CRISP: Curriculum-Inducing Primitive Informed Subgoal Prediction for Boosting Hierarchical Reinforcement Learning
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (30) · 25481-25489
DOI: 10.1609/aaai.v40i30.39743 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Utsav Singh und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Hierarchical reinforcement learning (HRL) leverages temporal abstraction to efficiently tackle complex long-horizon tasks. However, HRL often collapses because the low-level primitive’s continual updates make earlier sub-goals issued by the high-level policy obsolete, introducing non-stationarity that destabilizes training. We propose CRISP, a curriculum-driven framework that tackles this instability with three key ingredients: (1) primitive-informed parsing (PIP), which adaptively re-labels a handful of expert demonstrations to always generate reachable subgoals by the current low-level primitive; (2) an inverse-reinforcement-learning regularizer that steers the high-level policy toward the expert-induced subgoal distribution and stabilizes learning; and (3) a unified training loop that leverages these components to boost sample efficiency. Across six sparse-reward robotic navigation and manipulation benchmarks, CRISP improves success rates by more than 40% over strong hierarchical and flat baselines and successfully transfers to real-world tasks, demonstrating the promise of curriculum-based HRL for practical scenarios.
Zitieren
APA
Singh, U., & Namboodiri, V. P. (2026). CRISP: Curriculum-Inducing Primitive Informed Subgoal Prediction for Boosting Hierarchical Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 40(30), 25481-25489. https://doi.org/10.1609/aaai.v40i30.39743
Vancouver
Singh U, Namboodiri VP. CRISP: Curriculum-Inducing Primitive Informed Subgoal Prediction for Boosting Hierarchical Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(30):25481-25489. doi:10.1609/aaai.v40i30.39743
BibTeX
@article{Singh2026,
author = {Singh, Utsav and Namboodiri, Vinay P.},
title = {CRISP: Curriculum-Inducing Primitive Informed Subgoal Prediction for Boosting Hierarchical Reinforcement Learning},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {30},
pages = {25481--25489},
doi = {10.1609/aaai.v40i30.39743}
}
EUVIMED