Knowledge model quantitative evaluation for adaptive world modeling

Abstract

World modeling can provide environment information to applications for decision support and situation assessment. In a semantic world model like the Object-Oriented World Model (OOWM), knowledge about an application domain is modeled a priori. In practice, however, world modeling systems have to deal with an open world, where unforeseen real-world entities can occur during operations. To enable open-world modeling for the OOWM, an approach to adaptive knowledge management is presented. This approach proposes an information-theoretic model evaluation based on the Minimum Description Length principle.