Detection performance in self-organized wireless sensor networks

In designing a wireless sensor network, we suggest that one of the primary performance metrics should be the ability of the sensor network to detect, in a distributed manner, events of interest. Since wireless sensor networks will often have many nodes and will be deployed in remote environments, it is important that they have the ability to self-organize, forming clusters of sensor nodes. Thus, we consider the interaction of the self-organization and the distributed detection algorithms, with the primary goal of determining and optimizing the parameters that are responsible for the detection performance. In this paper, the self-organization is done using the linked cluster algorithm (LCA). The algorithm is distributed, and it uses two TDMA frames to allow each node to independently determine its neighbors. Every node uses its local connectivity matrix and its identification number to classify itself as either: (1) a cluster head, (2) a gateway, or (3) an ordinary node.