Domen Zupančič: Self learning systems in Ambient Intelligence

Presentation on Self learning systems in Ambient Intelligence given by Domen Zupančič.

The presentation provides a review of the research field ambient intelligence (AmI) and the general research orientation for PhD. Slightly more attention is given to ambient assistant living and energy optimization issues. The specification and definition of intelligent environment and an appropriate sensor network, which provides data about user activities, is required. That data should enable extraction of context; in combination with space and time labels that context will be used for further reasoning and decision making by agents. The context data is a key for AmI systems. The modular communication platform should be properly prepared to enable fully exploitation of the context and cooperation among various agents. The problem about conventional building automation is explained and the suggestion for a novel approach with multi-agent system is proposed. Some research projects about intelligent environment for intelligent homes and care systems are presented. The research aim is to apply the multi-agent approach, where a particular agent is represented as an autonomous node. That node can be either a physical device (sensor, actuator) or virtual (situation reasoner, activity monitor) with appropriate ontology which is used for automated reasoning based on context and ambient conditions.