Immobilization of smooth convex figures (Q1345110)

Immobilization problems were introduced by \textit{W. Kuperberg} [DIMACS, Workshop on Polytopes, Rutgers Univ. (1990)]. A collection of points \(H\) on the boundary of a plane figure \(\psi\) is said to immobilize \(\psi\) if any small rigid motion of \(\psi\) causes one point in \(H\) to penetrate the interior of \(\psi\). It has been shown in a paper of \textit{J. Gzyzowicz}, \textit{I. Stojmenovic} and the third author [`Immobilizing a shape', RR 90/11-18, Departement d'informatique, Université du Quebec à Hull (1990); see also Lect. Notes Comput. Sci. 519, 214-227 (1991; Zbl 0785.52004)] that any plane figure which is not a circular disk can be immobilized with at most four points, exactly four points being needed for a parallelogram. In spite of this Kuperberg conjectured that three points should be enough for some large classes of figures. Here the authors show that Kuperberg's conjecture is true for \(C^ 2\)- boundaries of convex domains (being different from disks). This is initiated by providing a local study of so-called slidings along curves. These studies lead to the introduction of some immobilization coefficient, involving geometric data from the tangents, the normals and the curvatures of three points in the boundary of the figure, which determines up to second order whether the three points immobilize it or not.