Plotting Ground Truth vs. FAST-LIO Trajectroy

[ryanpdoc57]

I was wondering what you used to plot both the ground truth and the fast-lio trajectory on the same plot? Thanks.

[ryanpdoc57]

If you are using plot juggler, can you help explain how to get the .txt file to plot the path for ground truth?

[Taeyoung96]

@ryanpdoc57
Thanks for your interest in our work!
Actually, I use evo to visualize fast-lio and ground-truth trajectories.

In the xxx.yaml file inside the config folder, set the traj_save_en parameter to true,
and append the absolute path to traj_file_path appropriately, you will get a fast-lio trajectory in TUM format.

Ground-truth utilized a txt file provided by M2DGR.
You can find more information in the evo repository.

I hope this helps.

[ryanpdoc57]

@Taeyoung96 Thank you so much! When I plot the trajectories, I am getting a slightly different starting spot (dots) which is causing more error than desired. I have also noticed that fast-lio performs much worse in bright environments (room_01). Have you tested any of the other M2DGR bag files besides the one that is posted? Do you have any tips on how to properly align the starting point?

[ryanpdoc57]

@Taeyoung96 Thank you so much! When I plot the trajectories, I am getting a slightly different starting spot (dots) which is causing more error than desired. I have also noticed that fast-lio performs much worse in bright environments (room_01). Have you tested any of the other M2DGR bag files besides the one that is posted? Do you have any tips on how to properly align the starting point?

I thought that the problem was due to my extrinsic translation parameters being incorrect, but I fixed them based on the information provided in the M2DGR read me, and I am still experiencing the same displacement between starting points

[Taeyoung96]

@ryanpdoc57
What options did you give yourself when you ran evo?
Can you tell me how you entered the commands in the terminal?

The -a option causes the evaluation to proceed towards aligning the entire trajectory, rather than focusing on the starting point.

Reducing errors requires an effort to make trajectories better. In your launch file,

filter_size_surf_param = LaunchConfiguration('filter_size_surf', default='1.0') # double
    filter_size_map_param = LaunchConfiguration('filter_size_map', default='1.0') # double

Can you try replacing the default with 0.1 instead of 1.0?
This is the parameter that corresponds to the voxel size.

Also, if you could share the full config file and launch file, it would be helpful for us to troubleshoot together.