Abstract
Indoor localization techniques proposed to date have assumed costly resources in terms of computation, power, or sensing modality for many wearable end-devices in the Internet of things (IoT). To make localization a universal feature for IoT devices, we propose EcoLoc, an indoor localization system using collaborative version of Conditional Random Fields (CCRF) integrated with our encounter model to generate the most probable locations. We have implemented EcoLoc on the Android tablet and the Broadcom WICED Sense IoT platform with a lower-power MCU, miniature inertial sensors, and Bluetooth Low-Energy (BLE) radio. Experimental results show that while operating without the aid of beacons, compared to the non-collaborative CRF, EcoLoc can shorten the convergence distance by up to 40% on tablet, and up to 50% on the WICED-Sense while incurring an extra current consumption of 15 mA.