Functional characterization of fault tolerant integration in distributed sensor networks

Abstract

Fault-tolerance is an important issue in network design because sensor networks must function in a dynamic, uncertain world. In this paper, we propose a functional characterization of the fault-tolerant integration of abstract interval estimates. This model provides a test bed for a general framework which we hope to develop to address the general problem of faulttolerant integration of abstract sensor estimates. We further propose a scheme for narrowing the width of the sensor output in a specific failure model and give it a functional representation.The main distinguishing feature of our model over the original Marzullo's model is in reducing the width of the output interval estimate significantly in most cases where the number of sensors involved is large.

Statistical Methods in Ai: Rare Event Learning Using Associative Rules and Higher-Order Statistics

“…So one cannot use non-parametric clustering when the number of clusters are not known a priori. So we argue the use of explicit overlap function (Prasad et al, 1991) feature induction in these cases works well. We further extend sensor data streams for fast predictive overlap models which can handle real-time constrains.…”

“…Most of the learning algorithms still have the training phase which make them unsuitable for real-time analytics. In the domain of sensor networks, distributed measurements have been successfully calibrated (Prasad et al, 1991) with fault tolerant (Prasad et al, 1993) algorithms to avoid false alarms. The techniques used in feature extraction and real-time fault tolerance does not address rare event learning directly.…”

“…Electronically generated streams do not contain labeled data and thus not suitable for supervised algorithms. We propose a powerful overlap function (Prasad et al, 1991) metric to identify the actionable events. The stream pre-processing algorithm is resilient to noise and uses instance based learning and at the same time, works in unsupervised settings (it can have labels).…”

Detection and diagnosis of data inconsistency failures in wireless sensor networks

“…Ganesan et al employed both approaches to enable the energy-efficient recovery of independent and geographically-correlated failures [8]. A review of the literature finds related studies investigating the implementation of reliable routing using arbitrary network topologies [15,16], the characterization of sensor fault modalities [17], and the grid coverage of surveillance and target locations in distributed sensor networks [18].…”

Efficient and Fault-Tolerant Feature Extraction in Wireless Sensor Networks

“…Therefore it's important to guarantee that faulty behavior of individual components does not affect the overall system behavior. Some of the early work in the area of distributed sensor networks focuses on reliable routing with arbitrary network topologies [18], [19], characterizing sensor fault modalities [3], [4], tolerating faults while performing sensor integration [20], and tolerating faults while ensuring sensor coverage [17]. A mechanism for detecting crash faults in wireless sensor networks is described in [29].…”