LATENT paper presents humanoid tennis skill learning from human motion data
Evidence notes
- The paper introduces LATENT, a method for learning athletic humanoid tennis skills from imperfect human motion data
- The official implementation is presented by a Tsinghua University and Galbot team
- The released codebase uses G1 tracking and evaluation tasks for the humanoid robot setup
Company context
Develops embodied AI robotics systems focused on general-purpose manipulation and real-world deployment. The company builds integrated hardware platforms and vision-language-action models to enable robots to operate autonomously across commercial and industrial environments.