適用於具非理想鏈結之無線感測網路之能量效率化分散式最佳線性不偏估計法

Abstract

有鑒於低傳送功率對無線感測網路(wireless sensor network)中之感測器而言為主要需求，如何在有限的總傳送能量下使效能最佳化的設計就愈顯重要。吾人考慮在一個異質性無線感測網路中利用最佳能量分配策略來對一個非隨機信號(deterministic signal)做分散式估計。感測器先將其觀測到之信號取樣成離散訊息後經過瑞利衰減通道(rayleigh fading channel)傳送至融合中心(fusion center)。接著此融合中心利用最佳不偏估計(best linear unbiased estimator)融合規則產生最終估計參數。本論文中提出數種只需得知長期雜訊變異量的統計特性即可求出最佳解之能量分配策略。於前半部，吾人提出的最佳能量分配策略建議對通道環境不佳或觀測品質不佳的感測器降低其所傳送訊息之量化解析度或進而將其關掉來節省能量。每個動態感測器的傳送位元數則由各自的通道衰減、路徑衰減(path loss)、局部觀測雜訊變異量(local observation noise variance)及能量限制來共同決定。於後半部，吾人提出兩個疊代式感測器殘餘能量配置演算法，來進一步提升估計準確度。根據模擬結果，於異質性感測環境中，相較於均衡式能量分配，吾人所提出的最佳能量分配策略可有顯著的效能改善。

As low transmitting power of sensors is a major requirement in wireless sensor networks (WSNs), optimizing their design under energy constraints is of primary importance. We consider an optimal power scheduling problem for the decentralized estimation of a deterministic signal in an inhomogeneous WSN. Sensors quantize their observations into discrete messages, which are transmitted to the fusion center (FC) over rayleigh fading channels. The FC which adopts the best-linear-unbiased-estimator (BLUE) fusion rule generates a final estimate. In this thesis, the optimal power allocation strategies which simply rely on long-term noise variance statistics are presented. In the first part, the proposed power scheduling scheme suggests that the sensors with bad channels or poor observation qualities should decrease their quantization resolutions or simply be shut off to save power. The bit load of each active sensor is determined jointly by individual channel fading gain, path loss, local observation noise variance, and the energy constraint. In the second part, two iterative allocation algorithms of residual energy at sensors are proposed to further enhance estimation accuracy. Numerical results show that in inhomogeneous sensing environments, significant performance improvement is possible when compared to the uniform quantization strategy.