mujoco-warp

Isaac Lab API, powered by MuJoCo-Warp, for RL and robotics research

GPU-optimized version of the MuJoCo physics simulator, designed for NVIDIA hardware.

Train a Unitree G1 humanoid to perform a double spin kick using mjlab

One policy. Any motion. Train universal whole-body tracking in mjlab—swap clips, not checkpoints.

Integrating a custom robot (ANYmal C) with mjlab's velocity task

mjlab reinforcement learning for the BDX-R robot

Implementation of Model Tensor Planning in JAX, TMLR 2025 & ICLR 2026.

VinRobotics RL training pipeline for high-loaded humanoids.

Predictive Style Matching (PSM), is method in which an offline predictor maps the robot’s lower-body state history and velocity commands to interpretable upper-body joint and gait targets that shape the rewards during training.

GPU-batched differential IK on MuJoCo Warp with a Mink-compatible API

🤖 Teach the Unitree G1 humanoid to perform a double spin kick using mjlab, with training data, pretrained models, and deployment instructions.

Train an RL agent to catch a descending rocket with mechanical chopstick arms, Mechazilla style.

Talk to a Unitree G1 humanoid inside live GPU physics: NVIDIA Newton + MuJoCo Warp, RL locomotion trained in Isaac Lab, natural-language commands in English, Spanish and Japanese.

Compare MuJoCo, Warp, Blender, ovrtx, and Mitsuba on one deterministic Franka FR3 scene

Two humanoid robots wrestle to push each other out of a ring. MuJoCo + MuJoCo Warp, TorchRL PPO, self-play.