Addressing Constraint Failures in Distributed Dialogue Protocols

Early agent communication languages lacked the means to coordinate the interaction between agents. The importance of coordination was better appreciated when open systems became in favor. Some solutions have been proposed; however, these normally rely on centralized gatekeepers. Distributed Dialogues, on the other hand, provide both communication and coordination protocols for multi-agent systems. They do that without the need to re-program agents every time they need to adapt to a new dialogue system and without the need for centralized gatekeepers for coordination. However, these dialogue protocols still aren’t tolerant to failure. When an agent does not respond appropriately, the dialogue simply stops. This usually happens when constraints are broken. This dissertation addresses these kinds of failure by offering ‘induced backtracking’ to explore other parts of the dialogue. It also proposes a negotiation protocol that could be used when some failed constraints are better dealt with by negotiation