Delivbot provides a complete ROS 2 workspace package for simulating and running the Delivbot mobile base. It bundles URDF, meshes, Gazebo simulation assets, and launch files for both simulation and real hardware bring-up.

• Unified robot.launch.py for real-robot bring-up (robot_state_publisher, lidar, RealSense, ODrive, joystick teleop, twist_mux).
• Modular launch files for each hardware subsystem: RealSense, Hesai lidar, and ODrive hoverboard driver.
• Gazebo simulation launch with twist mux, joystick teleop, ros2_control, and ROS ↔︎ Gazebo bridges.
• Ready-to-use URDF/Xacro, configuration, and RViz profiles for inspecting the robot.

# From the workspace root (e.g., ~/ros2_ws)
colcon build --packages-select delivbot --symlink-install
source install/setup.bash

Launch the full Gazebo simulation stack:

ros2 launch delivbot gazebo.launch.py

This starts Gazebo, spawns Delivbot, loads ros2_control controllers, the joystick teleop pipeline, and bridges available sensors to ROS topics.

For a URDF-only visualization in RViz:

ros2 launch delivbot display.launch.py

• Set joystick to D-Mode
• Enable/drive with left trigger (button 6) held.
• Turbo speed with right trigger (button 7).

The default SLAM stack now wraps lidarslam_ros2, which performs scan-matching SLAM with optional IMU fusion and a graph-based loop-closure back-end.

Prerequisites:

• Clone lidarslam_ros2 into your workspace so the packages scanmatcher, graph_based_slam, lidarslam, and lidarslam_msgs build alongside this robot package.
• Provide a point cloud on /lidar/points (the Gazebo bridge already publishes this) and optionally an IMU on /imu (see imu.launch.py and the Gazebo IMU sensor).

Start SLAM:

ros2 launch delivbot slam.launch.py

Key launch arguments (override as needed):

• pointcloud_topic (default /lidar/points)
• imu_topic (default /imu)
• use_imu (true to fuse IMU, false to ignore it)
• use_odom (set to true to fuse wheel odometry when available; defaults to false so SLAM still works without odom)
• initial_pose_topic (default initial_pose if you have a PoseStamped publisher)
• params_file (defaults to config/lidarslam_params.yaml)
• auto_initial_pose (true publishes a PoseStamped automatically; disable if you want to set it manually)
• initial_pose_{x,y,z,qx,qy,qz,qw,delay} (override the auto-published pose)

RViz quickstart:

• Fixed frame: map
• Add Pose display for /current_pose, Path for /path, and PointCloud2 for /map (global accumulated cloud).
• Loop-closed results appear on /modified_path and /modified_map.

To produce a 2D occupancy grid suitable for Nav2, run the projection pipeline:

ros2 launch delivbot grid_projection.launch.py

This wraps octomap_server to consume the /map point cloud and publish /map (occupancy grid) plus /octomap_* topics. Adjust parameters in config/octomap_params.yaml if you need different z-bounds or resolution.

The core SLAM tuning lives in config/lidarslam_params.yaml. Modify voxel sizes, registration method (NDT/GICP), scan ranges, and loop-closure thresholds there as you iterate.

robot.launch.py runs the physical robot stack while keeping Gazebo assets untouched. It loads the URDF, publishes TFs, launches joystick teleop, and can selectively start the hardware drivers.

ros2 launch delivbot robot.launch.py \
  launch_lidar:=true \
  launch_realsense:=true \
  launch_odrive:=true \
  launch_joystick:=true \
  cmd_vel_topic:=/cmd_vel

Joystick commands flow through twist_mux so /cmd_vel_joy and navigation topics are arbitrated automatically before reaching the ODrive driver (default /cmd_vel).

• Set joystick to X-Mode
• Enable/drive with left shoulder (button 4) held.
• Turbo speed with right shoulder (button 5).

Each subsystem launch file mirrors the ros2 run commands you would call manually, adding ergonomic launch arguments for consistent configuration.

Ensure the following packages/drivers are available in your workspace or installed on the system:

• ros_gz_sim, ros_gz_bridge, ros_gz_image
• hesai_ros_driver
• realsense2_camera
• odrive_hoverboard
• joy, teleop_twist_joy
• twist_mux
• scanmatcher, graph_based_slam, lidarslam, lidarslam_msgs (from lidarslam_ros2)

Refer to each dependency’s documentation for hardware setup (e.g., USB permissions for RealSense cameras and ODrive).

1. Simulation test: Verify navigation stack interaction and teleoperation via gazebo.launch.py.
2. Hardware dry-run: Connect the robot and start robot.launch.py, disabling specific drivers if they are not attached yet (e.g., launch_realsense:=false).
3. Parameter tuning: Provide YAML configuration files through the launch arguments as you refine driver settings (e.g., wheel parameters for the ODrive driver).
4. Monitoring: Use rqt_graph, rqt_tf_tree, and RViz with config/display.rviz to visualize topics and transforms.

For adjustments or new integrations, update the relevant launch file or configuration under config/, add documentation here, and rebuild with colcon. Contributions should include testing steps and documentation updates.