xbpeng/MimicKit

Health v2 · maintenance only

Adoption not part of the score

2254 stars · 288 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

MimicKit is a lightweight Python framework providing a suite of motion imitation methods (DeepMimic, AMP, ASE, AWR, and others) for training physics-based motion controllers. It supports multiple simulator backends including Isaac Gym, Isaac Lab, and Newton, and is designed with minimal dependencies.

Use cases

• train physics-based humanoid motion controllers from motion capture data
• apply adversarial motion priors to robot locomotion
• learn character animation skills with reinforcement learning
• compare motion imitation algorithms like DeepMimic, AMP, and ASE
• train quadruped controllers imitating reference motions
• research motion priors for physics simulation

When to choose

• you want a clean, minimal-dependency codebase for motion imitation research
• you need to train controllers in Isaac Gym, Isaac Lab, or Newton
• you want reference implementations of multiple motion imitation methods in one framework

When to avoid

• you need a feature-rich, highly modular production framework (consider ProtoMotions)
• you work outside physics simulation or motion control domains
• you require non-NVIDIA physics simulators

Facets

framework · maturity active

reinforcement-learning simulation machine-learning animation robotics reinforcement-learning simulation graphics python motion-imitation physics-simulation character-animation isaac-gym humanoid-control motion-priors linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="xbpeng/MimicKit")