LeggedGym-Ex#

GitHub Repo stars

A legged_gym based framework for training legged robots in Genesis, IsaacGym and IsaacSim.

Features#

Totally based on legged_gym

This framework keeps most APIs and conventions consistent with legged_gym, providing good readability and better control over training pipeline.
Integration of multiple simulators

We support training in either of three simulators: IsaacGym, Genesis and IsaacSim.

Quick tips for choosing among three simulators:
- Faster training but worse rendering -> IsaacGym
- Both training speed and support for fluid, soft materials -> Genesis
- More realistic rendering at the cost of training speed -> IsaacSim.

Incorporation of various methods in published RL papers

Method	Paper Link	Code
Periodic Gait Reward	Sim-to-Real Learning of All Common Bipedal Gaits via Periodic Reward Composition	go2_wtw
Walk These Ways	Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior	go2_wtw
System Identification	Learning Agile Bipedal Motions on a Quadrupedal Robot	go2_sysid
Teacher-Student	Rapid Locomotion via Reinforcement Learning	go2_ts
Explicit Estimator	Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion	go2_ee
Constraints as Terminations	CaT: Constraints as Terminations for Legged Locomotion Reinforcement Learning	go2_cat
DreamWaQ	DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning	go2_dreamwaq
SPO (Simple Policy Optimization)	Simple Policy Optimization	`legged_robot_config.py`
CTS (Concurrent Teacher Student)	CTS: Concurrent Teacher-Student Reinforcement Learning for Legged Locomotion	go2_cts
DeepMimic	DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills	g1_deepmimic
AMP (Adversarial Motion Priors)	AMP: Adversarial Motion Priors for Stylized Physics-Based Character Control	k1_amp