OpenMOSS-Team/FRoM-W1-Datasets

FRoM-W1是一个开源框架，旨在通过自然语言实现通用人形机器人全身运动控制。框架分为两个阶段：H-GPT（利用大量人类数据训练的大规模语言驱动的人类全身运动生成模型）和H-ACT（通过物理模拟中的强化学习预训练和微调的运动控制器，使人形机器人能够准确稳定地执行相应动作）。该框架已在Unitree H1和G1机器人上进行了广泛评估，展示了在模拟和现实环境中成功的语言到运动生成和稳定执行。

FRoM-W1 is an open-source framework designed to achieve general humanoid whole-body motion control using natural language. The framework operates in two stages: H-GPT (a large-scale language-driven human whole-body motion generation model trained using massive human data) and H-ACT (a motion controller pretrained and fine-tuned through reinforcement learning in physical simulation, enabling humanoid robots to accurately and stably perform corresponding actions). The framework has been extensively evaluated on the Unitree H1 and G1 robots, demonstrating successful language-to-motion generation and stable execution in both simulation and real-world settings.