Optimizing distortion for real-time data gathering in randomly deployed sensor networks

摘要

In several wireless sensor network applications, it is required to perform real-time reconstruction of the data field being sensed by the network. This task is generally carried out at a central location, e.g. sink node, using a continuous data gathering phase and relying on the known correlation properties of the underlying data field. Estimating the overall spatial and temporal distortion in the reconstructed field is an important step toward deciding the number of sensors to be deployed and the data collection algorithm to be used. However, estimating distortion in arbitrary networks is a challenging task. Existing work has focused on regular network deployments such as one- and two-dimensional girds. Such deployments are deemed infeasible in a realistic environment. In this paper, we consider one- and two-dimensional random networks. For the analysis purposes, we assume that the nodes are randomly deployed following Poisson distribution. We determine the total distortion function given the correlation coefficients of the field while assuming a simple data gathering protocol. Based on this, we also determine the optimal number of nodes to be deployed in the field that will minimize distortion.