{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,15]],"date-time":"2026-05-15T21:09:37Z","timestamp":1778879377311,"version":"3.51.4"},"reference-count":0,"publisher":"AI Access Foundation","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["jair"],"abstract":"<jats:p>Background: Multi-Agent Planning (MAP) involves coordinating the actions of multiple autonomous agents to achieve shared objectives. A prevalent formalism for MAP is the Multi-Agent Planning Domain Definition Language (MA-PDDL). While effective, existing MA-PDDL solvers typically require complete access to agents\u2019 action models\u2014specifically their preconditions and effects. However, manually creating these models is often intractable, requiring exhaustive domain expertise.\nObjectives: This work explores an alternative approach: automatically learning agents\u2019 action models from observed transitions. Since learned models may be inaccurate, planning with them can yield invalid or non-executable sequences. To mitigate this, we formalize a requirement for safety, ensuring that plans generated via the learned model remain sound with respect to the real unknown action model.\nMethods: Previous research introduced the Safe Action Model Learning (SAM) algorithm for single-agent domains. However, SAM is not suitable for MA-PDDL environments where observations include concurrently executed actions, since it cannot naturally disambiguate the individual contributions of the agents to the observed effects. To address this, we introduce Multi-Agent Safe Action Model Learning (MA-SAM), a safe action model learning algorithm designed to handle concurrent multi-agent observations. For scenarios where individual action effects remain ambiguous, we further propose MA-SAM+ , which learns the preconditions and effects of macro-actions representing concurrent execution of subsets of actions. We evaluate both algorithms on domains from the Competition of Distributed and Multi-Agent Planners (CoDMAP) benchmarks and a novel MAP domain inspired by the game Overcooked.\nResults: We establish a theoretical lower bound on the sample complexity for learning safe action models in multi-agent settings. We prove that MA-SAM does not achieve this lower bound in all cases, identifying specific conditions under which its sample complexity may become unbounded. Empirically, both MA-SAM and MA-SAM+ significantly outperform SAM-based baselines in coverage and applicability rates. While their performance is comparable in many settings, MA-SAM+ highly outperforms MA-SAM in some of the evaluated domains.\nConclusions: We present the first algorithms capable of learning safe MA-PDDL action models from concurrently executed actions, providing both theoretical foundations and empirical validation across diverse planning benchmarks<\/jats:p>","DOI":"10.1613\/jair.1.20494","type":"journal-article","created":{"date-parts":[[2026,5,15]],"date-time":"2026-05-15T20:17:43Z","timestamp":1778876263000},"source":"Crossref","is-referenced-by-count":0,"title":["Safe Learning of Multi-Agent Action Models from Concurrent Joint Action Observations"],"prefix":"10.1613","volume":"86","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-3502-1461","authenticated-orcid":false,"given":"Argaman","family":"Mordoch","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ori","family":"Karat","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Lea","family":"Shmilovich","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0008-7427-6380","authenticated-orcid":false,"given":"Yarin","family":"Benyamin","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6542-833X","authenticated-orcid":false,"given":"Brendan","family":"Juba","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0043-8179","authenticated-orcid":false,"given":"Roni","family":"Stern","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"16860","published-online":{"date-parts":[[2026,5,15]]},"container-title":["Journal of Artificial Intelligence Research"],"original-title":[],"link":[{"URL":"https:\/\/www.jair.org\/index.php\/jair\/article\/download\/20494\/27304","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.jair.org\/index.php\/jair\/article\/download\/20494\/27304","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,15]],"date-time":"2026-05-15T20:17:43Z","timestamp":1778876263000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.jair.org\/index.php\/jair\/article\/view\/20494"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,15]]},"references-count":0,"URL":"https:\/\/doi.org\/10.1613\/jair.1.20494","relation":{},"ISSN":["1076-9757"],"issn-type":[{"value":"1076-9757","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,15]]}}}