DEPLOY-RL: Active Boundary Discovery and Conservative Certification for Deployable Reinforcement Learning in Safety-Critical Continuous Processes

Jang, Yeojin; Baek, Minu; Gil, Gihun; Jung, Minsung; Park, Beomdo; Kwon, Woohyeon; Jang, Hyeonseok; Park, Junseong; Sengar, Neha; Saurez, Andres; Lee, Sangkeum

doi:10.24963/ijcai.2026/718

DEPLOY-RL: Active Boundary Discovery and Conservative Certification for Deployable Reinforcement Learning in Safety-Critical Continuous Processes

Yeojin Jang, Minu Baek, Gihun Gil, Minsung Jung, Beomdo Park, Woohyeon Kwon, Hyeonseok Jang, Junseong Park, Neha Sengar, Andres Saurez, Sangkeum Lee

Download BibTeX Cite

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence. AI4Tech: AI Enabling Technologies. Pages 6454-6461.

https://doi.org/10.24963/ijcai.2026/718

Keywords

Advanced AI4Tech: Data-driven AI4Tech
Advanced AI4Tech: AI4Tech foundations
Domain-specific AI4Tech: AI4Safety
Domain-specific AI4Tech: AI4Manufacturing

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.

Yeojin Jang, Minu Baek, Gihun Gil, Minsung Jung, Beomdo Park, Woohyeon Kwon, Hyeonseok Jang, Junseong Park, Neha Sengar, Andres Saurez, and Sangkeum Lee. 2026. DEPLOY-RL: Active Boundary Discovery and Conservative Certification for Deployable Reinforcement Learning in Safety-Critical Continuous Processes. In Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26. IJCAI, 6454–6461. https://doi.org/10.24963/ijcai.2026/718

@inproceedings{ijcai2026p718,
  title     = {DEPLOY-RL: Active Boundary Discovery and Conservative Certification for Deployable Reinforcement Learning in Safety-Critical Continuous Processes},
  author    = {Jang, Yeojin and Baek, Minu and Gil, Gihun and Jung, Minsung and Park, Beomdo and Kwon, Woohyeon and Jang, Hyeonseok and Park, Junseong and Sengar, Neha and Saurez, Andres and Lee, Sangkeum},
  booktitle = {Proceedings of the Thirty-Fifth International Joint Conference on
               Artificial Intelligence, {IJCAI-26}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {Diego Calvanese},
  pages     = {6454--6461},
  year      = {2026},
  month     = {8},
  note      = {AI4Tech: AI Enabling Technologies},
  doi       = {10.24963/ijcai.2026/718},
  url       = {https://doi.org/10.24963/ijcai.2026/718},
}

TY  - CPAPER
TI  - DEPLOY-RL: Active Boundary Discovery and Conservative Certification for Deployable Reinforcement Learning in Safety-Critical Continuous Processes
AU  - Jang, Yeojin
AU  - Baek, Minu
AU  - Gil, Gihun
AU  - Jung, Minsung
AU  - Park, Beomdo
AU  - Kwon, Woohyeon
AU  - Jang, Hyeonseok
AU  - Park, Junseong
AU  - Sengar, Neha
AU  - Saurez, Andres
AU  - Lee, Sangkeum
T2  - Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2026
SP  - 6454
EP  - 6461
DO  - 10.24963/ijcai.2026/718
UR  - https://doi.org/10.24963/ijcai.2026/718
ER  -