Abstract
This work presents a collaborative indoor localization system which provides a lightweight location estimation solution for resource constrained IoT devices. The proposed EcoLoc, an encounter-based collaborative indoor localization system, uses the chance of encounter to enable the sharing and composition of multiple trajectories which are generated by Pedestrian Dead Reckoning. A collaborative version of Conditional Random Field is developed to merge these trajectories and generate the most probable location while significantly shortening the convergence distance compared to the state-of-the-art techniques as particle filter. EcoLoc runs in realtime and can be distributed to resource-limited devices as opposed to running on centralized servers. Using the tablet and WICED Sense IoT platform, the convergence distance can be shorten by up to 40% on Android tablet and up to 50% on the WICED-Sense.