Robust and compact geographic/topological structures for routing and information dissemination in wireless sensor networks

• Authors: Leonidas J. Guibas;Qing Fang
• Venue: Robust and compact geographic/topological structures for routing and information dissemination in wireless sensor networks
• Year: 2007

This dissertation describes a new approach in organizing hundreds or more of autonomous nodes, each with sensing, computing and wireless communication capabilities. We build geometric and topological structures in a network during a discovery phase. These structures need to be compact enough to be stored at each node, yet robust enough not to be quickly invalidated by radio link dynamics. When nodes' geographic coordinates are known, we use the geometric structures to reduce average route length and to improve load balancing in geographic routing. When nodes' geographic coordinates are unknown, we can use the topological structures in devising hop-count based virtual addressing schemes to name the nodes and to guide routing and information dissemination. First, we study a geometric structure that we call a network hole. We define a network hole to be the region enclosed by a polygonal cycle that connects the nodes where local minima in greedy geographic routing can appear. We propose a distributed algorithm to identify the existence of such holes given nodes' location information and to identify cycles that mark the boundaries of the holes. We prove the correctness of the algorithm. Second, we propose GLIDER, a scheme that uses a two-tier combinatorial structure for network addressing and routing. The algorithm works by first discovering the global topology of the sensor field, partitioning the nodes into connected tiles. A tile is a region where the node placement is sufficiently dense that greedy methods can work well. As a byproduct, we map out the tile adjacency graph, the combinatorial Delaunay graph. The tile adjacency graph is then used for global route planning. We define a local coordinate system where a node's coordinates are measured by its distances to a selection of nearby nodes (landmarks) and propose a distance function for greedy routing within each tile. We prove any point in a continuous plane can be reached by greedy descent on this distance function. Third, we study data-centric information storage and retrieval for large-scale networks when nodes' geographic coordinates are unknown. We show how the combinatorial structures proposed in GLIDER could be used in solving the problem.