Embodiment-Aware Generalist Specialist Distillation for Unified Humanoid Whole-Body Control

Researchers introduce EAGLE, a reinforcement learning framework that creates unified control policies for multiple different humanoid robots without per-robot tuning. The system uses iterative generalist-specialist distillation to enable a single AI controller to manage diverse humanoid embodiments and support complex behaviors beyond basic walking.

• →EAGLE framework enables one AI policy to control multiple heterogeneous humanoid robots including Unitree H1, G1, and Fourier N1.
• →The system uses iterative distillation where robot-specific specialists train on individual embodiments then share knowledge with a generalist policy.
• →Testing was conducted on five different robots in simulation and four in real-world environments.
• →The approach eliminates the need for per-robot reward tuning while supporting complex behaviors like squatting and leaning.
• →Results show high tracking accuracy and robustness compared to existing methods for humanoid control.