Bayesian sensor fusion of Wi-Fi signal strengths and GNSS code and carrier phases for positioning in urban environments

In urban environments GNSS signals can be either blocked by buildings, so that the number of satellites with direct line-of-sight (LOS) is reduced considerably, or reflected by surfaces, so that signals from the same satellite are received over multiple paths. Taking into account all available signals would typically result in a rather poor position estimate. It is therefore essential to distinguish LOS from non-LOS or multipath-contaminated signals and include this information in the GNSS process model. This is done using a classification algorithm based on code (pseudorange) and carrier phase observations. On the other hand, Wi-Fi fingerprinting is complementary to GNSS in the sense that this approach benefits from signal degradations caused by shadowing through obstacles and reflections leading to unique variations in the radio map and less ambiguity in the mapping of signal strength measurements to positions. The fusion of GNSS and Wi-Fi measurements is done with a particle filter, which uses the probabilistic measurement process models for GNSS and Wi-Fi as inputs. The advantage of the particle filter is its ability to work with non-linear dynamical models and non-Gaussian probability distributions. The evaluation was done with a modular ARM-processor based hardware platform (miniLOK) with Android as operating system, which provides an interface to the raw data of the GNSS receiver. The algorithms were implemented in Java on top of the awiloc® software framework, which is a positioning system including Wi-Fi fingerprinting developed at Fraunhofer IIS. The paper starts with the construction of measurement process models for GNSS and Wi-Fi and explains how these models are integrated into the particle filter framework. The experimental results are presented together with a description of the setup and equipment. The achieved positioning uncertainty in urban environments is discussed.