Search papers, labs, and topics across Lattice.
This paper addresses the challenge of temporal asynchrony between high-level humanoid planners and low-rate trajectory execution by introducing an asynchronous upper body task-space tracking framework. Utilizing teacher-student distillation, the proposed method conditions a student policy on future trajectories and an execution-time index, effectively reducing frame drift through a sliding-window global reward system. Experimental results demonstrate that the framework outperforms synchronous and decoupled baselines in both simulation and hardware tests, particularly under low update rates and in adapting to out-of-distribution motions.
Asynchronous tracking for humanoid robots can significantly enhance performance and safety in dynamic environments, outperforming traditional methods even at low update rates.
High-level humanoid planners often output sparse task-space, low-rate trajectories, whereas whole-body controllers run at high frequency. This creates temporal asynchrony between the planning and execution, and structural incompleteness for full-body control. We propose an asynchronous upper body task-space tracking framework for humanoids. A student policy is initialized by teacher-student distillation, conditioned on the full cached future trajectory and an execution-time index, and trained with a sliding-window global reward to reduce frame drift without explicit frame estimation. For task-specific post-training, an MPC module completes sparse references into floating-base and upper-body guidance, while action- and FK level self-guidance constrain policy drift. Simulation and Unitree G1 hardware experiments show improved tracking under low update rates, stronger performance than synchronous and decoupled baselines, and safer adaptation to out-of-distribution motions.
