Agilex Robotics' Piper arm gets MuJoCo + ROS2 simulation workflow for embodied AI

A new open-source project from Agilex Robotics (hosted on GitHub by yanyuze1) provides a complete MuJoCo simulation environment for the Piper robotic arm, tightly integrated with ROS2 for control testing. The simulation foundation uses MuJoCo 3.3.0 as the physics engine and is designed for Ubuntu 22.04 with ROS2 Humble and C++17. It offers two build modes: one for launching the simulation standalone, and another that enables a ROS2 control plugin, allowing developers to send joint commands and test control policies directly in simulation.

The repository includes a detailed file structure, with assets like piper_body.xml and scene.xml, a CMake-based build system, and sample configurations. Users can clone the workspace, build dependencies, and run the simulation with a simple command like `./agilex_mujoco -r piper -s scene.xml`. The project is a practical tool for embodied AI research, enabling rapid iteration on robotic arm control without physical hardware. By combining MuJoCo's fast simulation with ROS2's modular architecture, it lowers the barrier for testing reinforcement learning, motion planning, and teleoperation pipelines.