Research map / Probabilistic and scientific modeling / Molecules and proteins
ECo-MoE: Embodiment-Conditioned Mixture of Experts Increases the Evolvability of Robots
Yibin Wang, Muhan Li, Zihan Guo, Sam Kriegman
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 |
|---|---|---|---|
| replace monolithic controller with embodiment-conditioned mixture of experts | universal controller · Flat Ground | fitness | No clear effect |
| replace monolithic controller with embodiment-conditioned mixture of experts | universal controller · Upright Locomotion | fitness | Improved |
| replace monolithic controller with embodiment-conditioned mixture of experts | universal controller · Potholes | fitness | Improved |
| add routing diversity regularization to the gate (ablation) | ECo-MoE · Upright Locomotion | expert usage | Improved |
| use 2 or 8 experts instead of 4 (ablation) | ECo-MoE · Upright Locomotion | fitness | No clear effect |
| replace monolithic controller with ECo-MoE | universal controller · Upright Locomotion | effective limb count and mass bias | Improved |
| add predesigned latent initialization (evo by demo) (ablation) | ECo-MoE · Upright Locomotion | fitness | Improved |
| add frozen pretrained expert to predesigned initialization (CoSteering) (ablation) | ECo-MoE · Upright Locomotion demo 1 | morphological similarity to demo | Improved |
| add frozen pretrained expert without predesigned initialization (PretrainOnly) (ablation) | ECo-MoE · Upright Locomotion | fitness | No clear effect |
| add frozen pretrained expert to predesigned initialization (ablation) | ECo-MoE · Upright Locomotion demo 3 | morphological alignment with demo | Hurt |
| remove clones of each design in population (ablation) | evolution setup · Upright Locomotion | fitness: same fitness at half compute | No clear effect |
Closest papers by meaning
- Convergent Functions, Divergent Forms (NeurIPS 2025 · similarity 0.90)
- Embracing Evolution: A Call for Body-Control Co-Design in Embodied Humanoid Robot (NeurIPS 2026 · similarity 0.89)
- BodyGen: Advancing Towards Efficient Embodiment Co-Design (ICLR 2025 · similarity 0.87)
- Efficient Language-instructed Skill Acquisition via Reward-Policy Co-Evolution (AAAI 2025 · similarity 0.87)
- DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.87)
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · 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.
