1. Project Description

This repository is the foundational ROS2-Gazebo simulation project for Chapter 1 of the Unitree robot dog series. The repository will be updated from time to time as the full project series progresses. You can check the current project plan on my Feishu page: Feishu Project Page. This is expected to be a very long-term project.

2. How to Use the Project

2.1 Build and Run Directly

Go to the workspace:

cd go2_sim_ws
colcon build
source install/setup.bash

After sourcing, the gazebo_sim package extends Gazebo model search paths (GZ_SIM_RESOURCE_PATH / IGN_GAZEBO_RESOURCE_PATH) and sets CycloneDDS CYCLONEDDS_URI to the XML shipped with the package (under share/gazebo_sim/dds/), avoiding broken file://~/... URIs where ~ is not expanded. To use your own DDS config, run unset CYCLONEDDS_URI after sourcing or export an absolute file:///... path.

Launch the simulation. The first launch may take longer because required world resources need to be downloaded:

ros2 launch gazebo_sim launch.py

After a successful launch, you should see something like this:

This project includes:

• A front monocular camera to simulate the GO2 front camera
• Two LiDAR sensors:
  • A front L1 LiDAR to simulate GO2’s built-in LiDAR
  • An external VLP16 LiDAR

Both LiDARs can publish LaserScan and PointCloud2 data simultaneously, making mapping and navigation workflows easier. A D435i camera has also been added.

Use keyboard teleoperation:

cd go2_sim_ws
source install/local_setup.bash
ros2 run teleop_twist_keyboard teleop_twist_keyboard --ros-args -r /cmd_vel:=/robot1/cmd_vel

After running the command above, you can control the robot dog with your keyboard. You can also use the following service to control behavior. Supported commands are walk, up, and sit.

ros2 service call /robot1/robot_behavior_command quadropted_msgs/srv/RobotBehaviorCommand "{command: 'walk'}"

2.2 Using Docker

Before using Docker, please read the Docker setup guide briefly, since environments may differ. In my setup, I use mounted volumes.

The following commands cover the main Docker workflow:

cd go2_sim_ws/src/docker
docker compose up -d --build --remove-orphans        # Build containers
docker compose up -d go2_sim                        # Start in detached mode; go2_sim is the compose service name
docker compose ps                                    # List containers
docker compose exec go2_sim bash                    # Enter container
docker compose down                                  # Remove containers

Inside the Docker container, follow the same steps as direct local usage:

colcon build
source install/setup.bash

Launch the simulation. First launch may take longer due to resource downloads:

ros2 launch gazebo_sim launch.py # Launch GO2 without extended sensors
ros2 launch gazebo_sim launch.py sensors:=true world:=warehouse.sdf # Launch GO2 with extended sensors in selected map

After successful startup, you should see:

As above, the project includes a front monocular camera and two LiDARs (GO2 front L1 + external VLP16), both of which can publish LaserScan and PointCloud2. This supports later mapping/navigation tasks. A D435i camera is also included.

cd go2_sim_ws
source install/local_setup.bash
ros2 run teleop_twist_keyboard teleop_twist_keyboard --ros-args -r /cmd_vel:=/robot1/cmd_vel

Once the command runs, you can control the robot dog with keyboard input. Behavior services support walk, up, and sit.

ros2 service call /robot1/robot_behavior_command quadropted_msgs/srv/RobotBehaviorCommand "{command: 'walk'}"

2.3 Mapping and Navigation

Multiple terminals are required. ros2 launch ... blocks the shell until you stop it with Ctrl+C. For mapping you must run Gazebo, Cartographer, and teleop at the same time, so use separate terminals (or tmux/screen panes). Do not paste all commands into one terminal in sequence—the first launch will block and the rest will never run.

In every new terminal, load the workspace:

cd go2_sim_ws
source install/setup.bash

Mapping (Cartographer SLAM)

1. Terminal 1 — Gazebo (extended sensors, warehouse world):

ros2 launch gazebo_sim launch.py sensors:=true world:=warehouse.sdf

2. Terminal 2 — Cartographer + RViz:

ros2 launch cartographer go2_cartographer.launch.py

3. Terminal 3 — Keyboard teleop (strongly recommended) so the robot explores; otherwise the map will stay poor or incomplete.

ros2 run teleop_twist_keyboard teleop_twist_keyboard --ros-args -r /cmd_vel:=/robot1/cmd_vel

4. Terminal 4 — Save the occupancy grid (run when you are happy with the map; writes warehouse_map.yaml and warehouse_map.pgm in the current directory):

cd go2_sim_ws
source install/setup.bash
mkdir -p maps_out && cd maps_out
ros2 run nav2_map_server map_saver_cli -t map -f warehouse_map

If you open a fresh terminal elsewhere, cd into go2_sim_ws, source install/setup.bash, then cd to the save directory and run map_saver_cli. For Nav2, pass the absolute path to warehouse_map.yaml as map:= (from maps_out, run realpath warehouse_map.yaml).

Navigation (Nav2)

When using the bundled warehouse map, keep the same world as for mapping (warehouse.sdf).

1. Terminal 1 — Gazebo:

ros2 launch gazebo_sim launch.py sensors:=true world:=warehouse.sdf

2. Terminal 2 — Nav2 (default map: installed navigation2/maps/warehouse_map.yaml):

ros2 launch navigation2 go2_navigation2.launch.py

To use a map you saved yourself:

ros2 launch navigation2 go2_navigation2.launch.py map:=/absolute/path/to/go2_sim_ws/maps_out/warehouse_map.yaml

Set the initial pose and goal in RViz; Nav2 plans and executes—keyboard teleop is not required as the primary interface.