Skip to Main Content
Wireless Sensor Networks (WSNs) are ad hoc networks formed by tiny, low powered, and low cost devices. WSNs take advantage of distributed sensing capability of the sensor nodes such that several sensors can be used collaboratively to detect events or perform monitoring of specific environmental attributes. Since sensor nodes are often exposed to harsh environmental elements, and normally operate in an unsupervised fashion over long periods of time, within their MTBF, some of them are subject to partial failure in form of A/D readings that are permanently off the correct levels. Additionally, due to glitches in timing and in hardware or software, even healthy sensor nodes can occasionally report readings that are outside of the expected range. In this paper we present a novel approach that combines spatial and temporal correlation of the data collected by neighboring sensors to combat both error modes described above. We combine the weighted averaging algorithm across multiple sensors, with the LMS adaptive filtering of individual sensor data, in order to improve fault tolerance of WSNs. We present performance gains achieved by combining these methods; and analyze the computational and memory costs of these algorithms.