Research map / Reinforcement learning / Visual and representation RL
Chebyshev Policies and the Mountain Car Problem: Reinforcement Learning for Low-Dimensional Control Tasks
Stefan Huber, Hannes Unger, Georg Schäfer, Jakob Rehrl
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace MLP policy with Chebyshev policy (degree 3) | ARS · MountainCarContinuous-v0 | regret: 2.72 -> 0.44 | Improved |
| replace MLP policy with Chebyshev policy (degree 3) | PPO · MountainCarContinuous-v0 | regret: 5.48 -> 1.29 | Improved |
| train Chebyshev policy with REINFORCE instead of MLP with REINFORCE | REINFORCE · MountainCarContinuous-v0 | regret: 0.77 compared to 2.72 of ARS | Improved |
| replace MLP policy with Chebyshev policy | ARS · Pendulum-v1 | average return: -218.3 -> -150.8 | Improved |
| replace MLP policy with Chebyshev policy | PPO · Pendulum-v1 | average return: -176.2 -> -162.8 | Improved |
| replace MLP policy with Chebyshev policy | ARS · Aero 2 (real) | average return: -718.4 -> -164.2 | Improved |
| replace MLP policy with Chebyshev policy | PPO · Aero 2 (real) | average return: -182.0 -> -55.8 | Improved |
| replace MLP policy with Chebyshev policy | PPO · Aero 2 (simulation) | average return: -84.6 -> -49.2 | Improved |
| replace MLP policy with Chebyshev policy | ARS · Aero 2 (simulation) | average return: -721.8 -> -125.2 | Improved |
| reduce PPO MLP from [64,64] to [32,32] (ablation) | PPO · Mountain Car (PPO MLP architectures, 20 start positions) | mean return: 94.22 -> 92.60 | Hurt |
| reduce PPO MLP from [64,64] to [16,16] (ablation) | PPO · Mountain Car (PPO MLP architectures, 20 start positions) | mean return: 94.22 -> -0.22 | Hurt |
| train MLP policies with REINFORCE across 8 optimizers and 6 architectures | REINFORCE · Mountain Car (MLP) | goal reached | Hurt |
Closest papers by meaning
- Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (AAAI 2025 · similarity 0.86)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.86)
- Dual RL: Unification and New Methods for Reinforcement and Imitation Learning (ICLR 2024 · similarity 0.86)
- PWM: Policy Learning with Multi-Task World Models (ICLR 2025 · similarity 0.86)
- Oracle-Efficient Reinforcement Learning for Max Value Ensembles (NeurIPS 2024 · similarity 0.86)
- Diffusion Policy Policy Optimization (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
