{"_id":"14830","status":"public","author":[{"id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0002-4681-1699","first_name":"Dorde","last_name":"Zikelic","full_name":"Zikelic, Dorde"},{"first_name":"Mathias","last_name":"Lechner","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87"},{"full_name":"Henzinger, Thomas A","last_name":"Henzinger","first_name":"Thomas A","orcid":"0000-0002-2985-7724","id":"40876CD8-F248-11E8-B48F-1D18A9856A87"},{"id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0002-4561-241X","first_name":"Krishnendu","full_name":"Chatterjee, Krishnendu","last_name":"Chatterjee"}],"publication":"Proceedings of the 37th AAAI Conference on Artificial Intelligence","department":[{"_id":"ToHe"},{"_id":"KrCh"}],"page":"11926-11935","acknowledgement":"This work was supported in part by the ERC-2020-AdG 101020093, ERC CoG 863818 (FoRM-SMArt) and the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie Grant Agreement No. 665385.","volume":37,"publication_identifier":{"issn":["2159-5399"],"eissn":["2374-3468"]},"year":"2023","language":[{"iso":"eng"}],"issue":"10","conference":{"end_date":"2023-02-14","start_date":"2023-02-07","location":"Washington, DC, United States","name":"AAAI: Conference on Artificial Intelligence"},"date_created":"2024-01-18T07:44:31Z","date_updated":"2024-01-22T14:08:29Z","project":[{"grant_number":"101020093","name":"Vigilant Algorithmic Monitoring of Software","call_identifier":"H2020","_id":"62781420-2b32-11ec-9570-8d9b63373d4d"},{"name":"Formal Methods for Stochastic Models: Algorithms and Applications","call_identifier":"H2020","_id":"0599E47C-7A3F-11EA-A408-12923DDC885E","grant_number":"863818"},{"_id":"2564DBCA-B435-11E9-9278-68D0E5697425","call_identifier":"H2020","name":"International IST Doctoral Program","grant_number":"665385"}],"publication_status":"published","month":"06","oa_version":"Preprint","keyword":["General Medicine"],"abstract":[{"text":"We study the problem of learning controllers for discrete-time non-linear stochastic dynamical systems with formal reach-avoid guarantees. This work presents the first method for providing formal reach-avoid guarantees, which combine and generalize stability and safety guarantees, with a tolerable probability threshold p in [0,1] over the infinite time horizon. Our method leverages advances in machine learning literature and it represents formal certificates as neural networks. In particular, we learn a certificate in the form of a reach-avoid supermartingale (RASM), a novel notion that we introduce in this work. Our RASMs provide reachability and avoidance guarantees by imposing constraints on what can be viewed as a stochastic extension of level sets of Lyapunov functions for deterministic systems. Our approach solves several important problems -- it can be used to learn a control policy from scratch, to verify a reach-avoid specification for a fixed control policy, or to fine-tune a pre-trained policy if it does not satisfy the reach-avoid specification. We validate our approach on 3 stochastic non-linear reinforcement learning tasks.","lang":"eng"}],"publisher":"Association for the Advancement of Artificial Intelligence","type":"conference","related_material":{"record":[{"status":"public","id":"14600","relation":"earlier_version"}]},"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","day":"26","title":"Learning control policies for stochastic systems with reach-avoid guarantees","intvolume":"        37","date_published":"2023-06-26T00:00:00Z","external_id":{"arxiv":["2210.05308"]},"citation":{"mla":"Zikelic, Dorde, et al. “Learning Control Policies for Stochastic Systems with Reach-Avoid Guarantees.” <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, vol. 37, no. 10, Association for the Advancement of Artificial Intelligence, 2023, pp. 11926–35, doi:<a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">10.1609/aaai.v37i10.26407</a>.","ista":"Zikelic D, Lechner M, Henzinger TA, Chatterjee K. 2023. Learning control policies for stochastic systems with reach-avoid guarantees. Proceedings of the 37th AAAI Conference on Artificial Intelligence. AAAI: Conference on Artificial Intelligence vol. 37, 11926–11935.","short":"D. Zikelic, M. Lechner, T.A. Henzinger, K. Chatterjee, in:, Proceedings of the 37th AAAI Conference on Artificial Intelligence, Association for the Advancement of Artificial Intelligence, 2023, pp. 11926–11935.","chicago":"Zikelic, Dorde, Mathias Lechner, Thomas A Henzinger, and Krishnendu Chatterjee. “Learning Control Policies for Stochastic Systems with Reach-Avoid Guarantees.” In <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, 37:11926–35. Association for the Advancement of Artificial Intelligence, 2023. <a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">https://doi.org/10.1609/aaai.v37i10.26407</a>.","ieee":"D. Zikelic, M. Lechner, T. A. Henzinger, and K. Chatterjee, “Learning control policies for stochastic systems with reach-avoid guarantees,” in <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, Washington, DC, United States, 2023, vol. 37, no. 10, pp. 11926–11935.","ama":"Zikelic D, Lechner M, Henzinger TA, Chatterjee K. Learning control policies for stochastic systems with reach-avoid guarantees. In: <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>. Vol 37. Association for the Advancement of Artificial Intelligence; 2023:11926-11935. doi:<a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">10.1609/aaai.v37i10.26407</a>","apa":"Zikelic, D., Lechner, M., Henzinger, T. A., &#38; Chatterjee, K. (2023). Learning control policies for stochastic systems with reach-avoid guarantees. In <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i> (Vol. 37, pp. 11926–11935). Washington, DC, United States: Association for the Advancement of Artificial Intelligence. <a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">https://doi.org/10.1609/aaai.v37i10.26407</a>"},"doi":"10.1609/aaai.v37i10.26407","ec_funded":1,"article_processing_charge":"No","quality_controlled":"1"}