When and Why the Topological Coverage Criterion Works

Abstract

In their seminal work on homological sensor networks, de Silva and Ghrist showed the surprising fact that it's possible to certify the coverage of a coordinate-free sensor network even with very minimal knowledge of the space to be covered. Here, coverage means that every point in the domain (except possibly those very near the boundary) has a nearby sensor. More generally, their algorithm takes a pair of nested neighborhood graphs along with a labeling of vertices as either boundary or interior and computes the relative homology of a simplicial complex induced by the graphs. This approach, called the Topological Coverage Criterion (TCC), requires some assumptions about the underlying geometric domain as well as some assumptions about the relationship of the input graphs to the domain. The goal of this paper is to generalize these assumptions and show how the TCC can be applied to both much more general domains as well as very weak assumptions on the input. We give a new, simpler proof of the de Silva-Ghrist Topological Coverage Criterion that eliminates any assumptions about the smoothness of the boundary of the underlying space, allowing the results to be applied to much more general problems. The new proof factors the geometric, topological, and combinatorial aspects, allowing us to provide a coverage condition that supports thick boundaries, k-coverage, and weighted coverage, in which sensors have varying radii. From Sensor Coverage to Data CoverageProblems in Homological Sensor Networks (HSNs) are usually stated in the vocabulary of sensor networks. There are sensors and coverage regions, and an important problem is to determine if the sensing region of a collection of sensors covers a given domain, given only the neighborhood relationships between the sensors and some indication of which sensors are near the boundary. The locations (coordinates) of the points is not assumed, nor is the shape (topology) of the domain to be covered. Although phrased in the language of sensor networks, the problem may be understood more generally as one of data coverage; it answers when a data set * Supported by NSF grants CCF-1464379 and CCF-1525978.† University of Connecticut. ‡ University of Connecticut. § University of Connecticut.sufficiently covers a domain, and holes in coverage can be viewed as gaps in the data. A surprising result by de Silva & Ghrist is that there exists a computable, sufficient condition called the Topological Coverage Criterion (TCC) to certify coverage without knowing the locations of the sensors when the domain's boundary is smooth [4]. All that is required is that the sensors have unique identifiers, can detect nearby sensors, can differentiate whether neighboring sensors are "close" or "very close" (in a technical sense to be defined below), and can detect if the boundary of the domain is close. Developed over a series of papers [8,17,5,4], the most general version of the TCC can be understood as a purely geometric problem on point sets P in R d with unknown coordinates in an unknown d...

Visualizing Sensor Network Coverage with Location Uncertainty

“…Their model, while based on unknown node location, nevertheless assumes those locations are deterministic. Cavanna et al [8] recently generalized the assumptions on the boundaries to make the results applicable to general domains. Gamble et al [14] extend this concept further to consider a timevarying network.…”

“…One could also explore sparsification schemes that simplifies a redundant cover using appropriate choice of generators [10]. If our tool could be extended to include polygonal boundaries instead of just rectangular ones, one could explore further the various boundary configurations within TCC [8].…”

The Generalized Persistent Nerve Theorem

“…It is at the heart, either implicitly or explicitly, of many foundational algorithms in the rapidly growing field of topological data analysis. Example problems include surface reconstruction [2,3,4], function reconstruction [5], homology inference [6,7,8] , coordinate-free sensor network coverage [9,10], shape analysis [11], data modeling [12,13,14], and clustering [15,16].…”

“…This also has implications for triangulations of covers of surfaces that have marginal measurement errors, because the errors can cause the cover elements to no longer be convex for example. Nerves are also in coverage testing for homological sensor networks [10,9], however the idealized model of Euclidean balls as coverage regions differs significantly from the very jagged coverage regions measured in practice, particularly when taking into account the affect of physical obstacles on real-life sensors' detection ranges.…”

Stability of 2-Parameter Persistent Homology

“…In Section 3, we prove the main stability results. Some of these depend on the multicover nerve theorem for bifiltrations, which we prove in Section 4, following [14]. Appendix A presents our computational study of the stability of degree-Rips bifiltrations.…”