Research map / Reinforcement learning / Policy gradient and actor-critic
Policy Gradient Method For Robust Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 robust rather than vanilla policy gradient when R>0 | Garnet problems and Taxi · Garnet problems G(12,6) and G(20,10); Taxi | worst-case performance | Helped |
| Set R=0, removing model-mismatch robustness (ablation) | Robust Policy Gradient · Garnet problems and Taxi | performance | No clear effect |
| Increase the smoothing parameter σ (ablation) | Smoothed Robust Policy Gradient · Garnet G(12,6) | objective value | Helped |
| Use a small smoothing parameter, such as σ=1 (ablation) | Smoothed Robust Policy Gradient · Garnet G(12,6) | objective value | Hurt |
| Use robust rather than vanilla actor-critic | Actor-Critic · Garnet G(30,20) | worst-case accumulated discounted reward | Helped |
| Use robust actor-critic rather than RARL | Taxi · Taxi | worst-case accumulated discounted reward | Helped |
| Smooth the robust objective with LogSumExp to enable gradient-based optimization | Smoothed Robust Policy Gradient · Robust RL, direct policy parameterization | iterations to ε-global optimum: \mathcal{O}(\epsilon^{-3}) | Helped |
| Set R=0, removing model-mismatch robustness (ablation) | Smoothed Robust Policy Gradient · Non-robust RL | iterations to ε-global optimum: \mathcal{O}(\epsilon^{-3}) -> \mathcal{O}(\epsilon^{-2}) | Helped |
Closest papers by meaning
- Constrained Reinforcement Learning Under Model Mismatch (ICML 2024 · similarity 0.91)
- Robust Offline Reinforcement Learning with Linearly Structured $f$-Divergence Regularization (ICML 2025 · similarity 0.89)
- Minimax Optimal and Computationally Efficient Algorithms for Distributionally Robust Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Adversarial Diffusion for Robust Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Model-Free Robust $\phi$-Divergence Reinforcement Learning Using Both Offline and Online Data (ICML 2024 · similarity 0.89)
- Improve Robustness of Reinforcement Learning against Observation Perturbations via l∞ Lipschitz Policy Networks (AAAI 2024 · similarity 0.89)
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.