The big strength of intelligent agents - their communicative abilities - can often change into a big hindrance in real-world applications of collaborative problem solving. One can often observe that communication overhead annihilates the merits of partitioning the given problem. In this article, I introduce a concept called dynamic reconfiguration that allows for adapting the multi agent system structure to the given problem structure thus saving communication costs. The main focus lies on a novel consensus protocol that supports this reconfiguration and assumes benevolent, trustworthy agents. Hence, it is like the whole reconfiguration approach itself mainly applicable for rather closed systems, e. g. information or control systems. The protocol claims to enhance the efficiency of consensus finding by using several different phases of negotiation.
展开▼