Mirte Lab Cleaner Documentation

The Mirte Master Lab Cleanup Robot is an autonomous mobile manipulator designed to identify, retrieve, and sort objects in a laboratory environment without human intervention. The system combines frontier-based exploration, coverage path planning, 3D object detection, and colour-based classification to fully automate the lab cleaning task.

The robot platform is the Mirte Master, a mechanum-drive mobile robot with a 4-DOF arm and 5-bar-linkage gripper. It uses an gripper-mounted Orbbec Astra depth camera for 3D perception and 2D classification. Navigation is handled by the ROS 2 Nav2 stack, motion planning by MoveIt 2, and task coordination by a py_trees behaviour tree.

---

Package Documentation

This is the top-level documentation for the mirte_lc ROS 2 project. Each ROS 2 package has its own API documentation:

Package	Description
mirte_lc_gazebo	Gazebo simulation worlds and launch files
mirte_lc_labclean	Main behaviour tree and lab cleaning logic
mirte_lc_moveit_cpp	MoveIt C++ action server for arm control
mirte_lc_msgs	Custom ROS 2 message, service, and action definitions
mirte_lc_nav2	Nav2 coverage navigation and frontier-based exploration
mirte_lc_vision	Object detection and 3D localisation via YOLO + point cloud

The system is composed of eleven ROS 2 nodes that communicate through topics, services, and actions. The nodes are organised into four functional layers: navigation, manipulation, perception, and task coordination.

Key interfaces between nodes:

Interface	Type	Kind	Between
/perception/depth/detected_objects	mirte_lc_msgs/DetectedObjectArray	Topic	object_locator → labclean_tree
/labclean_navigator/coverage	mirte_lc_msgs/NavigateCoverage	Action	labclean_tree → labclean_action_server
/labclean_navigator/set_state	mirte_lc_msgs/ServeCoverageStatus	Service	labclean_tree → labclean_action_server
/move_to_position	mirte_lc_msgs/MoveToPosition	Action	labclean_tree → mirte_lc_moveit_action_server
/navigate_to_pose	nav2_msgs/NavigateToPose	Action	labclean_tree, labclean_action_server → Nav2
/navigate_through_poses	nav2_msgs/NavigateThroughPoses	Action	labclean_action_server → Nav2
/perception/planar/get_detected_objects	mirte_lc_msgs/GetDetectedObjects	Service	labclean_tree → yolo_detector
/global_costmap/costmap	nav_msgs/OccupancyGrid	Topic	Nav2 → object_locator, labclean_action_server
/io/power/power_watcher	battery msg	Topic	Hardware → labclean_tree
/explore/status	explore_lite_msgs/ExploreStatus	Topic	explore_server → labclean_tree

---

Quickstart

Clone and build the full workspace:

cd ~/ros2_ws/src
git clone https://github.com/matt-rbt/Mirte_Lab_Clean
cd ~/ros2_ws
vcs import src/ < src/mirte_lc/sources.repos
cd src/mirte-ros-packages && git submodule update --init --recursive && cd ../..
rosdep install -y --from-paths src/ --ignore-src --rosdistro humble
colcon build --symlink-install

Launch the full stack on the real robot:

ros2 launch mirte_lc_labclean labclean_bringup.launch.py use_sim_time:=false

Launch in simulation:

# Terminal 1 — Gazebo
ros2 launch mirte_lc_gazebo gazebo_mirte_lc.launch.py

# Terminal 2 — full stack
ros2 launch mirte_lc_labclean labclean_bringup.launch.py use_sim_time:=true

See quickstart for the full installation guide including robot-only sparse checkout, clock synchronisation, and visualisation setup.

ROS 2 requires the clocks on the robot and the development laptop to be synchronised to within ~1 ms for TF and sensor fusion to work correctly. Chrony is used to sync the laptop to the robot:

On the robot — add to /etc/chrony/chrony.conf:

local stratum 8
allow 192.168.178.0/24
allow 192.168.42.0/24

On the laptop — add to /etc/chrony/chrony.conf:

server 192.168.42.148 iburst prefer
server 192.168.178.38 iburst
makestep 1.0 3

Verify synchronisation:

chronyc sources -v
# 192.168.x.x should show ^* (current best source)

---