Metraipolis › archive › dc_c96dc51fac13
Interleaved Projected Gradient Descent for Safe Imitation Learning — arXiv:2610.07167
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.07167
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T11:42:15.198Z
- SHA-256 of the archived text
2d8d1980b7ea1b1195b3c2d6a95de77939ca7bcddc42a2bd7b5e34f5dcb22708- Stable URL of this page
https://metraipolis.com/documento/dc_c96dc51fac13
How to cite
Interleaved Projected Gradient Descent for Safe Imitation Learning — arXiv:2610.07167 — arXiv (en), https://arxiv.org/abs/2610.07167 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c96dc51fac13 (Metraipolis archive dc_c96dc51fac13, sha256 2d8d1980b7ea1b1195b3c2d6a95de77939ca7bcddc42a2bd7b5e34f5dcb22708)
BibTeX
@misc{dcc96dc51fac13,
title = {Interleaved Projected Gradient Descent for Safe Imitation Learning — arXiv:2610.07167},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.07167},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c96dc51fac13; sha256 2d8d1980b7ea1b1195b3c2d6a95de77939ca7bcddc42a2bd7b5e34f5dcb22708},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.07167. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We propose an imitation-learning design for neural-network control policies under state and input constraints. Training alternates a standard imitation gradient step with a block of $k$ safety steps that pull the network's actions toward their projection onto the safe set; at run time, the controller is the trained network alone, with no safety filter. We analyze this scheme as inexact projected gradient descent in the space of policy actions. When the projected actions are recomputed at every safety step and each step moves the actions consistently toward the safe set, letting $k$ grow logarithmically yields asymptotic constraint satisfaction on the training states and bounds the distance to the constrained optimum of the imitation loss; with the projected actions held fixed, the same holds only if they are exactly representable by the network. On a nonlinear autonomous racing task, we compare our method with adding a weighted constraint-violation penalty to the imitation loss. With a sufficiently large weight, our method matches the lap time of unconstrained imitation while reducing the fraction of violating episodes from $15\%$ to $1\%$, about six times fewer than the penalty approach at its best weight. Its lap times are less sensitive to the weight, which instead sets how quickly violations vanish during training. In racing, the safety corrections are sparse and the conditions of the analysis do not hold; the gain arises instead through the data collected during training. These gains come at the cost of additional training computation. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Systems and Control (eess.SY) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.07167 — 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 2026-10-07T11:19:24.878Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).