A general multi-agent epistemic planner based on higher-order belief change

• Published in: Artificial intelligence 2021-12, Vol.301, p.103562, Article 103562
• Main Authors: Wan, Hai, Fang, Biqing, Liu, Yongmei
• Format: Article
• Language: English
• Subjects: Algorithms; Belief change; Canonical forms; Community planning; Epistemic planning; Epistemology; Knowledge bases (artificial intelligence); Multi-agent epistemic logic; Multiagent systems; Planning; Reasoning; Revisions
• ISSN: 0004-3702; 1872-7921
• DOI: 10.1016/j.artint.2021.103562

In recent years, multi-agent epistemic planning has received attention from both dynamic logic and planning communities. Existing implementations of multi-agent epistemic planning are based on compilation into classical planning and suffer from various limitations, such as generating only linear plans, restriction to public actions, and incapability to handle disjunctive beliefs. In this paper, we consider centralized multi-agent epistemic planning from the viewpoint of a third person who coordinates all the agents to achieve the goal. We treat contingent planning, resulting in nonlinear plans. We model private actions and hence handle beliefs, formalized with the multi-agent KD45 logic. We handle static propositional common knowledge, which we call constraints. For such planning settings, we propose a general representation framework where the initial knowledge base (KB) and the goal, the preconditions and effects of actions can be arbitrary KD45n formulas, and the solution is an action tree branching on sensing results. In this framework, the progression of KBs w.r.t. actions is achieved through the operation of belief revision or update on KD45n formulas, that is, higher-order belief revision or update. To support efficient reasoning and progression, we make use of a normal form for KD45n called alternating cover disjunctive formulas (ACDFs). We propose reasoning, revision and update algorithms for ACDFs. Based on these algorithms, adapting the PrAO algorithm for contingent planning from the literature, we implemented a multi-agent epistemic planner called MEPK. Our experimental results show the viability of our approach.