(1 \le s < +\infty \), we get from the Hölder inequality that
for all \(i \in \{ 0,\ldots ,N-1 \}\), and thus \(\Vert {\mathcal {I}}^{\mathbb {T}}(u) \Vert ^s_{\mathrm {L}^s} \le \Vert u \Vert ^s_{\mathrm {L}^s}\). \(\square \)
The next lemma is instrumental in order to approximate with a \(\mathrm {L}^s\)-norm (with any \(1 \le s < +\infty \)) any control \(u \in \mathrm {L}^\infty ([0,T],{\mathbb {R}}^m)\) with piecewise constant controls.
Lemma B.2
Let \(1 \le s < +\infty \). Given any \(u \in \mathrm {L}^s([0,T],{\mathbb {R}}^m)\), we have \(\Vert {\mathcal {I}}^{\mathbb {T}}(u) - u \Vert _{\mathrm {L}^s} \rightarrow 0\) as \(\Vert {\mathbb {T}}\Vert \rightarrow 0\).
Proof
Let \(u \in \mathrm {L}^s([0,T],{\mathbb {R}}^m)\) and let us prove that \(\Vert {\mathcal {I}}^{\mathbb {T}}(u) - u \Vert _{\mathrm {L}^s} \rightarrow 0\) as \(\Vert {\mathbb {T}}\Vert \rightarrow 0\). Our proof is based on a density argument. Therefore, in a first place, we deal with the case where u is essentially bounded, that is, \(u \in \mathrm {L}^\infty ([0,T],{\mathbb {R}}^m)\). From Lemma B.1, the domination \(\Vert {\mathcal {I}}^{\mathbb {T}}(u)(t) \Vert _{{\mathbb {R}}^m} \le \Vert u \Vert _{\mathrm {L}^\infty }\) is true for a.e. \(t \in [0,T]\) and thus, thanks to the Lebesgue dominated convergence theorem, we only need to prove that \({\mathcal {I}}^{\mathbb {T}}(u)(\tau ) \rightarrow u(\tau )\) as \(\Vert {\mathbb {T}}\Vert \rightarrow 0\) for a.e. \(\tau \in [0,T]\). For this purpose we set \(r(t) = \int _0^t u(\xi ) \, \mathrm{d}\xi \) for every \(t \in [0,T]\) and let \(\tau \in [0,T)\) being a Lebesgue point such that r is derivable at \(\tau \) with \({\dot{r}}(\tau ) = u(\tau )\). Given any \(\varepsilon > 0\), there exists \(\delta > 0\) such that
for every \(t \in [\tau -\delta ,\tau +\delta ] \cap [0,T] \backslash \{ \tau \}\). Take \({\mathbb {T}}\) a partition of [0, T] such that \(\Vert {\mathbb {T}}\Vert \le \delta \). There exists \(i \in \{ 0, \ldots , N-1 \}\) such that \(\tau \in [t_i,t_{i+1})\). Then
which concludes the proof in the case where u is essentially bounded. Now let us deal with the general case. Take \(\varepsilon > 0\) be arbitrary. By a density argument, there exists \(u' \in \mathrm {L}^\infty ([0,T],{\mathbb {R}}^m)\) such that \(\Vert u - u' \Vert _{\mathrm {L}^s} \le \frac{\varepsilon }{2}\). From the first step, there exists \(\delta > 0\) such that \(\Vert {\mathcal {I}}^{\mathbb {T}}(u') - u' \Vert _{\mathrm {L}^s} \le \frac{\varepsilon }{2}\) for any partition \({\mathbb {T}}\) of [0, T] such that \(\Vert {\mathbb {T}}\Vert \le \delta \). By linearity of \({\mathcal {I}}^{\mathbb {T}}\) and by Lemma B.1, we obtain that \(\Vert {\mathcal {I}}^{\mathbb {T}}(u) - u \Vert _{\mathrm {L}^s} \le \Vert {\mathcal {I}}^{\mathbb {T}}(u) - {\mathcal {I}}^T(u') \Vert _{\mathrm {L}^s} + \Vert {\mathcal {I}}^{\mathbb {T}}(u') - u' \Vert _{\mathrm {L}^s} \le \varepsilon \) for any partition \({\mathbb {T}}\) of [0, T] such that \(\Vert {\mathbb {T}}\Vert \le \delta \). The proof is complete. \(\square \)
Our objective now is to prove that, when \(\mathrm {U}\) is convex, the averaging operators project any integrable function with values in \(\mathrm {U}\) onto a piecewise constant function with values in \(\mathrm {U}\).
