Rank-Based Statistical Localization in Sensor Networks

Rank-Based Statistical Localization in Sensor Networks

Abstract:

In the applications of WSNs, positions of sensor nodes are critical prerequisite information for successful operation. Although remarkable progresses has been achieved in the so-called Range-Based methods utilizing inter-node range measurements, in realistic environment, it is generally hard for the distance measurement to exactly match the true Euclidean value, owing to the randomness and unpredictability of notorious wireless channels. Localization qualities of Range-Based methods, as a result, are closely related to the channel conditions; once the radio channel is contaminated with large measurement noise, localization accuracy would drastically deteriorate. We try to improve localization performance under the atrocious network conditions mentioned above by providing a robust Rank-Based method, as a special case and improvement of Range-Based methods. With a predefined signal model, locations of sensor node could be statistically estimated merely based on rank-order information of distance measurements. Without applying actual metric measurement values, our method has remarkable advantage over traditional Range-Based methods which suffer from the measurement insufficiency problem, and is more robust to the fluctuation of measurement noise.