Research

Paper

TESTING March 10, 2026

SCDP: Learning Humanoid Locomotion from Partial Observations via Mixed-Observation Distillation

Authors

Milo Carroll, Tianhu Peng, Lingfan Bao, Chengxu Zhou, Zhibin Li

Abstract

Distilling humanoid locomotion control from offline datasets into deployable policies remains a challenge, as existing methods rely on privileged full-body states that require complex and often unreliable state estimation. We present Sensor-Conditioned Diffusion Policies (SCDP) that enables humanoid locomotion using only onboard sensors, eliminating the need for explicit state estimation. SCDP decouples sensing from supervision through mixed-observation training: diffusion model conditions on sensor histories while being supervised to predict privileged future state-action trajectories, enforcing the model to infer the motion dynamics under partial observability. We further develop restricted denoising, context distribution alignment, and context-aware attention masking to encourage implicit state estimation within the model and to prevent train-deploy mismatch. We validate SCDP on velocity-commanded locomotion and motion reference tracking tasks. In simulation, SCDP achieves near-perfect success on velocity control (99-100%) and 93% tracking success in AMASS test set, performing comparable to privileged baselines while using only onboard sensors. Finally, we deploy the trained policy on a real G1 humanoid at 50 Hz, demonstrating robust real robot locomotion without external sensing or state estimation.

Metadata

arXiv ID: 2603.09574
Provider: ARXIV
Primary Category: cs.RO
Published: 2026-03-10
Fetched: 2026-03-11 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.09574v1</id>\n    <title>SCDP: Learning Humanoid Locomotion from Partial Observations via Mixed-Observation Distillation</title>\n    <updated>2026-03-10T12:21:11Z</updated>\n    <link href='https://arxiv.org/abs/2603.09574v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.09574v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Distilling humanoid locomotion control from offline datasets into deployable policies remains a challenge, as existing methods rely on privileged full-body states that require complex and often unreliable state estimation. We present Sensor-Conditioned Diffusion Policies (SCDP) that enables humanoid locomotion using only onboard sensors, eliminating the need for explicit state estimation. SCDP decouples sensing from supervision through mixed-observation training: diffusion model conditions on sensor histories while being supervised to predict privileged future state-action trajectories, enforcing the model to infer the motion dynamics under partial observability. We further develop restricted denoising, context distribution alignment, and context-aware attention masking to encourage implicit state estimation within the model and to prevent train-deploy mismatch. We validate SCDP on velocity-commanded locomotion and motion reference tracking tasks. In simulation, SCDP achieves near-perfect success on velocity control (99-100%) and 93% tracking success in AMASS test set, performing comparable to privileged baselines while using only onboard sensors. Finally, we deploy the trained policy on a real G1 humanoid at 50 Hz, demonstrating robust real robot locomotion without external sensing or state estimation.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.RO'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-03-10T12:21:11Z</published>\n    <arxiv:comment>6 pages, 8 figures, 5 tables, iRos</arxiv:comment>\n    <arxiv:primary_category term='cs.RO'/>\n    <author>\n      <name>Milo Carroll</name>\n    </author>\n    <author>\n      <name>Tianhu Peng</name>\n    </author>\n    <author>\n      <name>Lingfan Bao</name>\n    </author>\n    <author>\n      <name>Chengxu Zhou</name>\n    </author>\n    <author>\n      <name>Zhibin Li</name>\n    </author>\n  </entry>"
}