leggedrobotics/rsl_rl

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

2916 stars · 666 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

RSL-RL is a GPU-accelerated, lightweight reinforcement learning library for robotics research, featuring algorithms like PPO and Student-Teacher Distillation. It is designed for rapid prototyping with a minimal, readable codebase and supports multi-GPU training, and is used by Isaac Lab, Legged Gym, mjLab, and MuJoCo Playground.

Use cases

• train RL policies for legged robots
• run PPO training on GPU for robotics simulations
• perform student-teacher distillation for robot learning
• prototype new robot learning algorithms quickly
• train policies in Isaac Lab or Legged Gym
• scale robot learning training across multiple GPUs

When to choose

• doing robotics RL research with Isaac Lab, Legged Gym, mjLab, or MuJoCo Playground
• you want a minimal, hackable RL codebase instead of a large framework
• you need multi-GPU, high-throughput PPO training for simulated robots

When to avoid

• you need general-purpose RL beyond robotics (e.g., Atari, game benchmarks)
• you need a broad algorithm zoo rather than PPO-centric methods
• your project does not use GPU-accelerated simulation environments

Facets

library · maturity active

reinforcement-learning machine-learning gpu-computing simulation robotics reinforcement-learning machine-learning simulation python cross-platform ppo student-teacher-distillation legged-robots isaac-lab mujoco robot-learning multi-gpu gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="leggedrobotics/rsl_rl")