Research map / 3D vision / Human pose estimation
FARM: Frame-Accelerated Augmentation and Residual Mixture-of-Experts for Physics-Based High-Dynamic Humanoid Control
Jing Tan, Shiting Chen, Yangfan Li, Weisheng Xu, Renjing Xu
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 frame-accelerated augmentation alone to the FC controller without MoE (ablation) | FC · HDHM | MPJPEg: 111.3 -> 111.4 | No clear effect |
| add residual MoE on top of frame acceleration (ablation) | FC + FA · HDHM | MPJPEg: 111.4 -> 95.0 | Improved |
| use residual MoE (frozen base) instead of Full-MoE (ablation) | FARM · AMASS-test | MPJPEg: 47.3 -> 42.1 | Improved |
| use residual MoE instead of Full-MoE (ablation) | FARM · HDHM | MPJPEg: 104.9 -> 95.0 | Improved |
| add speed-aware router (ablation) | FARM · HDHM | MPJPEg: 97.5 -> 95.0 | Improved |
| use dynamic expert assignment instead of Top1 (ablation) | FARM · HDHM | MPJPEg: 97.5 -> 95.0 | Improved |
| use dynamic expert assignment instead of Top2 (ablation) | FARM · HDHM | MPJPEg: 98.1 -> 95.0 | Improved |
| add frame acceleration to residual MoE (ablation) | FARM · HDHM | MPJPEg: 97.0 -> 95.0 | Improved |
| fine-tune on failure cases only instead of full data (ablation) | FARM · HDHM | MPJPEg: 99.2 -> 95.0 | Improved |
| add FARM to FC controller | FC · AMASS-test | MPJPEg: 40.8 -> 42.1 | Hurt |
| add FARM to FC controller | FC · HDHM | success rate: 92.3% -> 95.6% | Improved |
| vary max acceleration factor and number of experts away from default (ablation) | FARM · HDHM, AMASS-test | MPJPEg | Hurt |
Closest papers by meaning
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (NeurIPS 2026 · similarity 0.87)
- Universal Humanoid Motion Representations for Physics-Based Control (ICLR 2024 · similarity 0.87)
- KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills (NeurIPS 2025 · similarity 0.86)
- H-GAP: Humanoid Control with a Generalist Planner (ICLR 2024 · similarity 0.86)
- From Experts to a Generalist: Toward General Whole-Body Control for Humanoid Robots (NeurIPS 2025 · similarity 0.86)
- SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework (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.
