An action language for multi-agent domains (Q2060726)

This paper presents \(\mathrm{m}\mathcal{A}^*\), an action language for reasoning about actions, agents' knowledge and agents' beliefs in a multi-agent setting. The language includes actions that cause the world to be changed, sensing actions, and actions by which agents can announce states of facts to other agents (called ``announcement'' actions). The language also deals with the issue of observability, that is, circumstances in which agents get full knowledge about actions being performed and their effects, those in which this knowledge is only partial or null. The authors introduce the syntax of the language and a Kripke-structures-based semantics of it (pointed Kripke structures are used to represent the possible states of the world and of agents' knowledge and beliefs). As for the dynamic component of the semantics, a transition function is associated to actions so that, given a state, a set of resulting states is returned. The paper then proves several interesting properties pertaining the thus introduced formal setting with a particular enphasis on practical employability of the language. The paper is well equipped with many clarifying examples which help reader's intuition about the formal framework as it is stepwisely introduced. Finally, it is worth mentioning that the paper also provides an in-depth, ample comparative discussion of relevant related work. So, generally speaking, the paper presents a very interesting proposal in the context of languages designed for reasoning about actions and their effects on world states, agents' knowledge and beliefs and, in my opinion, it is a ``must-read'' for anyone working in this field.