Saved in:
Bibliographic Details
Main Authors: Huber, Stefan, Unger, Hannes, Schäfer, Georg, Rehrl, Jakob
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.22305
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910278450413568
author Huber, Stefan
Unger, Hannes
Schäfer, Georg
Rehrl, Jakob
author_facet Huber, Stefan
Unger, Hannes
Schäfer, Georg
Rehrl, Jakob
contents We analytically solve the Mountain Car problem, a canonical benchmark in RL, and derive an optimal control solution, closing a gap after 36 years. This enables us to reveal two surprising insights: The optimal control is quite simple, yet modern RL agents display a large gap to optimality. Motivated by the analysis of the optimal control, we introduce Chebyshev policies as a universal (i.e. dense) class of RL policies from first principles. They can be trained as drop-in replacements of neural nets, reducing the regret by a factor of 4.18, while requiring 277 times fewer parameters, fostering sample efficiency, explainability and realtime capability. Chebyshev policies are evaluated on further RL tasks, including a real-world nonlinear motion control testbed. They consistently improve performance over neural nets with PPO, ARS and REINFORCE. Our results demonstrate how Chebyshev policies offer a compelling and lightweight alternative or addition to neural nets for low-dimensional control tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2605_22305
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Chebyshev Policies and the Mountain Car Problem: Reinforcement Learning for Low-Dimensional Control Tasks
Huber, Stefan
Unger, Hannes
Schäfer, Georg
Rehrl, Jakob
Machine Learning
We analytically solve the Mountain Car problem, a canonical benchmark in RL, and derive an optimal control solution, closing a gap after 36 years. This enables us to reveal two surprising insights: The optimal control is quite simple, yet modern RL agents display a large gap to optimality. Motivated by the analysis of the optimal control, we introduce Chebyshev policies as a universal (i.e. dense) class of RL policies from first principles. They can be trained as drop-in replacements of neural nets, reducing the regret by a factor of 4.18, while requiring 277 times fewer parameters, fostering sample efficiency, explainability and realtime capability. Chebyshev policies are evaluated on further RL tasks, including a real-world nonlinear motion control testbed. They consistently improve performance over neural nets with PPO, ARS and REINFORCE. Our results demonstrate how Chebyshev policies offer a compelling and lightweight alternative or addition to neural nets for low-dimensional control tasks.
title Chebyshev Policies and the Mountain Car Problem: Reinforcement Learning for Low-Dimensional Control Tasks
topic Machine Learning
url https://arxiv.org/abs/2605.22305