Résumé: An Embedded multi-agent system (Embedded MAS) is an embedded cognitive system based on agents cooperating to control hardware devices. These agents are autonomous and proactive entities capable of decision-making and can constantly acquire new knowledge via interaction with other agents and the environment. Since the interaction between agents is relevant for acquiring new knowledge, issues such as the communicability and mobility of agents from different Embedded MAS must be highlighted. The classification of a MAS as Open or Closed only considers the mobility of agents, but communicability also needs to be considered. For this, we extend the notion of openness in these systems to consider the existence of Totally Closed and Limited Open MAS, to consider agents from an Embedded MAS without the ability to move or communicate or when they lose the ability to communicate but still can move to other systems. In cooperative missions where several devices adopt Embedded MAS, they should not become totally closed since they lose the ability to cooperate and could put the mission at risk. Some existent works considering Embedded MAS relies upon IoT infrastructures to guarantee communicability and mobility. But, in cases where these infrastructures are temporarily or permanently unavailable, the system becomes totally closed. Even when alternatives exist, they do not use cryptography. Therefore, we present a middleware for supporting the development of Embedded MAS, considering radiofrequency ad-hoc communication to reduce the dependency on centralized infrastructures. An extended protocol supports message exchange between devices using cryptography. We also present a proof of concept application and a formalization of our model.

Langue: Anglais