Localization Using Magnetic Field Distortions

Current localization algorithms are prone to failure in demanding environments due to restrictions imposed by the commonly applied sensor techniques. For example, GPS-based localization fails in indoor environments or under water, vision-based sensors fail in difficult lightning conditions, etc. The Earth's magnetic field is up to now mainly used for heading estimation (compass), but has significant local properties that can be exploited to estimate not only the orientation, but the full pose including position.

The talk will give a brief introduction to magnetic field distortions in different environments, how to deal with modelling inherent system disturbances of mobile robots and introduce the developed solution to handle magnetic field properties in a probabilistic way and how to integrate them in a 6 DOF localization algorithm. I will also present ongoing work related to the Ph.D. topic using distributed magnetometers to further enhance the localization solution.