arXiv++ Combinatorics

A convex polygon $A$ is related to a convex $m$-gon $K= \bigcap_{i=1}^m k_i^+$, where $k_1^+,..., k_m^+$ are the $m$ halfplanes whose intersection is equal to $K$, if $A$ is the intersection of halfplanes $a_1^+,...,a_l$, each of which is a translate of one of the $k_i^+$-s. The planar family ${\cal A}$ is related to $K$ if each $A \in {\cal A}$ is related to $K$. We prove that any family of pairwise intersecting convex sets related to a given $n$-gon has a finite piercing number which depends on $n$. In the general case we show $O(3^{n^3})$, while for a certain class of families, we decrease the bound to $4(n-2)$, and for $n=3,4$ the bound is 3 and 6 respectively.

The complexity of two related geometrical (indeed, combinatorial) problems is considered, measured by the number of queries needed to determine the solution. It is proved that one cannot check in a linear in N number of queries whether N intervals cover a whole interval, or whether for N pairs of intervals on two lines there is a pair of points intersecting each of these pairs of intervals ("piercing all pairs of intervals"). The proofs are related to examples which show that there is no "Helly property" here - the whole set of N may cover the whole interval (resp. may have no pair of points piercing all pairs of intervals) while any proper subset does not. Also, for the piercing problem we outline an algorithm, taking O(N log N) steps, to check whether there is a pair of points piercing all pairs of intervals and if there is, to find it.