Quadratic obstructions to Small-Time Local Controllability for multi-input systems

Théo Gherdaoui · 2024

Read Quadratic obstructions to Small-Time Local Controllability for multi-input systems by Théo Gherdaoui (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-17384.txt.

Introduction

Théo Gherdaoui presents a necessary condition for the small-time local controllability of multi-input control-affine systems. Controllability asks whether a system's controls can steer it anywhere nearby in arbitrarily short time. Obstructions show where that is impossible, no matter how clever the control.

The condition is formulated on vectors obtained by evaluating Lie brackets of the system's vector fields at zero. It involves both the direction and the amplitude of those vectors. The proof adapts a general method developed for the single-input case, built on a Magnus-type representation formula. In that formula, the state is approximated by a combination of bracket evaluations whose coefficients are functionals of time and the controls. Interpolation inequalities then yield the obstructions. The result sharpens the theory engineers use to judge whether a machine can be maneuvered as intended.

Catalog record

CREATORThéo Gherdaoui
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE123 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-17384

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Quadratic obstructions to Small-Time Local Controllability for multi-input systems Théo Gherdaoui Abstract We present a necessary condition for the small-time local controllability of multi-input control-affine systems on ℝ d \mathbb{R}^{d} . This condition is formulated on the vectors of ℝ d \mathbb{R}^{d} resulting from the evaluation at zero of the Lie brackets of the vector fields: it involves both their direction and their amplitude. The proof is an adaptation to the multi-input case of a general method introduced by Beauchard and Marbach in the single-input case – see [ 4 ] . It relies on a Magnus-type representation formula: the state is approximated by a linear combination of the evaluation at zero of the Lie brackets of the vector fields, whose coefficients are functionals of the time and the controls. Finally, obstructions to small-time local controllability result from interpolation inequalities. Keywords: Control theory, ODEs, Obstruction for controllability, Lie Brackets, Small-time local exact controllability, control-affine systems. 1 Introduction 1.1 Definitions We consider the control-affine system \left\{\begin{array}[]{rcl}x^{\prime}(t)&=&f_{0}(x(t))+u(t)f_{1}(x(t))+v(t)f_{2}(x(t))\\ x(0)&=&0\end{array}\right., (1.1) where the state is x(t)\in\mathbb{R}^{d} , the controls are scalar functions u(t),v(t)\in\mathbb{R} and f_{0},f_{1} and f_{2} are real-analytic vector fields on a neighborhood of 0 in \mathbb{R}^{d} with f_{0}(0)=0 . The last hypothesis ensures that 0 is an equilibrium of the free system ( i.e. with (u,v)\equiv 0) . For each t>0 , u,v\in L^{1}((0,t),\mathbb{R}) , there exists a unique maximal mild solution to ( 1.1 ), which we will denote by x(\cdot;(u,v)) . As we are interested in small time and small controls, the solution is well-defined up to time t . In this article, we study the small-time local controllability of system ( 1.1 ) in the sense of Definition 1.1 below, which requires the following notion. For t>0 , m\in\mathbb{N} and p\in[1,+\infty] , we consider the usual Sobolev space W^{m,p}((0,t),\mathbb{R}) equipped with the usual norm \left\|u\right\|_{W^{m,p}}:=\left\|u\right\|_{L^{p}}+\cdots+\|u^{(m)}\|_{L^{p}}. The following concept was introduced by Beauchard and Marbach in [ 5 ] for scalar-input systems. Definition 1.1 ( W^{m,p}\times W^{m^{\prime},p^{\prime}} -STLC) . Let m,m^{\prime}\in\mathbb{N} , p,p^{\prime}\in[1,+\infty] . We say that system ( 1.1 ) is W^{m,p}\times W^{m^{\prime},p^{\prime}} -STLC when, for every t,\rho>0 , there exists \delta>0 , such