Dynamic Connectivity in Disk Graphs

Let $S$ be a set of $n$ sites, each associated with a point in $\mathbb{R}^2$ and a radius $r_s$ and let $\mathcal{D}(S)$ be the disk graph on $S$. We consider the problem of designing data structures that maintain the connectivity structure of $\mathcal{D}(S)$ while allowing the insertion and deletion of sites. For unit disk graphs we describe a data structure that has $O(\log^2n)$ amortized update time and $O((\log n)/(\log\log n))$ amortized query time. For disk graphs where the ratio $\Psi$ between the largest and smallest radius is bounded, we consider the decremental and the incremental case separately, in addition to the fully dynamic case. In the fully dynamic case we achieve amortized $O(\Psi \lambda_6(\log n) \log^{9}n)$ update time and $O(\log n)$ query time, where $\lambda_s(n)$ is the maximum length of a Davenport-Schinzel sequence of order $s$ on $n$ symbols. This improves the update time of the currently best known data structure by a factor of $\Psi$ at the cost of an additional $O(\log \log n)$ factor in the query time. In the incremental case we manage to achieve a logarithmic dependency on $\Psi$ with a data structure with $O(\alpha(n))$ query and $O(\log\Psi \lambda_6(\log n) \log^{9}n)$ update time. For the decremental setting we first develop a new dynamic data structure that allows us to maintain two sets $B$ and $P$ of disks, such than at a deletion of a disk from $B$ we can efficiently report all disks in $P$ that no longer intersect any disk of $B$. Having this data structure at hand, we get decremental data structures with an amortized query time of $O((\log n)/(\log \log n))$ supporting $m$ deletions in $O((n\log^{5}n + m \log^{9}n) \lambda_6(\log n) + n\log\Psi\log^4n)$ overall time for bounded radius ratio $\Psi$ and $O(( n\log^{6} n + m \log^{10}n) \lambda_6(\log n))$ for general disk graphs.