# OmniClone: Engineering a Robust, All-Rounder Whole-Body Humanoid Teleoperation System

Source: [OmniClone: Engineering a Robust, All-Rounder Whole-Body Humanoid Teleoperation System](<https://www.sota2.com/research/paper/69b8e03d0636e5bf4afbeda2>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-18T05:00:11.457Z

Authors: Yixuan Li, Le Ma, Yutang Lin, Yushi Du, Mengya Liu, Kaizhe Hu, Jieming Cui, Yixin Zhu, Wei Liang, Baoxiong Jia, Siyuan Huang

Published: 2026-03-15T11:13:04.000Z

## Abstract

Whole-body humanoid teleoperation enables humans to remotely control humanoid robots, serving as both a real-time operational tool and a scalable engine for collecting demonstrations for autonomous learning. Despite recent advances, existing systems are validated using aggregate metrics that conflate distinct motion regimes, masking critical failure modes. This lack of diagnostic granularity, compounded by tightly coupled and labor-intensive system configurations, hinders robust real-world deployment. A key open challenge is building a teleoperation system that is simultaneously robust, versatile, and affordable for practical use. Here we present OmniClone, a whole-body humanoid teleoperation system that achieves high-fidelity, multi-skill control on a single consumer GPU with modest data requirements. Central to our approach is OmniBench, a diagnostic benchmark that evaluates policies across stratified motion categories and difficulty levels on unseen motions, exposing the narrow specialization of prior systems. Guided by these diagnostics, we identify an optimized training data recipe and integrate system-level improvements: subject-agnostic retargeting and robust communication, that collectively reduce Mean Per-Joint Position Error (MPJPE) by over 66% while requiring orders-of-magnitude fewer computational resources than comparable methods. Crucially, OmniClone is control-source-agnostic: a single unified policy supports real-time teleoperation, generated motion playback, and Vision-Language-Action (VLA) models, while generalizing across operators of vastly different body proportions. By uniting diagnostic evaluation with practical engineering, OmniClone provides an accessible foundation for scalable humanoid teleoperation and autonomous learning.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Humanoid Motion Tracking on OmniBench Walk (Slow)](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-walk-slow.md>): Humanoid motion tracking; OmniBench Walk (Slow); Success Rate (SR): 100.

- [Humanoid Motion Tracking on OmniBench Walk (Fast)](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-walk-fast.md>): Humanoid motion tracking; OmniBench Walk (Fast); Success Rate: 100.

- [Humanoid Motion Tracking on OmniBench Walk Medium](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-walk-medium.md>): Humanoid motion tracking; OmniBench Walk Medium; Success Rate (SR): 100.

- [Humanoid Motion Tracking on OmniBench Jump (High)](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-jump-high.md>): Humanoid motion tracking; OmniBench Jump (High); Success Rate (SR): 90.

- [Humanoid Motion Tracking on OmniBench Jump Medium](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-jump-medium.md>): Humanoid motion tracking; OmniBench Jump Medium; Success Rate: 100.

- [Squat on OmniBench Low 1.0 (test)](<https://www.sota2.com/research/sota/squat-on-omnibench-low-1-0-test.md>): Squat; OmniBench Low 1.0 (test); Success Rate (SR): 100.

- [Loco-Manip on OmniBench Medium 1.0 (test)](<https://www.sota2.com/research/sota/loco-manip-on-omnibench-medium-1-0-test.md>): Loco-Manip; OmniBench Medium 1.0 (test); Success Rate (SR): 100.

- [Loco-Manip on OmniBench Low 1.0 (test)](<https://www.sota2.com/research/sota/loco-manip-on-omnibench-low-1-0-test.md>): Loco-Manip; OmniBench Low 1.0 (test); Success Rate (SR): 100.

- [Humanoid Motion Tracking on OmniBench Run (Fast)](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-run-fast.md>): Humanoid motion tracking; OmniBench Run (Fast); Success Rate (SR): 100.

- [Squat on OmniBench High 1.0 (test)](<https://www.sota2.com/research/sota/squat-on-omnibench-high-1-0-test.md>): Squat; OmniBench High 1.0 (test); Success Rate (SR): 100.

- [Squat on OmniBench Medium 1.0 (test)](<https://www.sota2.com/research/sota/squat-on-omnibench-medium-1-0-test.md>): Squat; OmniBench Medium 1.0 (test); SR: 100.

- [Loco-Manip on OmniBench High 1.0 (test)](<https://www.sota2.com/research/sota/loco-manip-on-omnibench-high-1-0-test.md>): Loco-Manip; OmniBench High 1.0 (test); Success Rate (SR): 100.

- [Humanoid Motion Tracking on OmniBench Run Medium](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-run-medium.md>): Humanoid motion tracking; OmniBench Run Medium; Success Rate (SR): 100.

- [Humanoid Motion Tracking on OmniBench Jump (Low)](<https://www.sota2.com/research/sota/humanoid-motion-tracking-on-omnibench-jump-low.md>): Humanoid motion tracking; OmniBench Jump (Low); Success Rate: 100.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.14327>)

[PDF](<https://arxiv.org/pdf/2603.14327.pdf>)
