Knowledge Modelling in Two-Level Decision Making for Robot Navigation

In recent years, social robotics has become a popular research field. It aims to develop robots capable of communicating and interacting with humans in a personal and natural way. Social robots have the objective to provide assistance as a human would do it. Social robotics is a multidisciplinary field that brings together different areas of science and engineering, such as robotics, artificial intelligence, psychology and mechanics, among others (Breazeal, 2004). In this sense, an interdisciplinary group of the University of Almeria is developing a social robot based on the Peoplebot platform (ActivMedia Robotics, 2003). It has been specifically designed and equipped for human-robot interaction. For that purpose, it includes all the basic components of sensorization and navigation for real environments. The ultimate goal is that this robot acts as a guide for visitors at our university (Chella et al., 2007). Since the robot can move on indoor/outdoor environments, we have designed and implemented a two-level decision making framework to decide the most appropriate localization strategy. Knowledge modelling is a process of creating a model of knowledge or standard specifications about a kind of process or product. The resulting knowledge model must be interpretable by the computer; therefore, it must be expressed in some knowledge representation language or data structure that enables the knowledge to be interpreted by software and to be stored in a database or data exchange file. CommonKADS is a comprehensive methodology that covers the complete route from corporate knowledge management to knowledge analysis and engineering, all the way to knowledge-intensive systems design and implementation, in an integrated fashion (Schreiber et al., 1999). There are several studies on the knowledge representation and modelling for robotic systems. In some cases, semantic maps are used to add knowledge to the physical maps. These semantic maps integrate hierarchical spatial information and semantic knowledge that is used for robot task planning. Task planning is improved in two ways: extending the capabilities of the planner by reasoning about semantic information, and improving the planning efficiency in large domains (Galindo et al., 2008). Other studies use the CommonKADS methodology, or any of its extensions, to model the knowledge; some of the CommonKADS extensions that have been used in robotics are CommonKADS-RT, for real time systems, and CoMoMAS (Conceptual Modelling of Multi-Agent Systems), for multi-