Understanding Reward Shaping in Planning: A Theoretical and Empirical Analysis
Proceedings of the International Symposium on Combinatorial Search · 14.08.2026 · 19 (1) · 286-290
DOI: 10.1609/socs.v19i1.43101 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the International Symposium on Combinatorial Search, online erschienen am 14.08.2026. Verfasst von Oleksii Shuhailo und 3 weiteren Autor:innen.
Abstract (Originalsprache)
Reward design is a critical yet often underexplored component of reinforcement learning for planning problems. In this work we analyze how different reward shaping strategies affect learning performance when using Proximal Policy Optimization (PPO). We evaluate several reward formulations derived from the objective function while keeping the learning algorithm and model architecture fixed. Experiments on benchmarks from the International Planning Competition 2023 show that commonly used shaping strategies do not consistently outperform sparse rewards. While some formulations provide competitive performance, our results highlight the difficulty of designing effective reward signals for policy gradient methods in planning domains.
Zitieren
APA
Shuhailo, O., Ghaffari, M., Chvalovský, K., & Pevný, T. (2026). Understanding Reward Shaping in Planning: A Theoretical and Empirical Analysis. Proceedings of the International Symposium on Combinatorial Search, 19(1), 286-290. https://doi.org/10.1609/socs.v19i1.43101
Vancouver
Shuhailo O, Ghaffari M, Chvalovský K, Pevný T. Understanding Reward Shaping in Planning: A Theoretical and Empirical Analysis. Proceedings of the International Symposium on Combinatorial Search. 2026;19(1):286-290. doi:10.1609/socs.v19i1.43101
BibTeX
@article{Shuhailo2026,
author = {Shuhailo, Oleksii and Ghaffari, Mohsen and Chvalovský, Karel and Pevný, Tomáš},
title = {Understanding Reward Shaping in Planning: A Theoretical and Empirical Analysis},
journal = {Proceedings of the International Symposium on Combinatorial Search},
year = {2026},
volume = {19},
number = {1},
pages = {286--290},
doi = {10.1609/socs.v19i1.43101}
}
EUVIMED