Lemma B.3
Assume that \(\mathrm {U}\) is convex. If \(u \in \mathrm {L}^1([0,1],\mathrm {U})\), then \(\int _0^1 u(\xi ) \mathrm{d}\xi \in \mathrm {U}\).
Proof
Let \(u \in \mathrm {L}^1([0,1],\mathrm {U})\) and let us prove that \(\widetilde{u} \in \mathrm {U}\) where \(\widetilde{u}\) is defined by \(\widetilde{u} = \int _0^1 u(\xi ) \mathrm{d}\xi \). We first give a simpler argument when \(\mathrm {U}\) is furthermore assumed to be closed. In that context, by the Hilbert projection theorem, we have
for a.e. \(\xi \in [0,1]\), where \(\mathrm {proj}_\mathrm {U}( \widetilde{u} ) \in \mathrm {U}\) is the projection of \(\widetilde{u} \) onto \(\mathrm {U}\). Integrating the above inequality over [0, 1] yields \(\Vert \widetilde{u} - \mathrm {proj}_\mathrm {U}( \widetilde{u} ) \Vert _{{\mathbb {R}}^m}^2 \le 0\) and thus \(\widetilde{u} = \mathrm {proj}_\mathrm {U}( \widetilde{u} ) \in \mathrm {U}\).
Now we remove the closedness assumption made on \(\mathrm {U}\). Let us prove that \(\widetilde{u} \in \mathrm {U}\) by strong induction on the dimension \(d \in {\mathbb {N}}\) of the nonempty convex set \(\mathrm {U}\). If \(d=0\), the set \(\mathrm {U}\) is reduced to a singleton and the result is trivial. Now consider that \(d \ge 1\) and assume that the result is true at all steps from 0 to \(d-1\). By contradiction assume that \(\widetilde{u} \notin \mathrm {U}\). By separation, there exists \(\psi \in {\mathbb {R}}^m \backslash \{ 0_{{\mathbb {R}}^m} \}\) such that \(\langle \psi , \omega - \widetilde{u} \rangle _{{\mathbb {R}}^m} \le 0 \) for all \(\omega \in \mathrm {U}\). We infer that the null integral \( \int _0^1 \langle \psi , u(\xi ) - \widetilde{u} \rangle _{{\mathbb {R}}^m} \mathrm{d}\xi \) has a nonpositive integrand. Thus this integrand is zero almost everywhere on [0, 1]. Therefore, u is with values in the convex set \(\mathrm {U}\cap ( \widetilde{u} + \psi ^\bot )\), where \(\psi ^\bot \) stands for the standard hyperplane defined by orthogonality with the nonzero vector \(\psi \). Since \(\mathrm {U}\cap ( \widetilde{u} + \psi ^\bot )\) is a nonempty convex set of dimension strictly inferior than d, thanks to our induction hypothesis we get that \(\widetilde{u} \in \mathrm {U}\cap ( \widetilde{u} + \psi ^\bot )\), which raises a contradiction. \(\square \)
From Lemma B.3 and applying a simple affine change of variable in (7), we obtain the next proposition.
Proposition B.2
Assume that \(\mathrm {U}\) is convex. If \(u \in \mathrm {L}^1([0,T],\mathrm {U})\), then \({\mathcal {I}}^{\mathbb {T}}(u) \in \mathrm {PC}^{\mathbb {T}}([0,T],\mathrm {U})\) for any partition \({\mathbb {T}}\) of [0, T].
1.2 Truncated end-point mapping and \(\mathrm {L}^s\)-differential
For every \(M > 0\), we fix a mapping \(\Lambda ^M : {\mathbb {R}}^n \times {\mathbb {R}}^m \rightarrow {\mathbb {R}}\) of class \(\mathrm {C}^1\) satisfying
Let \(M > 0\). When replacing the dynamics f in the control system (CS) by the truncated dynamics \(f^M\), defined by \( f^M (x,u,t) = \Lambda ^M(x,u) f(x,u,t) \) for all \((x,u,t) \in {\mathbb {R}}^n \times {\mathbb {R}}^m \times [0,T]\) we obtain a new control system that we denote by (\(\mathrm {CS}^M\)). The main difference is that, for any control \(u \in \mathrm {L}^1([0,T],{\mathbb {R}}^m)\) (even unbounded), there exists a trajectory \(x \in \mathrm {AC}([0,T], {\mathbb {R}}^n)\), starting at \(x(0)=x^0\), such that \({\dot{x}}(t) = f^M ( x(t),u(t),t )\) for a.e. \(t \in [0,T]\). In that case the trajectory x is unique and will be denoted by \(x^M_u\). We now introduce, for any \(1 \le s \le +\infty \), the truncated end-point mapping \(\mathrm {E}^M : \mathrm {L}^s([0,T],{\mathbb {R}}^m)\rightarrow {\mathbb {R}}^n\) defined by \(\mathrm {E}^M(u) = x^M_u(T)\) for all \(u \in \mathrm {L}^s([0,T],{\mathbb {R}}^m)\). Note that the next proposition, derived from standard techniques in ordinary differential equations theory, is true for any \(1 < s \le +\infty \). The case \(s=1\) is discussed in Remark B.2.
