Research map / Reinforcement learning / Policy gradient and actor-critic
On the Equilibrium Between Feasible Zone and Uncertain Model in Safe Exploration
Yujie Yang, Zhilong Zheng, Shengbo Eben Li
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 |
|---|---|---|---|
| increase initial transition radius r0 from 1.0 to 10.0 (ablation) | SEE · Double integrator | feasible zone recall: 100.00 -> 100.00 | No clear effect |
| increase initial transition radius r0 from 1.0 to 10.0 (ablation) | SEE · Double integrator | iterations to convergence: 3 -> 11 | Hurt |
| increase initial transition radius r0 from 2.0 to 5.0 (ablation) | SEE · Pendulum | feasible zone recall: 52.50 -> 10.10 | Hurt |
| enlarge known true-model region from r_x=0.0, r_u=0.0 to r_x=2.0, r_u=1.0 (ablation) | SEE · Double integrator | feasible zone recall: 0.04 -> 100.00 | Improved |
| enlarge known true-model region from r_x=3.0, r_u=1.0 to r_u=2.0 (ablation) | SEE · Pendulum | feasible zone recall: 4.55 -> 52.05 | Improved |
| enlarge known true-model region from r_x=3.0 to 4.0 (r_u=2.0) (ablation) | SEE · Pendulum | feasible zone recall: 52.05 -> 62.71 | Improved |
| increase Lipschitz constant L from 1.73 to 2.00 (ablation) | SEE · Double integrator | feasible zone recall: 100.00 -> 1.08 | Hurt |
| increase Lipschitz constant L from 1.73 to 2.50 (ablation) | SEE · Double integrator | feasible zone recall: 100.00 -> 0.84 | Hurt |
| increase Lipschitz constant L from 3.00 to 4.00 (ablation) | SEE · Pendulum | feasible zone recall: 52.05 -> 8.55 | Hurt |
| increase action Lipschitz constant L_u from 2.00 to 3.00 (ablation) | SEE · Pendulum | feasible zone recall: 52.05 -> 9.88 | Hurt |
| increase state Lipschitz constant L_x from 3.00 to 4.00 (ablation) | SEE · Pendulum | feasible zone recall: 52.05 -> 52.05 | No clear effect |
| increase L_x from 3.00 to 4.00 on iteration count (ablation) | SEE · Pendulum | iterations to convergence: 14 -> 14 | No clear effect |
Closest papers by meaning
- Sampling-Based Safe Reinforcement Learning (NeurIPS 2026 · similarity 0.89)
- Safe Reinforcement Learning with Instantaneous Constraints: The Role of Aggressive Exploration (AAAI 2024 · similarity 0.88)
- Safeguarded Progress in Reinforcement Learning: Safe Bayesian Exploration for Control Policy Synthesis (AAAI 2024 · similarity 0.87)
- Balance Reward and Safety Optimization for Safe Reinforcement Learning: A Perspective of Gradient Manipulation (AAAI 2024 · similarity 0.87)
- Risk-Sensitive Reward-Free Reinforcement Learning with CVaR (ICML 2024 · similarity 0.87)
- Dyna-Style Safety Augmented Reinforcement Learning: Staying Safe in the Face of Uncertainty (NeurIPS 2026 · 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.
