leap-hand

LEAP Hand in-hand rotation in mjlab: cube yaw, SO(3) reorientation and palm-down fingertip rotation of 22 objects, trained with PPO and SuperMimic-style world-model RL. Checkpoints and real-robot videos.

Contact-grounded SE(3)-equivariant dexterous grasp generative flows

simple program to move robotic hand Lynxmotion AL5D with human hands through Leap Motion sensor

A reinforcement learning pipeline for dexterous grasping with the 16-DOF LEAP Hand in MuJoCo. Features a custom Gymnasium environment, PPO training via Stable Baselines 3, and a documented reward engineering curriculum achieving an 85% success rate on cube-lifting tasks.

Dexterous pick-and-place with a Franka + LEAP hand in MuJoCo: hardcoded control vs. a local VLM (Qwen2.5-VL) policy.

ROS 2 (Jazzy) deployment pipeline for a LEAP Hand dexterous grasping policy trained via PPO. Features modular node architecture, SB3 inference, and high-fidelity RViz2 visualization.