Bio-Inspired Data Acquisition in Sensor Networks

I. OVERVIEW Field estimation is an important application of wireless sensor networks. This type of application deploys sensor nodes in a specific region to remotely sense space-temporally variable processes. The spatial and temporal frequencies of sampling directly impact the quality of the estimation. Therefore, there is a tradeoff between the frequency and the number of samples transmitted, which is related to the energy consumption of the network. The most common solution to reduce the number of samples transmitted to the sink is to identify areas where different nodes present similar readings and reduce the sampling spatial frequency by deactivating some of these nodes [1], [2]. Nevertheless, this approach is not efficient in regions of the field with sharp spatial variations, or borders. There is, however, another sampling dimension to regard: the temporal dimension. While many works focus the deactivation of nodes, little effort has been made to determine when active nodes must collect and transmit samples. We propose to allow nodes to identify patterns in the behavior of the sensed processes and report only uncommon measures. This environment-aware behavior is similar to the response of living beings to the surrounding events.People and animals are continually receiving stimuli; however, it is impossible to handle consciously all these stimuli. The organisms develop the notion of periphery and center of attention [3]. While the periphery is handled in a sub-conscious manner, the center of attention is the event consciously treated. Generally, an event migrates from the periphery to the center of attention when it differs much from the periphery as a whole. The proposed bio-inspired scheme exploits specific features of the monitored processes in order to reduce the number of transmitted samples. Moreover, the scheme is fully distributed as each node identifies its own periphery (Section II). Thus, each node sends to the sink only the samples that differ from the usually sensed by the node. This scheme can save resources for nodes located in border regions as in smooth varying regions. This temporal technique is orthogonal to space-frequency-reduction techniques and both techniques can be used together to improve the system performance.