microduck_rl
RL training environments for Microduck (mjlab)
Discover how Microduck RL's _servo_joint_ids() helper isolates active joints, preventing accidental manipulation of physics elements and ensuring accurate reward functions in Isaac Gym simulations.
How Backlash Is Modeled in the Microduck Robot for Sim2real TransferLearn how Microduck models backlash for sim2real transfer using hinge joints, a custom encoder actuator, and environment wrappers. Improve robot control and performance.
How Microduck RL Handles NaN Values to Prevent Training CrashesLearn how Microduck RL prevents training crashes from NaN values using four mechanisms: reward sanitisation, advantage sanitisation, observation NaN policies, and nan-state termination.
Microduck RL MJCF Robot Models: Complete Guide to All 8 XML ConfigurationsExplore the 8 Microduck RL MJCF robot models for walking, ground-contact, roller-skate, and backlash variants. Access all configurations in this comprehensive guide.
How to Publish Microduck RL Policies to HuggingFace HubEasily publish Microduck RL policies to HuggingFace Hub. Export ONNX, generate manifests, and upload with a single command using uv run publish.
How to Export Microduck RL Checkpoints: A Complete Guide to ONNX ConversionExport Microduck RL checkpoints easily with the official CLI script. Convert trained models to ONNX format, embedding the observation normalizer for seamless deployment. Learn how now.
How the Observation Normalizer Is Baked Into the ONNX Export for Microduck RLLearn how Microduck RL bakes the observation normalizer into ONNX export using the EmpiricalNormalization layer for seamless model integration and improved performance.
Microduck RL Domain Randomization Strategy: How Non‑Accumulating Perturbations Enable Sim‑to‑Real TransferDiscover the Microduck RL domain randomization strategy. Learn how non-accumulating perturbations ensure effective sim-to-real transfer for robotics.
How to Train a Microduck RL Policy with Backlash Simulation: Complete GuideLearn to train a Microduck RL policy with backlash simulation. Follow our guide to select tasks, run training, and export your policy for hardware deployment. Get started today!
Episodic Trick Tasks in Microduck RL: Forward Roll, Ground Pick, and Ball Kick ExplainedExplore Microduck RL's episodic trick tasks like Forward Roll, Ground Pick, and Ball Kick. Learn about these fixed-duration maneuvers for robot learning.
Microduck RL Locomotion Tasks: Complete Task Catalog for Bipedal Robot TrainingExplore over 18 Microduck RL locomotion tasks for bipedal robot training, including velocity control, recovery, manipulation, and acrobatics. Train on flat or rough terrain.
How Backlash Is Simulated in Microduck RL Environments: A Technical Deep-DiveLearn how backlash is simulated in Microduck RL environments. Discover the three-layer system for modeling mechanical play, actuator behavior, and task configurations for accurate reinforcement learning.
Have a question about this repo?
These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:
curl -s "https://instagit.com/install.md" 
Maintain an open-source project? Get it listed too →