Research map / Reinforcement learning / Policy gradient and actor-critic
Robust Adaptive Multi-Step Predictive Shielding
Tanmay Ambadkar, Darshan Chudiwal, Greg Anderson, Abhinav Verma
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 RAMPS multi-step robust shield instead of SPICE with the same Koopman model | PPO agent · SafeHopper | cumulative violations: 459 -> 172 | Improved |
| use RAMPS shield instead of CMDP method SauteRL | PPO/SAC agent · SafeHopper | cumulative violations: 703 -> 49 | Improved |
| use RAMPS shield instead of CUP | PPO agent · SafeAnt | cumulative violations: 1883 -> 111 | Improved |
| use RAMPS shield instead of CMDP baselines | PPO agent · SafeHumanoid | cumulative violations: 172 -> 154 (CUP vs RAMPS+K+PPO); 11 with SAC | Mixed |
| use RAMPS + K + SAC instead of PPO | RAMPS · SafeAnt | cumulative violations: 111 -> 242 | Hurt |
| use Koopman model instead of linear regression | RAMPS · SafeCheetah | cumulative violations / reward: 7 -> 26 violations | Mixed |
| use Koopman model instead of linear regression | RAMPS · SafeCheetah | episodic reward | Improved |
| remove error-aware tightening term (ablation) | RAMPS shield · safety-gym environments | safety violations | Hurt |
| use 99th percentile error bound (ablation) | RAMPS shield · safety-gym environments | safety and reward | Improved |
| overly conservative CBF decay rate lambda (ablation) | RAMPS shield · safety-gym environments | safety and reward | Hurt |
| use RAMPS on 21-dim safety set | RAMPS vs CMDP baselines · SafeHumanoid (21-dim safety) | cumulative violations: 256 vs >3000 | Improved |
Closest papers by meaning
- Dynamic Model Predictive Shielding for Provably Safe Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Reinforcement Learning with Adaptive Regularization for Safe Control of Critical Systems (NeurIPS 2024 · similarity 0.87)
- A Physics-Informed Machine Learning Framework for Safe and Optimal Control of Autonomous Systems (ICML 2025 · similarity 0.86)
- Probabilistic Shielding for Safe Reinforcement Learning (AAAI 2025 · similarity 0.86)
- Safety Representations for Safer Policy Learning (ICLR 2025 · similarity 0.86)
- Dyna-Style Safety Augmented Reinforcement Learning: Staying Safe in the Face of Uncertainty (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.
