| Although the specifics of sensor network deployment scenarios are entirely application domain specific, it is envisioned that wireless sensor networks are densely deployed over large monitoring areas. The post-deployment discovery of location and topological information in arbitrarily deployed wireless sensor network is critical to the effective use of a wireless sensor network. Fundamental to wireless sensor networks is the problem of developing a low-cost GPS-free localization technique. Therefore, we first present ANIML, a straightforward, iterative, anchor-free, range-aware, relative localization technique for wireless sensor networks. Through simulation, despite using a non-idealized MAC, we show that ANIML provides good relative localization in uniform, C-shaped and non-uniform topologies. However, while knowing the physical positions of every node in the network provides information about the deployed topology of a wireless sensor network, it does not provide a complete view of a network's topology, such as the shape of the network deployment. The boundaries of the network have a physical correspondence to the environment in which the sensors are deployed. Therefore, we next present a robust, distributed technique that addresses the problem of boundary recognition in wireless sensor networks. We show that our boundary recognition technique constructs accurate perimeters (i.e. correctly bounding all nodes) in randomly deployed topologies of varying densities, perturbed grid topologies of varying densities and in sparsely populated/low-density topologies, in addition to highly irregularly shaped connectivity holes and networks. Lastly, we address the problem of edge detection in wireless sensor networks. Edge detection is the idea of reducing data analysis overhead through the geometric identification of sensed phenomena within a sensor network. We adapt our boundary recognition technique to address the more general problem of edge detection in wireless sensor networks. Our edge detection technique keeps inter-group communication to a minimum, while still constructing correct outer perimeters in the presence of anomalous perimeter crossings and phenomena wholly surrounded by other phenomena. We show that our technique constructs accurate perimeters in randomly deployed topologies of varying densities, perturbed grid topologies of varying densities and in sparsely populated/low-density topologies, in addition to highly irregularly shaped phenomena and networks. |
