An improvement of previously known upper bound of Multiple Strip Packing problem and probabilistic analysis of algorithm in case of large number of strips given
D. O. Lazareva,
N. N. Kuzyurinba a Institute for System Programming of the Russian Academy of Sciences
b Moscow Institute of Physics and Technology
Abstract:
In this article, an analog of previously proposed algorithm Limited Hash Packing for Multiple Strip Packing Problem is studied using probabilistic analysis. Limited Hash Packing is an on-line algorithm, which works in closed-end mode, knowing the number
$N$ of rectangles it has to pack before knowing the heights and width of the first rectangle. The algorithm proposes that width and heights of all rectangles have a uniform on
$[0,1]$ distribution and works in two stages. Firstly, it divides the
$k$ strips into
$d=\Theta(\max\{k,\sqrt N\})$ rectangular areas width of which equal
$\frac id,\forall i=\overline{1,\dots,d}$ such that the sum space of all this areas equals the expected space of all rectangles,
$\frac N4$. Secondly, it packs a rectangle area of minimal width, in which it fits, or, if rectangle doesn’t fit in any area, above all areas. It was shown, that for any number of strips
$k$ and any number of rectangles
$N$, the expected value of space not filled with rectangles of all strips from their lowest point to the highest point of the highest rectangle,
$E(S_{sp})\le 6\sqrt{N\ln N}+3k$. It was also shown, that
$E(S_{sp})\ge \frac k8 - \frac{\sqrt N}4$. This result proves that the previous bound is asymptotically tight in case when packing
$N$ rectangles into
$k\ge\sqrt{N\ln N}$ strips.
Keywords:
on-line algorithm, closed-end, probabilistic analysis, closed-end mode, Multiple Strip Packing, an algorithm for packing into limited areas Limited Hash Packing.
DOI:
10.15514/ISPRAS-2019-31(1)-9