Real-time perception of errors in constructing topological maps during motion

[AG-Col]

Thank you very much for your work and the open-source code. While running the program, I wanted to test the generation of the topological map under the condition of real-time robot perception and mapping. I made the following modifications, but the map was not correctly constructed during execution. Could you please advise on what might be the issue? I am eagerly looking forward to your response. Once again, thank you so much!

[dklee98]

Sorry for closing your issue!!
While I replying, I miss-clicked haha

[AG-Col]

Thank you very much for your attention. Is there a solution to this issue, or is there something I haven't corrected properly? I am looking forward to your response, Thank you.

[dklee98]

Hello!
Thank you for your interest in my research!

The issue you're encountering is caused by a blind spot in the robot's vicinity.
(If the graph initialization happens too quickly, this issue might occur—please check this.)

In your figure, the robot is standing in an empty space (without lidar scans).
Therefore, you should move the robot to an area with lidar data.

However, if you're using raw point clouds, the area around the robot will always appear empty.

That's why I recommend using an existing elevation map, such as (https://github.com/leggedrobotics/elevation_mapping_cupy), for a more accurate and detailed map.
(Alternatively, you can accumulate lidar scans.)

Additionally, if you want to update the graph as the robot moves, please set isUpdate to true.

Thank you for your interest, and feel free to reach out with any questions!