Attention-driven artificial agents

Abstract

In many (if not all) of the domains that are related with machines’ interaction with humans the human model is considered as the ideal prototype. Adaptation of the behaviour of an application as a function of its current environment known as context awareness is clearly one of these domains. The environment can be characterized as a physical location, an orientation or a user profile. A context-aware application can sense the environment and interpret the events that occur within. In this paper we present an attention-based model, inspired from the human brain, for constructing artificial agents. In this model adaptation is achieved through focusing to irregular patterns, so as to identify possible context switches, and adapting the behaviour goals accordingly. Simulation results, obtained using a health-monitoring scenario, are presented showing the efficiency of the proposed model.