Research map / Reinforcement learning / Imitation learning
Receding Horizon Inverse Reinforcement Learning
Yiqing Xu, Wei Gao, David Hsu
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 local receding-horizon control-sequence matching instead of global matching | RHIRL · CarRacing | return ratio: 0.09 -> 0.40 | Improved |
| Use a noise-disentangled state cost with explicit control-noise modeling | RHIRL · Hopper, mild and high control noise | return ratio: 0.82 -> 0.98; 0.43 -> 0.96 | Improved |
| Use a noise-disentangled state cost with explicit control-noise modeling | RHIRL · Walker2d, mild and high control noise | return ratio: 0.82 -> 0.99; 0.78 -> 0.95 | Improved |
| Increase receding horizon from K = 5 to K = 20 (ablation) | RHIRL · Hopper-v2 | final score: 1100 -> 3071.68 | Improved |
| Increase receding horizon from K = 20 to K = 100 (ablation) | RHIRL · Hopper-v2 | final score: 3071.68 -> 3083 | Improved |
| Increase receding horizon from K = 20 to K = 100 (ablation) | RHIRL · Hopper-v2 | environment steps to stabilize above 3000: fewer than 1e6 -> more than 1e7 | Hurt |
| Use RHIRL rather than f-IRL | IRL methods · Pendulum, no noise | return ratio: 1.07 -> 1.06 | No clear effect |
| Use RHIRL rather than f-IRL | IRL methods · LunarLander, no noise | return ratio: 0.76 -> 1.05 | Improved |
| Use RHIRL rather than f-IRL | IRL methods · Ant, no noise | return ratio: 0.87 -> 0.86 | No clear effect |
| Increase control noise in the evaluation environment | RHIRL learned cost · Pendulum | return ratio: 1.06 -> 1.07 -> 1.06 | No clear effect |
| Increase control noise in the evaluation environment | RHIRL learned cost · Ant | return ratio: 0.86 -> 0.55 -> 0.15 | Hurt |
Closest papers by meaning
- Hybrid Inverse Reinforcement Learning (ICML 2024 · similarity 0.90)
- Is Inverse Reinforcement Learning Harder than Standard Reinforcement Learning? A Theoretical Perspective (ICML 2024 · similarity 0.90)
- How does Inverse RL Scale to Large State Spaces? A Provably Efficient Approach (NeurIPS 2024 · similarity 0.90)
- Recent Advancements in Inverse Reinforcement Learning (AAAI 2024 · similarity 0.90)
- Offline Inverse RL: New Solution Concepts and Provably Efficient Algorithms (ICML 2024 · similarity 0.89)
- Massively Scalable Inverse Reinforcement Learning in Google Maps (ICLR 2024 · similarity 0.88)
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.
