Issue saving the simulation state with save_total_state in ROS2

[harry-graves]

Hi, I noticed an issue when trying to save the simulation state in a ROS2 environment. After changing filepath_est, filepath_std and filepath_gt, when using the save_total_state:=true flag, I get the following error:

[run_subscribe_msckf-1] terminate called after throwing an instance of 'boost::filesystem::filesystem_error'
[run_subscribe_msckf-1]   what():  boost::filesystem::create_directories: Invalid argument
[ERROR] [run_subscribe_msckf-1]: process has died [pid 611689, exit code -6, cmd '/home/ori/ros2_ws/install/ov_msckf/lib/ov_msckf/run_subscribe_msckf --ros-args -r __ns:=/ov_msckf --params-file /tmp/launch_params_f1udo1j0 --params-file /tmp/launch_params_8cgjp7ug --params-file /tmp/launch_params__47f7d76 --params-file /tmp/launch_params_vj12ovz1 --params-file /tmp/launch_params_7y3x4w5v'].

However I noticed this can be worked around by manually replacing "state_estimate.txt", "state_deviation.txt" and "state_groundtruth.txt" with your own full filepaths in the following file:

open_vins/ov_msckf/src/ros/ROS2Visualizer.cpp

Lines 108 to 110 in 17b73cf

	std::string filepath_est = "state_estimate.txt";
	std::string filepath_std = "state_deviation.txt";
	std::string filepath_gt = "state_groundtruth.txt";

This would then give me a stderr output when rebuilding, but it worked as a temporary fix.

[diplodocuslongus]

After changing filepath_est, filepath_std and filepath_gt

Where do you change these files?
I have the same error as yours when setting save_total_state:=true (ROS2 humble) but without changing any destination file path or file name.

Your fix does the trick to save the deviation and estimate states, but the groundtruth isn't saved though.

Thanks a lot anyway.

[harry-graves]

filepath_est, filepath_std and filepath_gt can be changed in estimator_config.yaml, though whatever filepaths you put here are overwritten by the fix I have given above.

I'm not too sure what you mean by saving the groundtruth, but the above workaround allowed me to save the estimated state, which includes the trajectory, to a text file.

[diplodocuslongus]

not too sure what you mean by saving the groundtruth

I was referring to the ground truth data that is included in a dataset when running OpenVINS offline.
And I meant that when running openVINS with a pre-existing dataset (e.g. EuROC), I don't see any state_groundtruth.txt file.

Also the data format saved in the estimated state file is different from that expected by ov_eval for, e.g., trajectory alignment.