Life Cycle of a SLAM System: Implementation, Evaluation and Port to the Project Tango Device

Abstract

Augmented Reality (AR) applications are becoming popular recently. Among other things, AR requires a precise real-time tracking to work properly. Simultaneous Localization and Mapping (SLAM) is one way to perform this task. Commonly used in robotics applications, SLAM creates a map of the environment to use it as input to compute the pose while uses the pose to increment the map. At the same time, mobile devices are evolving faster lately. Along with more processing power and memory capabilities, they are being embedded with several powerful resources, such as depth sensors. Taking this into account, this work introduces STAM, a Simple Tracking and Mapping system that was developed in desktop and evaluated in a challenging scenario. Additionally, STAM was ported to a mobile version, using the Android platform and Google's Tango tablet device. Finally, the system was evaluated concerning its desktop version. The desktop version presented better tracking performance in simple scenarios with respect to reprojection error, but it presented a few drawbacks when dealing with the most complex ones. Regarding the mobile version, it proved to be slower than its desktop counterpart. However, it was more precise.