Sparsifying Disk Intersection Graphs for Reliable Connectivity

Abstract: The intersection graph induced by a set $\Disks$ of $n$ disks can be dense. It is thus natural to try and sparsify it, while preserving connectivity. Unfortunately, sparse graphs can always be made disconnected by removing a small number of vertices. In this work, we present a sparsification algorithm that maintains connectivity between two disks in the computed graph, if the original graph remains well-connected'' even after removing an arbitraryattack'' set $\BSet \subseteq \Disks$ from both graphs. Thus, the new sparse graph has similar reliability to the original disk graph, and can withstand catastrophic failure of nodes while still providing a connectivity guarantee for the remaining graph. The new graphs has near linear complexity, and can be constructed in near linear time. The algorithm extends to any collection of shapes in the plane, such that their union complexity is near linear.