Supporting Health and Community Care with Multi-Agent Systems

2. Background

While a number of definitions have been proposed for identifying a “software agent” and a “multi-agent system” (see, e.g., Russell & Norvig, 2003; Wooldridge & Jennings, 1995; Genesereth & Ketchpel, 1994), there is not a single one which researchers generally agree on. Nevertheless, there is a common understanding that an agent is essentially an autonomous software entity that should at least be designed to operate continuously in dynamic and uncertain environments, reacting to events while showing an intelligent behaviour to pursue its own objectives. An agent usually provides interoperable interfaces for interacting with other agents, either concurrently or cooperatively, exchanging messages formulated according to some syntax, semantics and pragmatics. Since an agent behaves proactively, it requires some degree of trust by its user, and it can receive delegations from either human users or other agents in the form of required actions or desired goals, matched with permissions to access necessary resources. Additionally, some agents may also be able to perform complex reasoning at run-time and also learn and change their behaviour over time, to improve their performances. Mobile agents are even able to move for one computational node to another, to follow their own users or to exploit some local resource more efficiently. Agent-based systems are often realized by loosely coupling various agents, i.e. autonomous software entities, thus modelling a proper multi-agent system, characterised by a higher level of modularity and a richer descriptive model, if compared with a solitary agent working within its environment – either with the presence of users or not.