Proposition B.3
Let \(1 < s \le +\infty \) and \(M > 0\). The truncated end-point mapping \(\mathrm {E}^M:\mathrm {L}^s([0,T],{\mathbb {R}}^m)\rightarrow {\mathbb {R}}^n\) is of class \(\mathrm {C}^1\) and its Fréchet differential is given by
for all u, \(v \in \mathrm {L}^s([0,T],{\mathbb {R}}^m)\), where \(w^{u,M}_v \in \mathrm {AC}([0,T],{\mathbb {R}}^n)\) is the unique solution to
Remark B.1
Let \(1 < s \le +\infty \). For a given control \(u \in {\mathcal {U}}\), note that \(\mathrm {D}\mathrm {E}(u)\) given in (1) admits a natural extension (still denoted by) \(\mathrm {D}\mathrm {E}(u):\mathrm {L}^s([0,T],{\mathbb {R}}^m)\rightarrow {\mathbb {R}}^n\). The nontruncated setting is related to the truncated one as follows:
-
(i)
Let \(u \in {\mathcal {U}}\) and \(M > 0\) be such that \(\Vert x_u \Vert _{\mathrm {C}} \le M\) and \(\Vert u \Vert _{\mathrm {L}^\infty } \le M\). Then \(x^M_u = x_u\) and \(\mathrm {D}\mathrm {E}^M (u) = \mathrm {D}\mathrm {E}(u)\) when considering the above extension of \(\mathrm {D}\mathrm {E}(u)\).
-
(ii)
Let \(u \in \mathrm {L}^\infty ([0,T],{\mathbb {R}}^m)\). If there exists \(M > 0\) such that \(\Vert x^M_u \Vert _{\mathrm {C}} \le M\) and \(\Vert u \Vert _{\mathrm {L}^\infty } \le M\), then \(u \in {\mathcal {U}}\) and \(x_u = x^M_u\).
Remark B.2
Let \(M > 0\). In the case \(s=1\), it can be proved that the truncated end-point mapping \(\mathrm {E}^M:\mathrm {L}^1([0,T],{\mathbb {R}}^m)\rightarrow {\mathbb {R}}^n\) is Gateaux-differentiable and its Gateaux differential is given by (8). However, it is not Fréchet-differentiable (and thus not of class \(\mathrm {C}^1\)) in general, as shown in the next example.
Example B.2
Take \(T=n=m=1\), \(\mathrm {U}= {\mathbb {R}}\) and \(f(x,u,t) = u^2\) for all \((x,u,t) \in {\mathbb {R}}\times {\mathbb {R}}\times [0,T]\). Consider the starting point \(x^0 = 0\) and the constant control \(u \equiv 0\). In that context, with \(M=s=1\), it is clear that \(x^M_u \equiv 0\) and that the Gateaux differential \(\mathrm {D}^{\mathrm {G}} \mathrm {E}^M(u) :\mathrm {L}^1([0,T],{\mathbb {R}}^m)\rightarrow {\mathbb {R}}^n\) of the truncated end-point mapping \(\mathrm {E}^M:\mathrm {L}^1([0,T],{\mathbb {R}}^m)\rightarrow {\mathbb {R}}^n\) at u, given by the expression (8), is null. Now, taking the needle-like control variation \(u^\alpha _{(0,1)}\), as defined in (2), associated with the pair \((0,1) \in {\mathcal {L}}( f_u) \times \mathrm {U}\), we obtain that
Therefore, \(\mathrm {E}^M:\mathrm {L}^1([0,T],{\mathbb {R}}^m)\rightarrow {\mathbb {R}}^n\) is not Fréchet-differentiable at u. A similar example can be found in [29, Remark 3.1].
Rights and permissions
About this article
Cite this article
Bourdin, L., Trélat, E. Robustness under control sampling of reachability in fixed time for nonlinear control systems. Math. Control Signals Syst. 33, 515–551 (2021). https://doi.org/10.1007/s00498-021-00290-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00498-021-00290-2
Keywords
- Nonlinear control systems
- Reachability
- Sampled-data controls
- Piecewise constant controls
- Regular controls