Diploma Thesis University of Bremen

08.10.2013

Carsten Könemann

In the field of robotics, one of the most frequently performed tasks is the search for specific objects. It is an integral part of almost everything a robot does, because it first needs to know where things are in order to work with them. Current solutions either require the robot to search for objects as needed or utilize a lookup-table of objects and their locations. The former solution is very flexible but time-consuming, especially in larger environments. The latter is extremely fast, but also inflexible and high-maintenance. In this work, a new method is developed, that combines the advantages of both types of existing solutions: A system that continuesly looks out for objects that might become interesting in the future and dynamically creates a semantic map of its environment. Then, if a robot needs something it came across earlier, it already knows where to go.