Research map / Video / Video segmentation and tracking
Hierarchical Visuomotor Control of Humanoids
Josh Merel, Arun Ahuja, Vu Pham, Saran Tunyasuvunakool, Siqi Liu, Dhruva Tirumala, Nicolas Heess, Greg Wayne
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 |
|---|---|---|---|
| Use discrete selection among control fragments instead of the compared control approaches | Humanoid visuomotor control · Go-to-target, Walls, Gaps, Forage, Heterogeneous Forage | task performance | Improved |
| Use hierarchical control with pretrained low-level skills instead of training a flat policy from scratch | Humanoid visuomotor control · Walls, Go-to-target | task performance | Improved |
| Use hierarchical control with pretrained low-level skills instead of training a flat policy from scratch | Humanoid visuomotor control · Forage | task performance | Improved |
| Use control fragments rather than a fixed-turning-radius steerable controller | Humanoid visuomotor control · Go-to-target | task performance | Improved |
| Use control fragments rather than a structured switching controller with predefined transitions | Humanoid visuomotor control · Walls | forward velocity | Improved |
| Use control fragments rather than a structured switching controller with predefined transitions | Humanoid visuomotor control · Forage | task performance | Improved |
| Use a query-based selector instead of discrete control-fragment selection | Control-fragment high-level controller · Core tasks | task performance | Hurt |
| Add low-level policy pretraining by supervised imitation before reinforcement learning | Single-clip tracking policies · Motion-capture tracking | training convergence and resulting policy quality | Improved |
| Use control fragments longer than one time step (ablation) | Control-fragment controller · Go-to-target | early exploration | Improved |
| Use control fragments longer than one time step (ablation) | Control-fragment controller · Go-to-target | ultimate task performance | Hurt |
| Add functionally similar, redundant behaviors to the control-fragment set (ablation) | Control-fragment controller · Go-to-target | training speed | Improved |
| Add extraneous policies involving movements unrelated to locomotion (ablation) | Control-fragment controller · Go-to-target | training performance | No clear effect |
Closest papers by meaning
- Hierarchical World Models as Visual Whole-Body Humanoid Controllers (ICLR 2025 · similarity 0.91)
- HumanoidArena: Benchmarking Egocentric Hierarchical Whole-body Learning (NeurIPS 2026 · similarity 0.89)
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (NeurIPS 2026 · similarity 0.88)
- Universal Humanoid Motion Representations for Physics-Based Control (ICLR 2024 · similarity 0.87)
- Embracing Evolution: A Call for Body-Control Co-Design in Embodied Humanoid Robot (NeurIPS 2026 · 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.
