Infrastructure Engineering - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemTopological relationships between continuously evolving regions in geosensor networks( 2012)A key challenge facing many applications of new geosensor networks technology is to derive meaningful spatial knowledge from low-level sensed data. This research aims at designing a unified framework for computing global topological relationships and reasoning about their gradual changes from local point-wise sensor readings in the network. The thesis presents a formal model to represent spatial regions and fundamental topological relationships between these regions in discrete space (i.e., a geosensor network). Each node can provide only local observations of two spatial regions in the network. Consequently, each node can only observe local topology of two spatial regions and its node state can be locally computed using a 4-bit binary number. A set of fundamental topological relationships between two regions can then be characterized by these local node states. The main contribution of the research is to provide a computational framework for the detection of global, high-level, qualitative relationships and relationship changes from local, low-level, quantitative sensor measurements. The framework relies only on local information about a node’s own node state and its immediate (one hop) neighbors’. The computational framework enables the design and development of efficient and reliable algorithms for computing static topological relationships and dynamic relationship changes in the context of a geosensor network. The overall performance of the decentralized algorithms was evaluated empirically using agent-agent simulations. As expected, the BAG (boundary-based aggregation) algorithm outperforms the SAG (state-based aggregation) algorithm for computing static topological relationships between spatial regions. The BAG algorithm exploits the autocorrelation of spatial regions and restricts the computation on the boundary nodes in the network. The experimental results have shown that the DM (decentralized-movement) algorithm outperforms the decentralized algorithm for reasoning about gradual changes in topological relationships. The DM algorithm enables a coarser temporal monitoring of spatial regions and can still operate efficiently and reliably by relaxing some strong simplifying assumptions.