A few bits are enough: Energy efficient device-free localization

Abstract

A Device Free Localization (DFL) system can locate and track people wearing no wireless devices, due to the fact that a person standing at different locations attenuates wireless links differently. Since the DFL system usually consists of battery-powered sensors, energy efficiency is a critical issue. However, existing works mainly focus on improving localization accuracy by designing various metrics to characterize wireless link attenuation, and none of them considers energy efficiency, specifically. We present EE-Loc, an energy efficient localization system, for locating and tracking people with higher energy efficiency and comparable localization accuracy compared to the state-of-the-art DFL systems. EE-Loc has two main energy efficient components. First, EE-Loc has a radio tomographic imaging (RTI) component that uses only one bit information to describe link attenuation. The one bit information is derived from the Kullback–Leibler divergence (KL-divergence) of Received Signal Strength (RSS), and we prove that RTI with this one bit information is sufficient for localization. Second, EE-Loc has a tracking component that deactivates many unnecessary links through predicting the person’s location with Kalman filter to reduce energy consumption. We build a test-bed of EE-Loc using 16 sensors. The experimental results indicate that EE-Loc improves energy efficiency by 27.05% compared to Spin*> for locating a person, and reduces link measurements by 41.91% for tracking a person, without compromising the localization accuracy.