Deep Game Location Acquisition for Pokemon GO on Mobile Devices

Abstract

Advanced AR (Augmented Reality) games such as Pokemon Go attract million and million players because they integrate various sensors to impress excellent user experiences on mobile devices. In this paper, for energy saving on the mobile devices, we can use an accelerometer and a magnetometer to estimate the current user's location approaching interest of points on a game map without active GPS sensors and wireless network. In addition, we create an effective data structure with R-tree and adopt deep learning to efficiently answer the location query of Pokemons on the game location database. Our algorithm outperforms previous approaches in not only smaller running time but also the better power consumption for the AR games on the mobile devices