CSAIL Event Calendar: Previous Series
Semantic Maps for Mobile Robots - Perception, Representation, Reasoning and Actions
Speaker: Andrzej Pronobis , Royal Institute of Technology (KTH)
Robots have finally escaped from industrial workplaces and are making their way into our homes, offices and public spaces. In order to realize the dream of robot assistants performing tasks together with humans, we need to provide them with the capability of understanding complex, unstructured environments and express their knowledge in terms of human concepts. In this talk we present our work on endowing robot systems with the ability to acquire and exploit the semantics and structure that is present in man made environments. This provides the means for more efficient task execution and also plays a central role in enabling intuitive HRI and knowledge transfer between humans and robots. The latter opens up for making use of the vast amount of human spatial knowledge encoded in human readable form. More specifically, in this talk we present our work on a framework for large-scale, probabilistic semantic mapping, which exploits multiple, heterogenous sources of semantic information, including place appearance and geometry, objects, topology of space and human asserted knowledge. Furthermore, we show how semantic knowledge can be utilized for solving real-world problems, on the example of large-scale object search task, leading to improved performance and more intuitive and human-like behavior.