Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Learning to Control Self-Assembling Morphologies: A Study of Generalization via Modularity
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a shared-parameter DGN controller rather than a monolithic controller | Six-limb agents · Standing Up (training) | reward: 17518 vs 4104 | Helped |
| Use a shared-parameter DGN controller rather than a monolithic controller | Six-limb agents · Locomotion (training) | distance: 8.71 vs 0.96 | Helped |
| Use DGN instead of either monolithic baseline | Six-limb agents · Locomotion, Hurdles | distance: 6.39 vs -0.77; -3.12 | Helped |
| Use DGN instead of either monolithic baseline | Six-limb agents · Locomotion, Gaps in Terrain | distance: 3.25 vs -0.32; 2.09 | Mixed |
| Test the six-limb DGN policy with twice as many limbs | DGN trained on six limbs · Standing Up, 2x Limbs | reward: 19796 (113%) | Helped |
| Test the six-limb DGN policy with half as many limbs | DGN trained on six limbs · Standing Up, .5x Limbs | reward: 10839 (62%) | Hurt |
| Test the six-limb DGN policy with a different limb count | DGN trained on six limbs · Locomotion, 2x and .5x Limbs | distance: 5.47 (63%); 6.64 (76%) | Hurt |
| Use DGN rather than the monolithic policies in novel terrain and force conditions | Six-limb agents · Novel environments | reward or distance: Table 2 values vary by condition | Mixed |
| Add message passing to the DGN (ablation) | DGN controller · Standing Up | performance | Helped |
| Add message passing to the DGN (ablation) | DGN controller · Locomotion | performance | No clear effect |
| Add dynamic self-assembly to a modular DGN policy on fixed morphology (ablation) | Modular DGN controller · Standing Up and locomotion training | performance | Helped |
| Replace modular DGN control with a monolithic controller on fixed morphology (ablation) | Fixed-morphology agent · Standing Up and locomotion training | performance | Hurt |
Closest papers by meaning
- Convergent Functions, Divergent Forms (NeurIPS 2025 · similarity 0.88)
- Embracing Evolution: A Call for Body-Control Co-Design in Embodied Humanoid Robot (NeurIPS 2026 · similarity 0.88)
- DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.87)
- DiffeoMorph: Learning to Morph 3D Shapes Using Differentiable Agent-Based Simulations (NeurIPS 2026 · similarity 0.87)
- BodyGen: Advancing Towards Efficient Embodiment Co-Design (ICLR 2025 · similarity 0.86)
- SEEA-R1: Tree-Structured Reinforcement Fine-Tuning for Self-Evolving Embodied Agents (NeurIPS 2025 · 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.