# fan-ziqi/rl_sar

Simulation verification and physical deployment of robot reinforcement learning algorithms, suitable for quadruped robots, wheeled robots, and humanoid robots. "sar" represents "simulation and real"

Repository: https://github.com/fan-ziqi/rl_sar
Canonical: https://ross.abutalabs.com/products/rl_sar
Language: C++
License: Apache-2.0
License Family: permissive
Last push: 2026-08-16T10:59:49+00:00

## Health v2 (maintenance only)
Score: 79/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 63, longevity 65
- inputs: {"age_days": 910, "days_push": 17, "days_rel": 167, "gap_med": 50.0, "n_releases_24m": 11}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1436, forks 188 (observed 2026-08-28T04:04:43.622298+00:00)

## What it is
A C++ framework for simulation verification and physical deployment of reinforcement learning policies for robots, supporting quadruped, wheeled, and humanoid robots. It bridges sim-to-real workflows with support for IsaacGym/IsaacSim training, Gazebo/MuJoCo simulation, and ROS/ROS2 integration.

## Use cases
- deploy a trained RL locomotion policy on a Unitree quadruped robot
- verify a reinforcement learning policy in Gazebo or MuJoCo simulation before real deployment
- simulate and deploy humanoid robot whole-body tracking policies
- run sim-to-real workflows for wheeled-legged robots
- test IsaacGym or IsaacSim trained policies on real hardware via ROS or ROS2
- run RL policies with libtorch or onnxruntime inference

## When to choose
- you need to deploy RL locomotion or dance policies on supported quadruped, wheeled, or humanoid robots
- you want a unified sim-to-real pipeline with Gazebo, MuJoCo, IsaacGym, and IsaacSim support
- you work in the ROS/ROS2 ecosystem with C++ and need on-robot policy inference
- you want pre-trained policies for popular robots like Unitree A1, Go2, B2, and G1

## When to avoid
- you only need Python-based training without deployment - use legged_gym or IsaacLab directly
- your robot is not in the supported list and you cannot adapt the hardware interface
- you need Windows support - only Linux and macOS (MuJoCo only) are supported
- you need a general-purpose robot simulator rather than an RL deployment framework

## Facets
- artifact type: framework
- maturity: active
- function: simulation, robotics, machine-learning, reinforcement-learning, deployment
- domain: robotics, reinforcement-learning, simulation, machine-learning
- platform: cpp, python
- tags: quadruped-robots, humanoid-robots, wheeled-robots, gazebo, mujoco, isaacgym, isaac-sim, ros, ros2, sim-to-real, locomotion, libtorch, onnxruntime, linux, macos

## Member repositories
- fan-ziqi/rl_sar (main) score 79

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:43.622298+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:36:47.747184+00:00, confidence not recorded.
  - readme: https://github.com/fan-ziqi/rl_sar (fetched 2026-08-28T04:04:43.622298+00:00, sha 590e2772a19f)
- Data as of 2026-08-30T08:39:29.467469+00:00.