Agreeing on plans through iterated disputes

Autonomous agents transcend their individual capabilities by cooperating towards achieving shared goals. The different viewpoints agents have on the environment cause disagreements about the anticipated effects of plans. Reaching agreement requires the resolution of such inconsistencies and the alignment of the agents' viewpoints. We present a dialogue protocol that enables agents to discuss candidate plans and reach agreements. The dialogue is based on an argumentation process in the language of situation calculus. Agreement is reached through persuasion, thereby aligning the planning beliefs of the agents. We describe our abstract iterated dialogue protocol, and extend it for the specific problem of arguing about plans. We show that our method always terminates and produces sound results. Furthermore, we detail a set of extensions to simplify reasoning and reduce the exchanged information.