Challenges of Large-Scale Augmented Reality on Smartphones

Smartphones have been identified as most promising future devices for an Augmented Reality (AR) mass market. However, their use puts considerable constraints on the design and composition of AR applications. The key problem is to find a registration mechanism for accurate six degree of freedom (6DOF) self-localization with respect to the environment. Approaches based on Computer Vision (CV) have been shown to be promising, but the feasibility of many CV methods on smartphones is questionable. In this paper we discuss current and future challenges faced in developing AR on smartphones, in particular for large and unconstrained outdoor environments. We focus on the registration task, giving a survey and an assessment of existing approaches from AR and CV. From this survey, we identify a set of important issues still seeking for practical solutions, both in terms of the fundamental registration problem and for making AR on smartphones a unique experience. As will become apparent, despite recent advances, we are still far from arriving at a universal solution to the problem.