Flexible, interactive driving of universal humanoid controllers

Determine a flexible and interactive command mechanism for driving universal humanoid whole-body controllers, which currently execute diverse behaviors by tracking kinematic reference trajectories or via teleoperation, so that user intent can guide the controller during execution.

Background

Recent universal humanoid whole-body controllers can robustly execute diverse, coordinated motions on real hardware but are usually driven by either predefined trajectories or human teleoperation. Predefined trajectories limit adaptability to evolving intent, while teleoperation requires continuous human effort.

The paper highlights that bridging high-level human intent expression with low-level humanoid execution remains unresolved, motivating the need for an interactive mechanism to drive such controllers flexibly.

References

However, how to flexibly and interactively drive such universal controllers remains an open challenge.

TextOp: Real-time Interactive Text-Driven Humanoid Robot Motion Generation and Control  (2602.07439 - Xie et al., 7 Feb 2026) in Section 1 (Introduction)