Resource aware distributed detection and estimation of random events in wireless sensor networks

Abstract

Bu tezde, telsiz duyarga aglar icin daginik tesbit ve kestirim problemleri kaynak duyarliligi altinda incelenmistir. Ilk olarak daginik tesbit sorunu ele alinmistir. Amaclarin tumlestirme merkezi karar hata olasiliginin ve agin toplam enerji sarfiyatinin en aza indirgenmesi oldugu bir cok amacli en iyileme problemi tanimlanmistir. Bu problemden elde edilen sonuclar, ulasilabilecek en dusuk hata olasiligina yakin ama onemli olcude enerji tasarrufu saglayan karar esiklerini gostermistir.Olay yerinin hassas kestirimi turlu uygulamalar acisindan onemlidir. Tum duyarga verisini tek bir defada gonderilmesi yerine, tekrarli bir kestirim yontemi sunulmaktadir. Az sayida duyarganin verisi kullanilarak olay yeri once kabaca kestirilir. Yontemin bir diger tekrarinda, verisi istenecek duyargalar musterek bilgi veya sonsal Cramer-Rao alt siniri esasli metrikler yardimiyla secilmektedir. Musterek bilgi veya sonsal Cramer-Rao alt siniri temelli duyarga secim metrikleri kusursuz iletim kanallari varsayimi altinda benzer kestirim basarimi gosterseler de, sonsal Cramer-Rao alt siniri temelli duyarga secim metriginin karmasikligininn musterek bilgi temelli duyarga secim metrigine gore daha azdir.Son olarak, olcum gurultusunun her bir duyarga icin farkli oldugu ayrisik durum incelenmistir. Her bir duyarganin olcumunu temsilde kullandigi kuantalama hizi, olcumunun isaret-gurultu oranlarina bagli olarak belirlenmistir. Verilen bandgenisligi altinda, genel bir daginik kestirim problemi incelenmistir. Toplam bandgenisligini asmamak icin belirli bir kuantalama hizinda temsil edilen olcum, tumlestirme merkezine yine belirli bir gonderim olasiligi ile iletilmektedir. Her bir kuantalama hizinin, en uygun gonderim olasiligini bulmak icin toplam bandgenisligini ve band kullanimi kistaslari altinda ortalama Fisher bilgisinin tersi en aza indirgenmistir.

In this dissertation, we develop several resource aware approaches for detection and estimation in wireless sensor networks (WSNs). For the distributed detection problem in WSNs, we first find the trade-off solutions between two objectives: minimizing the probability of error and minimizing the total energy consumption. The Pareto-optimal solutions provide significant energy savings at the cost of a slight increase in the probability of error from its minimum achievable value.We next propose an iterative source localization algorithm where a small set of anchor sensors first arrive at a coarse location estimate. Then a number of non-anchor sensors are selected in an iterative manner to refine the location estimate. The iterative localization scheme reduces the communication requirements as compared to the one-shot location estimation while introducing some estimation latency. To select the sensors at each iteration, two metrics are proposed which are derived based on the mutual information and the posterior Cramer-Rao lower bound. In terms of computational complexity, the PCRLB-based metric is more efficient as compared to the MI-based metric, and under the assumption of perfect communication channels both sensor selection schemes achieve the similar estimation performance.We finally consider a heterogeneous sensing field and define a distributed parameter estimation problem where the quantization data rate of a sensor is determined as a function of its observation SNR. In order to find the optimal transmission probability of each possible quantization rate, the inverse of the average Fisher information is minimized subject to the total bandwidth and bandwidth utilization constraints.