Exploiting Ground Reflection for Robust 3D Smartphone Localization

The reliability of an indoor localization system is often limited by the capability of the systems of distinguishing the line-of-sight signals in environments with multipath. Besides, estimating the height of the target can be challenging due to dilution of precision (DOP) and non line of sight (NLOS) situations. However, the reflections can be used to infer information about the scenario. We present a novel approach which exploits the ground reflections to reduce the error in three dimensional localization and detect reliably the line-of-sight signals. The presented approach uses the geometry of the reflections as the model of the RANSAC method. In this model, ground reflections and line of sight signals can be found just solving a quadratic equation, which reduces the computational power required compared to other approaches. The positions are estimated using local optimization algorithms. In real experiments, the median error in the height estimation was reduced from 25 cm to 4.2 cm.