LAFAN1_Retargeting_Dataset
收藏资源简介:
LAFAN1重新定位数据集用于将人类运动捕捉数据重新定位到Unitree的人形机器人上,以使机器人运动更加自然。数据集包含了三个模型(H1, H1_2, G1)的重新定位轨迹,这些轨迹是通过基于Interaction Mesh和IK的数值优化实现的,考虑了末端执行器姿态约束以及关节位置和速度约束,以防止脚部滑动。数据集以CSV格式存储,每行对应原始运动捕捉数据的每一帧,记录了人形机器人所有关节的配置。
The LAFAN1 Retargeting Dataset is developed to retarget human motion capture data to Unitree humanoid robots, enabling the robots to exhibit more natural locomotion. The dataset includes retargeting trajectories for three robot models (H1, H1_2, G1), which are generated through numerical optimization based on Interaction Mesh and Inverse Kinematics (IK). The optimization process incorporates end-effector pose constraints, joint position constraints and joint velocity constraints to prevent foot slippage. The dataset is stored in CSV format, where each row corresponds to one frame of the original motion capture data and records the configuration of all joints of the humanoid robot.