that, for every x^{*}\in B(0,\delta) , there exist (u,v)\in W^{m,p}((0,t),\mathbb{R})\times W^{m^{\prime},p^{\prime}}((0,t),\mathbb{R}) with \left\|(u,v)\right\|_{W^{m,p}\times W^{m^{\prime},p^{\prime}}}\leq\rho and x(t;(u,v))=x^{*} . We say that system ( 1.1 ) is W^{m,p} -STLC when ( 1.1 ) is W^{m,p}\times W^{m,p} -STLC. Remark 1.2 . The historical notion of STLC corresponds to m=0 , p=\infty – see [ 8 , 26 , 24 ] . We are looking for necessary conditions for STLC formulated in terms of Lie brackets; the definition is recalled below. Indeed, Krener proved in [ 19 , Theorem 1 ] that all the information concerning the STLC is contained in the evaluation at 0 of the Lie brackets of the vector fields f_{0} , f_{1} and f_{2} . Definition 1.3 (Lie bracket of vector fields) . Let f,g:\Omega\to\mathbb{R}^{d} be two smooth (at least \mathcal{C}^{1} ) vector fields on an open subset \Omega of \mathbb{R}^{d} . We define [f,g]:x\in\Omega\mapsto\mathrm{D}g(x)\cdot f(x)-\mathrm{D}f(x)\cdot g(x). (1.2) 1.2 Illustrative examples We propose to study the following example of a quadratic system in an elementary way, as it is representative of the type considered in this article. Example 1.4 . Let us consider \left\{\begin{array}[]{rcl}x_{1}^{\prime}&=&u\\ x_{2}^{\prime}&=&v\\ x_{3}^{\prime}&=&x_{1}^{2}+x_{2}^{2}+\alpha x_{1}x_{2}\end{array}\right., (1.3) with \alpha\in\mathbb{R} . This is a control-affine system of the form ( 1.1 ) with f_{0}(x)=\begin{pmatrix}0\\ 0\\ x_{1}^{2}+x_{2}^{2}+\alpha x_{1}x_{2}\end{pmatrix},\quad f_{1}(x)=e_{1},\quad f_{2}(x)=e_{2}. For all x,y\in\mathbb{R} , x^{2}+y^{2}+\alpha xy=\left(x+\frac{\alpha}{2}y\right)^{2}+\frac{4-\alpha^{2}}{4}y^{2}. Consequently, if |\alpha|\leq 2 , then x_{3}^{\prime}\geq 0 and ( 1.3 ) is not L^{\infty} -STLC. Note that, when |\alpha|>2 , for every m\in\mathbb{N} , the system is W^{m,\infty}-STLC . Indeed, we can choose u:s\in(0,t)\mapsto\varphi^{\prime}(\frac{s}{t}) and v:s\in(0,t)\mapsto\psi^{\prime}(\frac{s}{t}) with \varphi,\psi\in\mathcal{C}^{\infty}_{c}((0,1),\mathbb{R}) . If \psi\equiv 0 , the state moves along +e_{3} . If \varphi\equiv-\frac{\alpha}{2}\psi , the state moves along -e_{3} , as x_{3}(t)=-\frac{\alpha^{2}-4}{4}t^{3}\int_{0}^{1}\psi^{2}. The purpose of this article is to determine assumptions on the evaluation at 0 of the Lie brackets of f_{0} , f_{1} and f_{2} under which such a system is, in a sense, embedded within the system ( 1.1 ). This analysis can be generalized to an affine system with integrators, as in the following example. Example 1.5 . Let k,k^{\prime}\in\mathbb{N}^{*} and \alpha\in\mathbb{R} be such that |\alpha|\leq 2 . As in the previous case, the following system is not L^{\infty} -STLC \left\{\begin{array}[]{rcl}x_{1}^{\prime}&=&u\\ x_{2}^{\prime}&=&x_{1}\\ &\vdots&\\ x_{k}^{\prime}&=&x_{k-1}\\ y_{1}^{\prime}&=&v\\ y_{2}^{\prime}&=&y_{1}\\ &\vdots&\\ y_{k^{\prime}}^{\prime}&=&y_{k^{\prime}-1}\\ z_{1}^{\prime}&=&x_{k}^{2}+y_{k^{\prime}}^{2}+\alpha x_{k}y_{k^{\prime}}\end{array}\right.. In the first two examples, there was an explicitly signed direction due to the presence of a positive definite quadratic form. Here, we present a more general affine system that includes terms which are harmless with respect to STLC. This system captures the essence of the phenomenon we are about to study. Example 1.6 . We consider the following system \left\{\begin{array}[]{rcl}x_{1}^{\prime}&=&u\\ x_{2}^{\prime}&=&x_{1}\\ x_{3}^{\prime}&=&v\\ x_{4}^{\prime}&=&\left(x_{1}^{2}+2x_{3}^{2}+\frac{1}{2}x_{1}x_{3}\right)-1028x_{2}^{2}-643vx_{1}^{2}-2vx_{3}\end{array}\right.. (1.4) First of all, for all x,y\in\mathbb{R} , x^{2}+2y^{2}+\frac{1}{2}xy\geq\frac{3}{4}\left(x^{2}+y^{2}\right) . Let T>0 . If u\in L^{1}(0,T) , we note u_{1}:t\in[0,T]\mapsto\int_{0}^{t}u(s)\mathrm{d}s . For all u,v\in L^{1}(0,T) , we then obtain \int_{0}^{T}\left(x_{1}^{2}+2x_{3}^{2}+\frac{1}{2}x_{1}x_{3}\right)(t)\mathrm{d}t\geq\frac{3}{4}\int_{0}^{T}\left(x_{1}^{2}+x_{3}^{2}\right)(t)\mathrm{d}t=\frac{3}{4}\left\|(u_{1},v_{1})\right\|_{L^{2}}^{2}. (1.5) Consequently, as