Event

PAKE: A Hierarchical Control Framework for Whole-Body Mobile Manipulation with Quadruped Robots

1 reports · 1 sources

Zhengmao He et al. propose PAKE, a whole-body mobile manipulation control framework that decomposes the problem into partial-reference motion generation and low-level imitation control, and validate it on a quadruped robot equipped with a six-degree-of-freedom robotic arm. PAKE introduces a Kinematic Normalizing Flow (KNF) model trained on a large-scale motion generation dataset to generate diverse partial motion references; the high-level controller navigates in the KNF latent space to exploit redundant solutions, while the low-level controller learns dynamics-consistent action execution. The paper reports that on a simulation benchmark containing 4,000 targets, PAKE outperforms the evaluated baselines in tracking accuracy and reachable workspace coverage; hardware deployment was evaluated in 24 trials across 8 different mobile manipulation tasks.