Situated Bayesian Reasoning Framework for Robots Operating in Diverse Everyday Environments

General-purpose robots operating in unstructured environments have the potential to benefit by leveraging abstract, commonsense knowledge for task execution. In this paper, we present an approach for automatically generating a compact semantic knowledge base, relevant to a robot’s particular operating environment, given only a small number of object labels obtained from object recognition or a robot’s task description. In order to cope with noise and non-deterministic data across our data sources, we formulate our representation as a statistical relational model represented as a Baysian Logic Network. We validate our approach in both abstract and real-world domains, demonstrating the robot’s ability to perform inference about object categories, locations and properties given a small amount of local information. Additionally, we present an approach for interactively validating the mined information with the help of a co-located user.