Maximilian Kunz

"High-Level Semantic Localization for Human-like Navigation"

Localization is an important task for every autonomous agent in a changing and unpredictable environment. Common solutions utilize Global Navigation Satellite Systems(GNSS) which return the correct global coordinate of the robot. Such a system is not without fault though. Small errors can occur due to weather effects and occlusion of the sky. Bigger errors and total failure can happen if the view to the sky is blocked which is the case when navigating close to buildings, in the forest or inside. When facing such problems, it is worth looking at how nature fixes those problems. We as humans do not have a navigation system inside, but we can find specific locations very well even with very limited, inaccurate high-level descriptions. In addition to that, we heavily rely on our knowledge of the word and our internal map of places we have already visited. In my research I want to focus on more human-like navigation which includes semantic information on an abstract level for robot localization.

Termin als Datei downloaden und in den Kalender importieren.