Launch with MuJoCo

Overview

This package enables controlling Doosan robots in the MuJoCo physics simulator using ROS 2. You can simulate various robot models in custom environments and interact with them using ROS 2 interfaces.

Supported Models:

H-SERIES: h2515 (blue), h2017 (blue)
M-SERIES: m1013 (white), m0609 (white), m1509 (white), m0617 (white)
A-SERIES: a0509 (blue), a0912 (blue)
E-SERIES: e0509 (white)

Prerequisites

Install the MuJoCo ROS 2 control plugin. It installs MuJoCo through mujoco_vendor, so no separate MuJoCo download or MUJOCO_DIR setting is needed.

sudo apt update
sudo apt install ros-jazzy-mujoco-ros2-control

Alternatively, install it with the other workspace dependencies through rosdep:

cd ~/ros2_ws
rosdep install --from-paths src --ignore-src -y

dsr_bringup2_mujoco.launch.py includes the launch file of dsr_mujoco, so build dsr_mujoco in your workspace as well.

Command

ros2 launch dsr_bringup2 dsr_bringup2_mujoco.launch.py [arguments]

Arguments

model: Name of the Doosan robot model. (Default: m1013)
gui: Launch RViz2 GUI. (true/false, Default: true)
gripper: Gripper model to attach, for example 2f85. (Default: none)
scene_path: Path to a MuJoCo XML scene file, relative to dsr_description2/mujoco_models/, for example demo/slope_demo_scene.xml. (Default: none)
headless: Run MuJoCo without the viewer window. (true/false, Default: false)
mj: Start the MuJoCo simulation. With false only the Doosan robot side starts. (true/false, Default: true)

Example

Launch a Robot in a Blank Scene

This command launches the m1013 robot in an empty MuJoCo environment. (every settings in default)

ros2 launch dsr_bringup2 dsr_bringup2_mujoco.launch.py

In a new terminal you can run command to move the robot.

ros2 run dsr_example dance

Pick & Place Demo

This example launches the m1013 robot with a gripper in a pre-configured scene and runs a pick-and-place demonstration.

Launch m1013 with gripper in the demo scene:

ros2 launch dsr_bringup2 dsr_bringup2_mujoco.launch.py model:=m1013 gripper:=2f85 scene_path:=demo/slope_demo_scene.xml

In a new terminal, run the Pick & Place script:

ros2 run dsr_example slope_demo

Customizing the MuJoCo Environment

You can create custom simulation environments by editing or creating new MuJoCo scene (.xml) files. Scene files are located in dsr_description2/mujoco_models/. In these files, you can add or modify elements such as floor, lighting, and objects.

Key MuJoCo XML tags for customization include:

<Compiler>: Defines global simulation settings such as calculation methods. The meshdir attribute is critical for locating model asset files.

<compiler angle="radian" meshdir="../" />

Note

For the robot model to load successfully, meshdir must point to the mujoco_models directory.

<Option>: Fine-tunes physics properties like contact models and integrators, which is important for stable grasping.

<option cone="elliptic" impratio="200" integrator="implicitfast"/>

For detailed guidance on creating scenes, refer to the official MuJoCo XML documentation.

Gripper Operation

Unlike the main DSR manipulators, grippers should use individual ROS 2 controllers.

2F85 Gripper Example

The 2F85 gripper uses the forward_command_controller/ForwardCommandController, which is a standard ROS 2 controller.

This controller manages the left_driver_joint of the gripper by /{ns}/mj/left_knuckle_position_controller/commands topics.

You can find an example of its usage in the dsr_example2/dsr_example/dsr_example/demo/slope_demo.py script.

To control the gripper from the terminal, publish a Float64MultiArray message to the /dsr01/mj/left_knuckle_position_controller/commands topic.

Open Gripper:

ros2 topic pub /dsr01/mj/left_knuckle_position_controller/commands std_msgs/msg/Float64MultiArray "{data: [0.0]}"

Close Gripper:

ros2 topic pub /dsr01/mj/left_knuckle_position_controller/commands std_msgs/msg/Float64MultiArray "{data: [0.8]}"