Distributed algorithms for bottleneck identification and segmentation in 3D wireless sensor networks

Abstract

Segmentation decomposes a network with complex and irregular shape into a set of subnetworks, each under a simple boundary condition without bottlenecks. It has a wide spectrum of applications in routing, coverage, localization, backbone construction and maintenance, and in-network data centric storage and retrieval. To our best knowledge, this is the first work that tackles the segmentation problem in 3D wireless sensor networks. We propose a fully distributed 3D segmentation scheme with mere network connectivity information. Each node on boundary computes its injectivity radius, which reflects the narrowness of the corresponding boundary area and thus is employed to locate the undesired bottlenecks. A cluster of connected boundary nodes with similar smallest injectivity radii form a bottleneck segment. A recursive process is applied to identify a set of such bottlenecks, which together divide the network boundary into segments. An internal non-boundary node simply joins the nearest segment, thus completing the segmentation of the entire 3D sensor network. Our simulations show that the proposed algorithm works efficiently under various sensor network models with different boundary conditions and noise levels, always yielding appropriate segmentation results. We further demonstrate that segmentation can effectively promote the performance of a range of applications in 3D wireless sensor networks.