Human activities and the environments in which they occur (real or virtual) are perceived in an increasingly sophisticated manner by artificial agents, and they can trigger complex interactions with "intelligent" software agents or "autonomous"robot. The research activities aims to study new architectures and methodologies to better manage the cycle "perception-understanding-action" of artificial autonomous systems, taking inspiration also from human cognitive models. The perception usually requires the elaboration of a large amount of raw data of the real environment (eg by artificial vision), or data originating from real and virtual sensors (soft sensors). Understanding instead requires capacity for representation and processing of knowledge at multiple levels of abstraction. Currently, hybrid architectures are the subject of study and experimentation in order to integrate frameworks of different types (sub-conceptual, conceptual and symbolic), to create conceptual structures (eg through conceptual spaces, semantic spaces, and ontologies), to create the association of these structures to symbolic constructs, and finally to allow learning, reasoning and action planning within models of the human mind. The goal of the research is to integrate both aspects of basic cognitive capabilities (understanding, learning, decision making, and communicating), and higher-level aspects such as emotions (by affect computing), creativity (by computational creativity paradigms), introspective capabilities, and motivations which strongly influence the actions in real environments. Another important goal is that concerning the social interaction of the system, through the development of language, the analysis of textual information (by social/semantic computing), and the models of human-robot interaction. The experiments deal with both autonomous software agents (eg chatbot), and humanoid robotic platforms.
Vedi anche: