Abstract
Network systems consisting of a large number of small, low-power, wireless sensing devices offer new ability to observe the physical world, especially for monitoring previously unobservable phenomena. In this thesis, we consider answering range queries that request certain of statistics on a subset of nodes in wireless sensor networks in an energy and space efficient way. An algorithm based upon a rate-distortion multi-resolution wavelet coding technique is presented for deliberately selecting coefficients on the continuous sensing data streams. The design goal is to approximate the requested statistics with bits delivered from sensors to the base station as few as possible and with limited space usage. We also deduce a rate-distortion control formula to determine how to optimally allocate bit rate for sensors under constrained total bit rate so as to minimize the overall distortion of the collected data. The experimental results show the R-D proposed method efficiently reduces the transmission amount using the combination of linear regression, spatial correlation and wavelet coding. Finally, we verify our R-D optimal allocation, which can efficiently allocate appropriate bit rate to every sensor node, so as to archive minimizing distortion for range queries.