Analysis of inertia based positioning systems
Summary
We want to create an immersive augmented reality framework. An essential part in this
framework is aligning a virtual environment with the real world. To align the virtual
environment we need to know our position and orientation in the real world as accurate as
possible. In order to obtain our position and orientation we need a positioning system. We also
want to have a smooth user experience, which means that the positioning system has to be
fast. Therefore we analyze an inertia based positioning system to see whether it meets our
requirements.
Inertia based positioning systems are subject to a cumulative integration error called
drift. Our first objectives is to analyze the drift and to find what causes this error. Then we
try to keep the drift within a certain error margin for a longer time by improving the
measurement signals; which appear to be the biggest effector on the drift. We also use
computer vision to verify the measurements. Our final objective is to analyze whether we can
use a map-based positioning system as reference, in order to meet the requirements to the
system.
The results we present show that we can improve the measurement signals using a
Kalman filter with models based on the Wiener process acceleration model. We have also
analyzed the drift over time and we concluded that using a map-based positioning system as
reference would suffice to keep the drift within considerable error margins. However, we strived
to reduce the drift using computer vision as well. Unfortunately, the hardware used in our
experiments was not capable to reduce the drift further.