Research map / Reinforcement learning / Visual and representation RL
Learning to Reconfigure: Configuration-Control Co-optimization of Reconfigurable Robots for Heterogeneous Locomotion
Xiaoyu Xiong, Kehan Liu, Huiyi Yan, Shengjie Wang, Yang Gao, Tao Du
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 epsilon-greedy exploration to the design tails (ablation) | Learning to Reconfigure · Fly-Walk (Mount) | normalized traversal progress S: 0.335 -> 1.978 | Improved |
| add epsilon-greedy exploration to the design tails (ablation) | Learning to Reconfigure · Wheel-Fly (Mount) | normalized traversal progress S: 3.820 -> 7.368 | Improved |
| add epsilon-greedy exploration to the design tails (ablation) | Learning to Reconfigure · Wheel-Leg | normalized traversal progress S: 8.923 -> 9.148 | Improved |
| add staged optimization (freeze lock angle tail initially) (ablation) | Learning to Reconfigure · Wheel-Fly (Mount) | normalized traversal progress S: 0.325 -> 7.368 | Improved |
| add staged optimization (freeze lock angle tail initially) (ablation) | Learning to Reconfigure · Fly-Walk (Mount) | normalized traversal progress S: 0.437 -> 1.978 | Improved |
| use hierarchical primitives plus scheduler instead of end-to-end single policy (ablation) | Learning to Reconfigure · 7 heterogeneous locomotion tasks | normalized traversal progress S: 9.83x average boost | Improved |
| use learned scheduler instead of rule-based heuristic scheduler (ablation) | Learning to Reconfigure · 7 heterogeneous locomotion tasks | normalized traversal progress S | Improved |
| use reconfigurable learned configurations instead of non-reconfigurable co-design (BodyGen, GLSO) | Learning to Reconfigure · 7 heterogeneous locomotion tasks | normalized traversal progress S: 10.52x BodyGen; 9.10x GLSO on average | Improved |
| use learned configurations instead of human domain-expert configurations | Learning to Reconfigure · 7 heterogeneous locomotion tasks | normalized traversal progress S: 1.19x on average | Improved |
| use PPO without reconfiguration | PPO · Car-Leg (Narrow) | normalized traversal progress S: 2.318 vs 2.618 (ours) | Mixed |
| train BodyGen/GLSO from the same initial morphology instead of scratch | BodyGen, GLSO · Fly-Walk (Mount) and Car-Leg | normalized traversal progress S: BodyGen 0.498 -> 0.452; GLSO 0.507 -> 0.496 | No clear effect |
Closest papers by meaning
- DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.89)
- Convergent Functions, Divergent Forms (NeurIPS 2025 · similarity 0.87)
- ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Embracing Evolution: A Call for Body-Control Co-Design in Embodied Humanoid Robot (NeurIPS 2026 · similarity 0.86)
- BodyGen: Advancing Towards Efficient Embodiment Co-Design (ICLR 2025 · similarity 0.85)
- Debate2Create: Robot Co-design via Multi-Agent LLM Debate (ICML 2026 · similarity 0.85)
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.
