Research map / Multi-agent systems and games / Human-AI and goal-conditioned
DMAP: a Distributed Morphological Attention Policy for learning to locomote with a changing body
Alberto Chiappa, Alessandro Marin Vargas, Alexander Mathis
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 |
|---|---|---|---|
| remove morphology encoding at test time (DMAP-ne) (ablation) | DMAP · Walker sigma=0.1 | mean episode reward: 1229 -> 337 | Hurt |
| remove morphology encoding at test time (ablation) | DMAP · Hopper sigma=0.1 | mean episode reward: 1842 -> 984 | Hurt |
| remove morphology encoding at test time (ablation) | DMAP · Half Cheetah sigma=0.1 | mean episode reward: 2261 -> 1493 | Hurt |
| remove morphology encoding at test time (ablation) | DMAP · Ant sigma=0.5 | mean episode reward: 960 -> 881 | Hurt |
| replace TCN encoder with DMAP attention encoder and distributed controllers (ablation) | TCN · Ant sigma=0.3 | mean episode reward: 251 -> 1623 | Improved |
| replace TCN encoder with DMAP (ablation) | TCN · Walker sigma=0.3 | mean episode reward: 518 -> 893 | Improved |
| replace TCN encoder with DMAP (ablation) | TCN · Hopper sigma=0.3 | mean episode reward: 1368 -> 1316 | No clear effect |
| replace TCN encoder with DMAP (ablation) | TCN · Half Cheetah sigma=0.1 | mean episode reward: 2278 -> 2261 | No clear effect |
| give Oracle morphology parameters instead of proprioception only (Simple) | Simple · Ant sigma=0.5 | mean episode reward: 391 -> 974 | Improved |
| give Oracle morphology parameters instead of proprioception only | Simple · Ant sigma=0.1 | mean episode reward: 2164 -> 2148 | No clear effect |
| give Oracle morphology parameters | Simple · Walker sigma=0.1 | mean episode reward: 1909 -> 1780 | Hurt |
| replace Oracle with two-step RMA regressed encoding | Oracle · Hopper sigma=0.5 | mean episode reward: 919 -> 730 | Hurt |
Closest papers by meaning
- Convergent Functions, Divergent Forms (NeurIPS 2025 · similarity 0.89)
- BodyGen: Advancing Towards Efficient Embodiment Co-Design (ICLR 2025 · similarity 0.87)
- Hierarchical World Models as Visual Whole-Body Humanoid Controllers (ICLR 2025 · similarity 0.87)
- DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.87)
- Distilling Morphology-Conditioned Hypernetworks for Efficient Universal Morphology Control (ICML 2024 · similarity 0.86)
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (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.
