An object and a group of observers in a normed space (Q643835)

From MaRDI portal
scientific article
Language Label Description Also known as
English
An object and a group of observers in a normed space
scientific article

    Statements

    An object and a group of observers in a normed space (English)
    0 references
    0 references
    2 November 2011
    0 references
    Let \(X\) be a Banach space with differentiable strictly convex norm \(\|\cdot\|,\, G\subset X\) be the closure of an open subset of \(X,\, t\in X\setminus G\) a moving object, and \(f\in X\) an observer. The object \(t\) is visible for \(f\) if \((t,f)\cap G=\emptyset,\) where \((t,f)=\{(1-\theta)t+\theta f : 0<\theta<1\}\) denotes the open segment with endpoints \(t\) and \(f\). One supposes that the location of \(f\) is known with an error \(h>0\), i.e., it is known that it belongs to the open ball \(V'_h(f)=\{x\in X : \|f-x\|'<h\},\) where \(\|\cdot\|'\) is a differentiable norm on \(X,\) possibly different from \(\|\cdot\|.\) The observer prefers a position such that any point of the ball \(V_r(t)=\{x\in X : \|t-x\|<r\}\) is visible for \(f\) for an as large as possible number \(r>0\). Putting \(K_r(t,f,h)=\text{conv}\left(V_r(t)\cup V'_h(f)\right)\setminus V'_h(f),\) this means that \(K_r(t,f,h)\cap G=\emptyset\) for the largest \(r\), and so the function \(r(t,f)=r_h(t,f,G)=\min\{r>0 : K_r(t,f,h)\cap G\neq\emptyset\}\) characterizes the visibility of \(t\) for \(f\). Let \(R(t,f,g)=\min\{r>0 : g\in K_r(t,f,h)\},\) so that \(r(t,f)=r(t,f,G)=\min\{R(t,f,g) : g\in G\}.\) Continuing the investigations on this topic, the author studies in this paper the differentiability properties of the function \(R\). He shows that the function \(R\) is differentiable in every direction \(\tilde t,\,\|\tilde t\|=1,\) and the directional derivative can be expressed in terms of the support functional \(l\) corresponding to the norm \(\|\cdot\|\), defined by the conditions: (i) \(\;l_x(x)=1=\|l_x\|\) for every \(x\in X\) with \(\, \|x\|=1,\) and (ii)\; \(l_{\lambda x}=\lambda\, l_x\) for every \(\lambda \geq 0\) and \(\|x\|=1.\) It is also shown that the derivative \(\partial R(t,f,g)/\partial t\) is strong-to-weak continuous in all the variables. As a consequence one obtains that if \(X\) is finite dimensional with strictly convex smooth norm, then the function \(r(t,f,g)\) is differentiable in any direction \(\tilde t\) and \(\partial R(t,f,g)/\partial \tilde t=\min\{ \partial R(t,f,g)/\partial \tilde t : g\in G(t)\},\) where \(G(t)=\{g\in G : R(t,f,g) =r(t,f,g)\},\) a result obtained by the author in a previous paper [Dokl. Math. 82, No. 3, 986--987 (2010); translation from Dokl. Akad. Nauk., Ross. Akad. Nauk. 435, No. 5, 595--597 (2010; Zbl 1226.46066)]. At the end of the paper one considers the case of several observers, meaning a compact set \(\mathcal F\) of observers in a finite dimensional strictly convex and smooth normed space. In this case, one proves the directional differentiability of the function \(\mathcal R(t)=\max\{r(t,f,g) : f\in \mathcal F\}.\)
    0 references
    moving object
    0 references
    visibility
    0 references
    Gâteaux derivative
    0 references
    optimization
    0 references
    smooth normed space
    0 references
    strictly convex normed space
    0 references

    Identifiers