A linear regression based cost function for WSN localization

Localization with Wireless Sensor Networks (WSN) creates new opportunities for location-based consumer communication applications. There is a great need for cost functions of maximum likelihood localization algorithms that are not only accurate but also lack local minima. In this paper we present Linear Regression based Cost Function for Localization (LiReCoFuL), a new cost function based on regression tools that fulfills these requirements. With empirical test results on a real-life test bed, we show that our cost function outperforms the accuracy of a minimum mean square error cost function. Furthermore we show that LiReCoFuL is as accurate as relative location estimation error cost functions and has very few local extremes.