Event

HIPC Iterative Planning and Control Framework Proposed for Humanoid Robots

1 reports · 1 sources

Yi Han and other authors propose HIPC, a real-time iterative planning and control framework for humanoid robots that couples generative models with motion tracking into a unified planning and control paradigm. The framework adopts goal-conditioned motion planning and is trained with a canonical motion representation to avoid gimbal-lock singularities under extreme poses; it also uses two strategies, deviation-decay recovery and historical re-anchoring, to mitigate out-of-distribution shifts caused by real-robot state feedback. HIPC runs as a real-time asynchronous planning-control system and adds delay compensation to reduce temporal mismatch caused by inference latency. The authors state that experiments show HIPC achieves the best goal-reaching accuracy among existing methods and exhibits robustness to physical perturbations and the ability to recover after falls.