Research map / Reinforcement learning / Policy gradient and actor-critic
When to Update Your Model: Constrained Model-based Reinforcement Learning
Tianying Ji, Yu Luo, Fuchun Sun, Mingxuan Jing, Fengxiang He, Wenbing Huang
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 |
|---|---|---|---|
| Use CMLO's event-triggered model updates instead of fixed model-training intervals (ablation) | CMLO with SAC · HalfCheetah | average return | Improved |
| Use CMLO's event-triggered model updates instead of fixed model-training intervals (ablation) | CMLO with SAC · Ant | average return | Improved |
| Use a fixed model-training interval of 250 instead of the event-triggered mechanism (ablation) | CMLO with SAC · HalfCheetah and Ant | average return | Hurt |
| Use CMLO's event-triggered model updates instead of fixed model-training intervals (ablation) | CMLO with TRPO · HalfCheetah | average return | Improved |
| Use CMLO's event-triggered model updates instead of fixed model-training intervals (ablation) | CMLO with TRPO · Ant | average return | Improved |
| Use event-triggered updates instead of the most-frequently triggered case (ablation) | CMLO with iLQR · DKitty-Stand | asymptotic performance | No clear effect |
| Use event-triggered updates instead of the most-frequently triggered case (ablation) | CMLO with iLQR · DKitty-Stand | computation cost | Improved |
| Use event-triggered updates instead of the most-frequently triggered case (ablation) | CMLO with iLQR · Panda-Reaching | asymptotic performance | No clear effect |
| Use event-triggered updates instead of the most-frequently triggered case (ablation) | CMLO with iLQR · Panda-Reaching | computation cost | Improved |
| Use CMLO's event-triggered model updates and constrained model-shift strategy | CMLO with SAC · Six continuous-control tasks: HalfCheetah, Hopper, Walker2d, Swimmer, Ant, and Humanoid | sample efficiency | Improved |
| Use CMLO's event-triggered model updates and constrained model-shift strategy | CMLO with SAC · Six continuous-control tasks: HalfCheetah, Hopper, Walker2d, Swimmer, Ant, and Humanoid | asymptotic performance | Improved |
| Use CMLO's event-triggered model updates and constrained model-shift strategy | CMLO with SAC · Six continuous-control tasks: HalfCheetah, Hopper, Walker2d, Swimmer, Ant, and Humanoid | asymptotic performance | No clear effect |
Closest papers by meaning
- Model-based RL as a Minimalist Approach to Horizon-Free and Second-Order Bounds (ICLR 2025 · similarity 0.89)
- OCEAN-MBRL: Offline Conservative Exploration for Model-Based Offline Reinforcement Learning (AAAI 2024 · similarity 0.88)
- Constrained Reinforcement Learning Under Model Mismatch (ICML 2024 · similarity 0.88)
- Towards General-Purpose Model-Free Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Last-Iterate Global Convergence of Policy Gradients for Constrained Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Robust Model Based Reinforcement Learning Using $\mathcal{L}_1$ Adaptive Control (ICLR 2024 · similarity 0.87)
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.