in Examples 1.4 and 1.5 , the first three terms of the last line define a positive definite quadratic form. Let us show that the last three terms do not prevent us from drawing the conclusion. We have \left|\int_{0}^{T}x_{2}(t)^{2}\mathrm{d}t\right|=\left|\int_{0}^{T}\left(\int_{0}^{t}u_{1}(s)\mathrm{d}s\right)^{2}\mathrm{d}t\right|\leq\int_{0}^{T}\left\|u_{1}\right\|_{L^{2}(0,t)}^{2}t\mathrm{d}t\leq\frac{T^{2}}{2}\left\|(u_{1},v_{1})\right\|_{L^{2}}^{2}, (1.6) \left|\int_{0}^{T}v(t)x_{1}(t)^{2}\mathrm{d}t\right|=\left|\int_{0}^{T}v(t)u_{1}(t)^{2}\mathrm{d}t\right|\leq\left\|v\right\|_{L^{\infty}}\left\|u_{1}\right\|_{L^{2}}^{2}\leq\left\|(u,v)\right\|_{L^{\infty}}\left\|(u_{1},v_{1})\right\|_{L^{2}}^{2}. (1.7) Then, the fourth and fifth terms are negligible compared to \left\|(u_{1},v_{1})\right\|_{L^{2}}^{2} as \left(T,\left\|(u,v)\right\|_{L^{\infty}}\right)\to 0 . Using ( 1.5 ), ( 1.6 ) and ( 1.7 ), an explicit integration of the solution from 0 leads to \displaystyle x_{4}(T) \displaystyle\geq\left(\frac{3}{4}-514T^{2}-643\left\|(u,v)\right\|_{L^{\infty}}\right)\left\|(u_{1},v_{1})\right\|_{L^{2}}^{2}-x_{3}(T)^{2} \displaystyle\geq C\left\|(u_{1},v_{1})\right\|_{L^{2}}^{2}-x_{3}(T)^{2}, for every C\in\left(0,\frac{3}{4}\right) , for small times T and small controls in L^{\infty} . As a result, any target in the set \{x\in\mathbb{R}^{4}\;;\;x_{4}+x_{3}^{2}<0\} is not reachable, so the system ( 1.4 ) is not L^{\infty} -STLC. 1.3 Drift for proving obstructions Our strategy to deny W^{m,p}\times W^{m^{\prime},p^{\prime}} -STLC consists in proving that system ( 1.1 ) has a drift. Definition 1.7 (Drift) . Let e\in\mathbb{R}^{d} , N\subset\mathbb{R}^{d} be a vector subspace, m,m^{\prime}\in\mathbb{N} , p,p^{\prime}\in[1,+\infty] , \alpha\in\mathbb{R} and \Delta:L^{1}_{\mathrm{loc}}(\mathbb{R}^{+},\mathbb{R})^{2}\to\mathbb{R}^{+} . We say that system ( 1.1 ) has a drift along e parallel to N with strength \Delta as \left(t,t^{\alpha}\left\|(u,v)\right\|_{W^{m,p}\times W^{m^{\prime},p^{\prime}}}\right)\to 0 when there exist C>0 , \beta>1 and \rho>0 such that, for every t\in(0,\rho) and (u,v)\in W^{m,p}((0,t),\mathbb{R})\times W^{m^{\prime},p^{\prime}}((0,t),\mathbb{R}) with t^{\alpha}\left\|(u,v)\right\|_{W^{m,p}\times W^{m^{\prime},p^{\prime}}}\leq\rho , \mathbb{P}x(t;(u,v))\geq C\Delta(u,v)-C\left|x(t;(u,v))\right|^{\beta}, (1.8) where \mathbb{P} is a linear form, satisfying \mathbb{P}(e)>0 and \mathchoice{{\mathbb{P}\,\smash{\vrule height=5.46666pt,depth=1.27501pt}}_{\,N}}{{\mathbb{P}\,\smash{\vrule height=5.46666pt,depth=1.27501pt}}_{\,N}}{{\mathbb{P}\,\smash{\vrule height=3.82668pt,depth=0.85472pt}}_{\,N}}{{\mathbb{P}\,\smash{\vrule height=2.71904pt,depth=1.44028pt}}_{\,N}}\equiv 0 . Lemma 1.8 . With the notations of Definition 1.7 , if the system ( 1.1 ) has a drift along e parallel to N with strength \Delta as \left(t,t^{\alpha}\left\|(u,v)\right\|_{W^{m,p}\times W^{m^{\prime},p^{\prime}}}\right)\to 0 , then, the system ( 1.1 ) is not W^{m,p}\times W^{m^{\prime},p^{\prime}} -STLC. Proof. The solution x(t;(u,v)) cannot reach targets of the form x^{*}=-ae with a>0 small because this would entail -a\mathbb{P}(e)\geq C\Delta(u,v)-C\left|-ae\right|^{\beta}\geq-C\left|e\right|^{\beta}a^{\beta}, which is impossible for a small since \beta>1 and \mathbb{P}(e)>0 . ∎ 1.4 Algebraic background We use the definitions and notations of Beauchard and Marbach in [ 4 ] . Let X:=\{X_{0},X_{1},X_{2}\} be a set of 3 non-commutative indeterminates. Definition 1.9 (Free algebra) . We note \mathcal{A}(X) the free algebra generated by X over the field \mathbb{R} , i.e. the unital associative algebra of polynomials of the indeterminates X_{0} , X_{1} and X_{2} . Definition 1.10 (Free Lie algebra) . For a,b\in\mathcal{A}(X) , we define the Lie bracket of a and b as [a,b]:=ab-ba

Read the complete text