Zur \(L^ 2\) Bestapproximation eines konvexen Körpers durch einen bewegten konvexen Körper. (On the \(L^ 2\)-best approximation of a convex body by a moved convex body) (Q748887)

Let \(E^ n\) be the n-dimensional Euclidean space, \(S^{n-1}\) the unit sphere and do the element of surface area of \(S^{n-1}\). The \(\delta^ 2\)-distance of two convex bodies K and L of \(E^ n\) is defined by the \(L^ 2\)-norm \[ \delta^ 2(K,L):=(\int_{S^{n-1}}(h_ L(x)-h_ K(x))^ 2do(x))^{1/2}=\| h_ L-h_ K\|_ 2 \] of \(h_ L-h_ K\) where \(h_ K,h_ L: S^{n-1}\to {\mathbb{R}}\) denote the support functions of K and L. We consider the problem of minimizing the distance \(\delta^ 2(\phi K,L)\) in the group of Euclidean motions \(\phi\) : \(E^ n\to E^ n\). A motion \(\phi\) that minimizes \(\delta^ 2(\phi K,L)\) is called optimal for (K,L) (in the sense of the metric \(\delta\)). The mean result in the general case (n arbitrary) shows that the Steiner points of \(\phi\) K and L coincide if \(\phi\) is optimal for (K,L). In the case \(n=2\) the optimal location of two convex discs is discussed. Finally, using finite Fourier transforms of the support functions, the deciding theorem with respect to the practical computation of a plane optimal motion is proved.