An integrated active perception module for a distributed cognitive architecture

This paper presents an active perception planning module embedded in a distributed functional cognitive architecture for complex environments. It discusses the functional module integration over the wish list concept, enabling distributed planning, reasoning and decision actions. Further the perception planning approach is depicted along with its components: the probabilistic framework for scene modeling, the correct association of observation data into the scene model, the probabilistic computation of object occlusions and the probabilistic action planning in order to select the most profitable prospective viewpoint. In an experimental setting a complete perception loop is presented and the achieved results are discussed. Perception and manipulation tasks in an everyday kitchen environment are chosen as the evaluation scenario. It can be stated that the so far developed active perception planning approach convinces in its applicability to high dimensional state spaces, its quick computation and its efficient planning strategy.