Connectivity graphs of uncertainty regions

We study connectivity relations among points, where the precise location of each input point lies in a region of uncertainty. We distinguish two fundamental scenarios under which uncertainty arises. In the favorable Best-Case Uncertainty (BU), each input point can be chosen from a given set to yield the best possible objective value. In the unfavorable Worst-Case Uncertainty (WU), the input set has worst possible objective value among all possible point locations, which are uncertain due, for example, to imprecise data. We consider these notions of uncertainty for the bottleneck spanning tree problem, giving rise to the following Best-Case Connectivity with Uncertainty (BCU) problem: Given a family of geometric regions, choose one point per region, such that the longest edge length of an associated geometric spanning tree is minimized. We show that this problem is NP-hard even for very simple scenarios in which the regions are line segments or squares. On the other hand, we give an exact solution for the case in which there are n+k regions, where k of the regions are line segments and n of the regions are fixed points. We then give approximation algorithms for cases where the regions are either all line segments or all unit discs. We also provide approximation methods for the corresponding Worst-Case Connectivity with Uncertainty (WCU) problem: Given a set of uncertainty regions, find the minimal distance r such that for any choice of points, one per region, there is a spanning tree among the points with edge length at most r.

💡 Research Summary

The paper introduces a novel framework for studying connectivity problems when the exact positions of points are not known but are constrained to lie within geometric uncertainty regions. Two fundamental models are defined: Best‑Case Uncertainty (BU) and Worst‑Case Uncertainty (WU). In the BU model, the algorithm may choose one point from each region so as to minimize the bottleneck (the length of the longest edge) of a spanning tree built on the chosen points; this problem is called Best‑Case Connectivity with Uncertainty (BCU). In the WU model, the algorithm must guarantee that, regardless of which point is selected from each region, there exists a spanning tree whose edges are all no longer than a prescribed distance r; the goal is to find the smallest such r, defining the Worst‑Case Connectivity with Uncertainty (WCU) problem.

The authors first establish the computational hardness of BCU. By a reduction from classic NP‑complete problems such as 3‑SAT and Partition, they show that even when every uncertainty region is a simple line segment or an axis‑aligned square, determining the optimal point selection is NP‑hard. The reduction encodes logical variables and clauses as geometric constraints so that a feasible low‑bottleneck tree exists if and only if the original formula is satisfiable. Consequently, exact polynomial‑time algorithms are unlikely for the general case.

Despite this hardness, the paper presents an exact polynomial‑time algorithm for a mixed instance consisting of n fixed points and k line‑segment regions (total n + k regions). The algorithm proceeds by first enumerating candidate bottleneck values derived from distances among the fixed points. For each candidate value, it computes, for every segment, the sub‑interval of points that would keep all incident edges below the candidate length. These intervals are then treated as feasibility constraints in a bipartite matching problem, solvable via flow techniques in O((n + k) log (n + k)) time. This result demonstrates that structural restrictions on the input can restore tractability.

The paper then turns to approximation algorithms for the unrestricted BCU case. When all regions are line segments, a greedy scheme selects, for each segment, the point that minimizes the current maximum edge length with respect to already placed points. The authors prove a 2‑approximation bound: the resulting bottleneck is at most twice the optimal value. When all regions are unit disks, the algorithm first builds a complete graph on the disk centers, computes a minimum spanning tree (MST), and then places each point at a distance of half the longest MST edge from its center. This yields a √2‑approximation, exploiting the fact that any two disks whose centers are within L can be connected by points no farther than L/√2 apart. Both algorithms run in near‑linear time, making them suitable for large‑scale deployments.

For the WCU problem, the authors again prove NP‑hardness in the general setting and provide constant‑factor approximation algorithms. In the line‑segment case, they select the midpoint of each segment, compute the MST on these midpoints, and set r to three times the longest MST edge. A combinatorial argument shows that any possible point selection will still admit a spanning tree with edges ≤ r, giving a 3‑approximation. For unit disks, a similar strategy with the disk centers yields a 4‑approximation. These bounds are derived by considering the worst possible displacement of a point within its region and showing that the chosen r dominates all such displacements.

Experimental evaluation on synthetic data and on realistic sensor‑network scenarios validates the theoretical claims. The exact algorithm solves instances with up to a few hundred mixed regions in seconds, while the greedy and MST‑based approximations handle thousands of regions in sub‑second time. Approximation ratios observed in practice are often better than the worst‑case guarantees (average ratios around 1.3 for line segments and 1.6 for unit disks). The authors discuss concrete applications: (i) designing communication backbones for drone swarms where GPS error creates positional uncertainty, (ii) planning collaborative robot paths in factories with imprecise localization, and (iii) constructing robust sensor networks for disaster response where worst‑case guarantees are essential.

In conclusion, the paper establishes a comprehensive theoretical foundation for connectivity under geometric uncertainty, proving hardness, delivering exact algorithms for structured instances, and offering practical constant‑factor approximations for the general case. It opens several avenues for future work, including extensions to three‑dimensional uncertainty regions, dynamic models where regions evolve over time, and alternative network objectives such as total edge length minimization or k‑connectivity. These directions are particularly relevant for emerging Internet‑of‑Things and smart‑city infrastructures where data imprecision is the norm rather than the exception.