Welcome to ROSMASTER M3 PRO repository
4. Map mapping markers
Map Mapping Markers
Map Mapping Markers1. Course Content2. Preparation2.1 Map Preparation2.2 Start the MicroROS Chassis Agent2.3 Start the Road Network Container2.4 Start Odometry, tf, Robotic Arm Assistant, Camera, Aritag Detection Nodes2.5 Start the MCP Service2.6 Start Road Network Navigation3. Record Map Mapping
1. Course Content
[!IMPORTANT]
- This section is a comprehensive course that demonstrates and explains the use and operation steps of the sand table map. For detailed parameter debugging and source code explanation, refer to the prerequisite course: [05-OpenClaw SLAM Mapping Navigation]
- Mark map mapping for target locations that need to be used in the sand table map.
2. Preparation
2.1 Map Preparation
- First, complete the construction of the sand table map and pose map according to the [Build Sand Table Map] course.
2.2 Start the MicroROS Chassis Agent
- Skip if already started
sh start_agent.sh
2.3 Start the Road Network Container
xxxxxxxxxxroadnet start
2.4 Start Odometry, tf, Robotic Arm Assistant, Camera, Aritag Detection Nodes
xxxxxxxxxxros2 launch m3pro_bringup car_base.launch.py relocation:=True2.5 Start the MCP Service
xxxxxxxxxxros2 launch m3pro_bringup mcp_service.launch.py

2.6 Start Road Network Navigation
- Start inside the roadnet container
xxxxxxxxxxros2 launch road_net_route roadnet_nav2.launch.py rviz_config_file:=/root/volumes/code_ws/src/road_net_route/rviz/nav2_relocaton.rviz
3. Record Map Mapping
- The layout and area naming of the map site are arbitrary and can be modified according to needs. The following reference area naming cases refer to:
- Shelf 1
- Shelf 2
- Parking lot 1
- Cargo warehouse
- Cargo sorting area
- Shelf 3
- Shipping area
- Garbage sorting area
- Trash bin locations: including kitchen waste, recyclable waste, other waste, hazardous waste

- Move the robot to the location that needs to be recorded, and use the robot_control CLI tool to record the map mapping.
xxxxxxxxxxrobot_control call-tool RecordMapLocation --name xxx --symbol x
Important: Before recording a location, always check whether the laser contour matches the environmental edge features. If there is a deviation, use the 2D Estimate tool for manual relocation before recording the position.
Recording example:
- Record the shipping area. Move the robot to the location near the shipping area that needs to be recorded.

- Ensure that the cargo in the area is fully visible in the field of view.


- Use the robot_control CLI tool to record the position of the robot's base_link in the map coordinate system.
xxxxxxxxxxrobot_control call-tool RecordMapLocation --name 发货区 --symbol B
- --name specifies the name of the recording point
- --symbol specifies the symbol of the recording point, used by OpenClaw for identification and differentiation.

- The positions of different types of trash bins need to be recorded separately. Place the robot in front of the trash bin to be recorded, then perform the recording.

- Check the map mapping file map_mapping.yaml. The file path is:
xxxxxxxxxx$HOME/M3Pro_ws/src/m3pro_bringup/config/map_mapping.yaml
- Check whether all required locations have been recorded.
