7. ORB_SLAM2 Octomap Map Building

This example is implemented on Orin NX, and cannot be implemented on a virtual machine. It just explains how the process is implemented. If you want to implement this function on your own motherboard, you need to compile the entire function package and connect related peripherals.

octomap official website: http://octomap.github.io/

octomap source code: https://github.com/OctoMap/octomap

octomap wiki: http://wiki.ros.org/octomap

octomap_server: http://wiki.ros.org/octomap_server

The operating environment and reference configuration of software and hardware are as follows:

• Reference vehicle: ROSMASTER X3

• Robot hardware configuration: Arm series master control, Silan A1 laser radar, depth camera

• Robot system: Ubuntu 20.04

• PC virtual machine: Ubuntu (20.04) + ROS2 (Foxy)

• Usage scenario: Use on a relatively clean 2D plane

7.1, Introduction

Octomap is an octree-based 3D map creation tool that can display complete 3D graphics including barrier-free areas and unmapped areas, and sensor data based on occupancy grids can be fused and updated in multiple measurements; Maps are available in multiple resolutions, data can be compressed, and storage is compact. In fact, the code of octomap mainly consists of two modules: octomap, a 3D map creation tool, and octovis, a visualization tool. Compared with point clouds, it can save a lot of space. The map created by octomap looks like this: (different resolutions from left to right)

7.2 , octomap map based on camera

Execute the following launch file in the terminal:

1. Start the camera

xxxxxxxxxx
#gemini camera startup
ros2 launch astra_camera gemini.launch.xml

2. Start octomap_server map

xxxxxxxxxx
ros2 launch yahboomcar_slam camera_octomap_launch.py

3. Start rviz in docker or virtual machine [recommended]:

xxxxxxxxxx
ros2 launch yahboomcar_slam display_octomap_launch.py

4. Use the remote control or keyboard to control the node to slowly move the robot to build the map. Losing key frames may cause the map to fail.

7.3, Octomap construction based on orbslam and pointcloud_mapping

1, Start the camera

xxxxxxxxxx
#gemini camera startup
ros2 launch astra_camera gemini.launch.xml

2, Start orbslam to publish the camera pose, color map and depth map. Depending on the performance of different master controllers, the waiting time here is about 10 seconds

xxxxxxxxxx
ros2 launch yahboomcar_slam orbslam_base_launch.py

3, Start octomap_server to build a map

xxxxxxxxxx
ros2 launch yahboomcar_slam orbslam_pcl_octomap_launch.py 

4. Start rviz:

xxxxxxxxxx
ros2 launch yahboomcar_slam display_octomap_launch.py

5. Use the remote control or keyboard to control the node to slowly move the robot to build a map. Losing key frames may cause the map to fail.

7.4、Node analysis

7.3.1、Display calculation graph

xxxxxxxxxx
rqt_graph

7.3.2、Details of each node

7.3.3、TF transformation

xxxxxxxxxx
ros2 run tf2_tools view_frames.py