Network Localization on Unit Disk Graphs

TL;DR

This paper introduces a novel approach to network localization on unit disk graphs that leverages the property that lack of connectivity provides useful information, enabling more efficient algorithms for large, sparse networks.

Contribution

It reformulates the localization problem to reduce search space and demonstrates a practical depth-first tree-search algorithm that is efficient for sparse networks with limited anchors.

Findings

Reduced search space for localization problem

Tree-search algorithm with linear memory complexity

Significant speed improvements over traditional methods

Abstract

We study the problem of cooperative localization of a large network of nodes in integer-coordinated unit disk graphs, a simplified but useful version of general random graph. Exploiting the property that the radius $r$ sets clear cut on the connectivity of two nodes, we propose an essential philosophy that "no connectivity is also useful information just like the information being connected" in unit disk graphs. Exercising this philosophy, we show that the conventional network localization problem can be re-formulated to significantly reduce the search space, and that global rigidity, a necessary and sufficient condition for the existence of unique solution in general graphs, is no longer necessary. While the problem is still NP-hard, we show that a (depth-first) tree-search algorithm with memory O(N) ($N$ is the network size) can be developed, and for practical setups, the search complexity and speed is very manageable, and is magnitudes less than the conventional problem, especially when the graph is sparse or when only very limited anchor nodes are available.