Improved RSS-based Localization Using Regression Approach in UWSNs

This paper proposes an approach to solve an issue of location estimation in underwater wireless sensor networks, where received signal strength (RSS) measurements are collected for localization. We first introduce an RSS model which reflects the environment characteristics, then formulate the problem of estimating sensor locations by extracting the RSS measurements and the correlation among them. Second, from a certain subset of potential anchor nodes, we estimate the sensor location while keeping the error, communication overhead, and response time low. By applying linear regression technique, we also find the first-order polynomial that best fits a given set of RSS measurements. Then, error control is executed to cancel out the impact of noisy ranging effect during localization process. The results allow us to access a quick coarse range for estimating a predicted current location. Simulation results indicate the effectiveness of the proposed approach.