Research map / Reinforcement learning / Policy gradient and actor-critic
Reduced Policy Optimization for Continuous Control with Hard Constraints
Shutong Ding, Jingya Wang, Yali Du, Ye Shi
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 |
|---|---|---|---|
| add two-stage decision process (construction plus GRG projection) to DDPG with Lagrangian relaxation (ablation) | DDPG · Safe CartPole | episodic reward: 15.5000 -> 200.0000 | Improved |
| add two-stage decision process to SAC with Lagrangian relaxation (ablation) | SAC · Safe CartPole | episodic reward: 40.5000 -> 200.0000 | Improved |
| add two-stage decision process to DDPG (ablation) | DDPG · Spring Pendulum | episodic reward: 76.4383 -> 175.1558 | Improved |
| add two-stage decision process to SAC (ablation) | SAC · Spring Pendulum | episodic reward: 149.6762 -> 182.4221 | Improved |
| add two-stage decision process to DDPG (ablation) | DDPG · OPF with Battery Energy Storage | episodic reward: -22.0649 -> 42.9710 | Improved |
| add two-stage decision process to SAC (ablation) | SAC · OPF with Battery Energy Storage | episodic reward: -19.4738 -> 42.0764 | Improved |
| add two-stage decision process to DDPG (ablation) | DDPG · Spring Pendulum | max instantaneous equality violation: 0.3805 -> 0.0000 | Improved |
| RPO-SAC versus Safety Layer | RPO-SAC · Safe CartPole | max instantaneous equality violation: 1.7099 -> 0.0000 | Improved |
| RPO versus Safe RL baseline CUP | RPO-SAC · Spring Pendulum | episodic reward: 0.8599 -> 182.4221 | Improved |
| RPO versus CPO | RPO-DDPG · OPF with Battery Energy Storage | max instantaneous equality violation: 0.4413 -> 0.0001 | Improved |
| RPO versus DDPG-L on inequality violation | RPO-DDPG · OPF with Battery Energy Storage | max instantaneous inequality violation: 0.0000 -> 0.0002 | Hurt |
Closest papers by meaning
- Constrained Reinforcement Learning Under Model Mismatch (ICML 2024 · similarity 0.88)
- Feasible Policy Optimization for Safe Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- Last-Iterate Global Convergence of Policy Gradients for Constrained Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- On the Global Optimality of Policy Gradient Methods in General Utility Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Reward Penalties on Augmented States for Solving Richly Constrained RL Effectively (AAAI 2024 · similarity 0.86)
- Absolute State-wise Constrained Policy Optimization: High-Probability State-wise Constraints Satisfaction (NeurIPS 2026 · 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.
