Planning object informed search for robots in household environments

In the current state-of-the-art, social robots performing non-trivial tasks often spend most of their time finding and modeling objects. In this paper we present the extension of a cognitive architecture that reduces the time and effort a robot needs to retrieve objects in a household scenario. We upgrade our previous Passive Learning Sensor algorithm into a full fledged agent that is part of the CORTEX robotics cognitive architecture. With its planning capabilities, this new configuration allows the robot to efficiently search, pick and deliver different objects from different locations in large households environments. The contribution presented here dynamically extends the robot's knowledge of the world by making use of memories from past experiences. Results obtained from several experiments show that, both, the new software agent and the integrated cognitive architecture, constitute an important step towards robot autonomy. The experiments show that the find-and-pick task is greatly accelerated.