Metraipolis › archive › dc_242bbc8767c3

SafePG: Safe and Globally Optimal Reinforcement Learning with Hard Constraints — arXiv:2609.14076

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.14076
License (as archived)
CC0-1.0
Archived
2026-10-16T02:55:49.519Z
SHA-256 of the archived text
1c26b9973079bd4b432d2f756b78403038e14527d791f2a17d496b2a951fd79e
Stable URL of this page
https://metraipolis.com/documento/dc_242bbc8767c3

How to cite

SafePG: Safe and Globally Optimal Reinforcement Learning with Hard Constraints — arXiv:2609.14076 — arXiv (en), https://arxiv.org/abs/2609.14076 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_242bbc8767c3 (Metraipolis archive dc_242bbc8767c3, sha256 1c26b9973079bd4b432d2f756b78403038e14527d791f2a17d496b2a951fd79e)
BibTeX
@misc{dc242bbc8767c3,
  title = {SafePG: Safe and Globally Optimal Reinforcement Learning with Hard Constraints — arXiv:2609.14076},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.14076},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_242bbc8767c3; sha256 1c26b9973079bd4b432d2f756b78403038e14527d791f2a17d496b2a951fd79e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.14076. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We present an optimal and convergent model-free policy gradient (PG) reinforcement learning (RL) framework for controlling nonlinear dynamical systems under hard safety constraints. We first construct a class of stochastic wrapper policies centered around a deterministic controller, thereby enabling exploration in unknown environments while preserving the underlying deterministic control structure. We then define a class of parameterized safe-by-construction control policies by truncating these stochastic policies onto hard safety constraints. We next establish, via measure-theoretic arguments, that the potentially nonconvex RL objective under the truncated policy class, as well as its policy gradients, are well-defined. We then develop a model-free PG algorithm based on stochastic gradient ascent that directly searches over these truncated policies and leverage gradient dominance to establish convergence and optimality guarantees. Finally, we validate this framework through simulations on a safe quadrotor navigation problem.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-15 in eess.SY).
Categorie arXiv: Systems and Control (eess.SY).
Pagina del lavoro: https://arxiv.org/abs/2609.14076 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T02:45:08.529Z dalla pagina https://arxiv.org/catchup/eess.SY/2026-09-15?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).