{"type":"ScholarlyArticle","title":"DEPLOY-RL: Active Boundary Discovery and Conservative Certification for Deployable Reinforcement Learning in Safety-Critical Continuous Processes","authors":[{"name":"Yeojin Jang","slug":"yeojin-jang","given":"Yeojin","family":"Jang","affiliation":""},{"name":"Minu Baek","slug":"minu-baek","given":"Minu","family":"Baek","affiliation":""},{"name":"Gihun Gil","slug":"gihun-gil","given":"Gihun","family":"Gil","affiliation":""},{"name":"Minsung Jung","slug":"minsung-jung","given":"Minsung","family":"Jung","affiliation":""},{"name":"Beomdo Park","slug":"beomdo-park","given":"Beomdo","family":"Park","affiliation":""},{"name":"Woohyeon Kwon","slug":"woohyeon-kwon","given":"Woohyeon","family":"Kwon","affiliation":""},{"name":"Hyeonseok Jang","slug":"hyeonseok-jang","given":"Hyeonseok","family":"Jang","affiliation":""},{"name":"Junseong Park","slug":"junseong-park","given":"Junseong","family":"Park","affiliation":""},{"name":"Neha Sengar","slug":"neha-sengar","given":"Neha","family":"Sengar","affiliation":""},{"name":"Andres Saurez","slug":"andres-saurez","given":"Andres","family":"Saurez","affiliation":""},{"name":"Sangkeum Lee","slug":"sangkeum-lee","given":"Sangkeum","family":"Lee","affiliation":""}],"abstract":"Reinforcement learning (RL) policies often outperform classical controllers in simulation, yet rarely reach production in safety-critical processes. The barrier is that there is no principled way to answer “Is this policy safe to deploy?” with statistical guarantees. We introduce DEPLOY-RL, a post-training certification framework built on one key insight: deployment certification requires discovering where failures occur (boundary discovery), not measuring how much everywhere (uniform reconstruction). Our contributions: (1) a contract-coupled acquisition function that concentrates sampling on certification-critical boundaries, achieving ≈ 2× sample efficiency with a semi-empirical ambiguity reduction bound (domain-calibrated convergence guarantee); (2) conformal risk control providing finite-sample false-go guarantees (≤ α) under explicit deployment contracts; (3) a three-way decision framework (Deploy/No-Deploy/Abstain) with fail-safe PID/MPC fallback. In simulations on papermaking (industrial digital twin) and Tennessee Eastman (public benchmark), DEPLOY-RL achieves 4.4% false-go rate (vs. 8.6% for the best baseline) while retaining 88.6% policy coverage—the only method achieving <5% false-go with >85% coverage among 14 baselines under our evaluation protocol.","doi":"10.24963/ijcai.2026/718","url":"https://ijcai.org/proceedings/2026/718","pdf":"https://ijcai.org/proceedings/2026/0718.pdf","markdown":"https://ijcai.org/proceedings/2026/718.md","bibtex":"https://ijcai.org/proceedings/2026/bibtex/718","volume":{"slug":"2026","title":"Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2026","url":"https://ijcai.org/proceedings/2026/","year":2026,"location":"Bremen, Germany","dates":"15-21 August 2026","editors":"Diego Calvanese","isbn":"978-1-956792-09-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2026"},"section":"AI4Tech: AI Enabling Technologies","track":"AI4Tech: AI Enabling Technologies","number":718,"pages":{"first":6454,"last":6461},"keywords":["Advanced AI4Tech: Data-driven AI4Tech","Advanced AI4Tech: AI4Tech foundations","Domain-specific AI4Tech: AI4Safety","Domain-specific AI4Tech: AI4Manufacturing"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}