Intelligent association selection of embedded agents in intelligent inhabited environments

Discovering frequent user--environment interactions in intelligent environments

“…So far, the most common approach relates the actions of the user (gathered by O-type sensors) with the status of the environment (gathered by C-type sensors) (Gal et al, 2001) (Hagras et al, 2004). We consider, as have other authors (Duman et al, 2008) (Jakkula et al, 2007), that detecting relationships between the user's actions facilitates understanding of the user's behavior, using information about the environment to contextualize such a relationship. Besides facilitating the understanding of the user's behavior, finding these relationships also allows us to use relative time references instead of absolute times.…”

“…Recording the changes caused by the user in the environment, they generated membership functions as well as fuzzy rules that mapped data into fuzzy rules. Due to the excessive number of rules they generated, they introduced an improvement (Duman et al, 2008) where they identified relevant and important associations between given actions, so that irrelevant aspects of the rules (and, by extension, some rules as well) were removed.…”

A Fuzzy Based Architecture for Learning Relevant Embedded Agents Associations in Ambient Intelligent Environments

“…The Intelligent Association System (IAS) defines the architectural framework for the intelligent association exploration method as depicted in figure 1 [5]. The IAS framework combines all devices and their services to a decentralized service-oriented overlay network architecture.…”

“…The F-IAS agents uses the structural notions and descriptions of Fuzzy Cognitive Maps (FCMs) [5] allowing causal evaluation among the associated services. The most significant characteristics of FCMs that comply with the F-IAS agents are: * The FCMs structure allows the F-IAS agents to dynamically reorganize themselves, e.g.…”

“…* The association among the F-IAS agents and services is indicated with a weight value, which illustrates the strength of the causation and as a result represent the importance of the association. * An unsupervised mechanism (mostly Hebbian learning [5]) for FCMs gradually learns the association strengths after correlating the changes of two associated entities. * The overall system is event-driven and decentralized.…”