GigaBrain-WBC-0.5: A Behavior World Model for Robust Humanoid Whole-Body Tracking with Environment Interaction
GigaBrain-WBC-0.5: A Behavior World Model for Robust Humanoid Whole-Body Tracking with Environment Interaction
This paper introduces InterTrack, the first behavior world model (BWM) designed for robust whole-body tracking on humanoids that includes environment interaction. The system uses a Transformer to predict actions, states, and behavior distributions conditioned on the environment. An automatic annotation pipeline reconstructs 3D support geometry from retargeted motions to scale training data. At deployment, the policy handles implausible commands with best-effort responses. Quantitative results show an 81.3% success rate on terrain interaction (4.3x the best baseline), a 99.3% fall-recovery rate, and improved free-space tracking compared to three leading baselines. The work also demonstrates real-time cross-terrain whole-body teleoperation with object interaction and robust recovery from falls.
The authors present InterTrack, a behavior world model for humanoid whole-body tracking that achieves an 81.3% success rate on terrain interaction and a 99.3% fall-recovery rate in their reported experiments.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Control
- TeleoperatedOpen source S4
Reported numbers
Reported success rate
81.3%
Reported trials
View original evidence
est-effort" manner. Quantitatively, InterTrack achieves an 81.3% success rate on terrain interaction (4.3 times the best evaluated baseline) and a 99.3% fall-recovery rate, while also improving free-space tracking and outperforming three leading tracking baselines across all of these regimes. To our knowledge, we provide the first demonstration of real-time cross-terrain whole-body teleoperation on a humanoid robot,
Open source S4Reported success rate
99.3%
Reported trials
View original evidence
est-effort" manner. Quantitatively, InterTrack achieves an 81.3% success rate on terrain interaction (4.3 times the best evaluated baseline) and a 99.3% fall-recovery rate, while also improving free-space tracking and outperforming three leading tracking baselines across all of these regimes. To our knowledge, we provide the first demonstration of real-time cross-terrain whole-body teleoperation on a humanoid robot,
Open source S4
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
est-effort" manner. Quantitatively, InterTrack achieves an 81.3% success rate on terrain interaction (4.3 times the best evaluated baseline) and a 99.3% fall-recovery rate, while also improving free-space tracking and outperforming three leading tracking baselines across all of these regimes. To our knowledge, we provide the first demonstration of real-time cross-terrain whole-body teleoperation on a humanoid robot,
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org