As robots become more common, it becomes increasingly useful for them to communicate and effectively share knowledge that they have learned through their individual experiences. Learning from experiences, however, is oftentimes embodiment-specific; that is, the knowledge learned is grounded in the robot’s unique sensors and actuators. This type of learning raises questions as to how communication and knowledge exchange via social interaction can occur, as properties of the world can be grounded differently in different robots. This is especially true when the robots are heterogeneous, with different sensors and perceptual features used to define the properties. In this paper, we present methods and representations that allow heterogeneous robots to learn grounded property representations, such as that of color categories, and then build models of their similarities and differences in order to map their respective representations. We use a conceptual space representation, where object properties are learned and represented as regions in a metric space, implemented via supervised learning of Gaussian Mixture Models. We then propose to use confusion matrices that are built using instances from each robot, obtained in a shared context, in order to learn mappings between the properties of each robot. Results are demonstrated using two perceptually heterogeneous Pioneer robots, one with a web camera and another with a camcorder.

Abstract. We introduce a framework for practical, essentially empirical exploitation of emergent, primarily automatically extracted knowledge. Efficient and meaningful machine processing of knowledge extracted, e.g., by ontology learning from natural language texts, has been largely an open problem to date. To address this gap, we propose a light-weight, similarity-based knowledge representation framework and respective simple, yet quite practical inference services. Our approach has been motivated by and applied to a life science use case. 1

Abstract. In this report, we present a solution for robust scalable extraction and exploitation of knowledge from unstructured text. The robustness is achieved by an application of our novel light-weight, similarity-based knowledge representation framework and respective inference services. The scalability is ensured by the framework’s straightforward anytime implementation on the top of a relational database back-end. The potential of our work is exemplified within an oncological literature search scenario, which was motivated by and evaluated with domain experts. 1

Abstract. Behaviour recognition in a video scene consists of several distinct sub-tasks: objects or object parts must be recognised, classified and tracked, qualitative spatial and temporal properties must be determined, behaviour of individual objects must be identified, and composite behaviours must be determined to obtain an interpretation of the scene as a whole. In this paper, we describe how these tasks can be distributed over three processing stages (low-level analysis, middle layer mediation and high-level interpretation) to obtain flexible and efficient bottom-up and top-down processing. The approach is implemented in the system SCENIC and currently applied to two domains: dynamic indoor scenes and static building scenes. We include details of an experiment where an ongoing table-laying scene is recognised. 1

Abstract — This paper explores methods and representations that allow two perceptually heterogeneous robots, each of which represents concepts via grounded properties, to transfer knowledge despite their differences. This is an important issue, as it will be increasingly important for robots to communicate and effectively share knowledge to speed up learning as they become more ubiquitous. We use Gärdenfors ’ conceptual spaces to represent objects as a fuzzy combination of properties such as color and texture, where properties themselves are represented as Gaussian Mixture Models in a metric space. We then use confusion matrices that are built using instances from each robot, obtained in a shared context, in order to learn mappings between the properties of each robot. These mappings are then used to transfer a concept from one robot to another, where the receiving robot was not previously trained on instances of the objects. We show in a 3D simulation environment that these models can be successfully learned and concepts can be transferred between a ground robot and an aerial quadrotor robot. I.

The complex analyses required by artificial intelligence applications need both a flexible structure for information representation and a quick and e#cient method for information categorization. The latter has a great impact on the former because it can increase or not the knowledge database dimension. This situation can sometimes lead to undesired complexity. In this paper a genetic algorithm approach was used in order to obtain two major desired e#ects. The first one is to increase the disjunction level in knowledge representation and the second is to increase the processing speed. Comparisons were made between our solutions and the weighting results of ReliefF algorithm.