Prior knowledge employment based on the K-L and Tanimoto distances matching for intelligent autonomous robots

Abstract

Modern autonomous robots are performing complex tasks in a real dynamic environment. This requires real-time reactive and proactive handling of arising situations. A basis for such situation awareness and handling can be a world modeling subsystem that acquires information from sensors, fuses it into existing world description and delivers the required information to all other robot subsystems. Since sensory information is affected by uncertainty and lacks for semantic meaning, the employment of a predefined information, that contains concepts and descriptions of the surrounding world, is crucial. This employment implies matching of the world model information to prior knowledge and subsequent complementing of the dynamic descriptions with semantic meaning and missing attributes. The following contribution describes a matching mechanism based on the Kullback-Leibler and Tanimoto distances and direct assignment of the prior knowledge for the model complementation. © Springer-Verlag Berlin Heidelberg 2012.