Serial distributed detection strategies for wireless sensor networks

Abstract

The interest in wireless sensor networks (WSNs) has been significantly increased over the past years due to their promising future in wide range of application areas like health, military and home. Various technical issues of WSNs have been investigated recently such as topology management, efficient routing protocols and collaborative signal processing. This thesis considers serial distributed detection in WSNs. Unlike traditional distributed detection algorithms where error-free transmissions of local decisions to the fusion center are assumed, lossless communication is not applicable in WSNs since wireless transmission channels are subjected to fading and interference. Suggested distributed detection algorithms in WSNs should deal with the channel uncertainty due to fading and noisy effects of non-ideal channel under low power transmission. In this thesis, we first propose suboptimal fusion rules to the optimal fusion rule for serial distributed detection. In particular, we derive the low and high SNR approximations of the optimal rule in order to relieve some requirements of the optimal fusion rule. Then, we investigate effects of node failure to the serial distributed detection performance and we suggest more robust decision fusion rules under node failure. Lastly, we analyze effects of decision feedback at serial network topology. In order to improve serial distributed detection performance we propose feedback strategies and derive appropriate decision fusion rules for suggested strategies.