After running python dagger_training.py --settings_file=config/dagger_settings.yaml an exception occurs and it keeps printing Crashing into something!

[Nuyoah0123]

I tried to follow the tutorial in the Collect your own dataset section of Readme to collect my own data, I changed the resume_training parameter in the dagger_settings.yaml file to False and ran the dagger_training.py file, but it ran The result doesn't seem to be satisfactory.

As shown in the picture, the drone started after a short flight according to the reference trajectory, spinning in place and getting stuck in the middle of an obstacle.
I would like to know why this is happening and how I should fix the problem?

[Nuyoah0123]

I plotted the xyz data from the ellipsoid_trajectory.csv file, the odometry.csv file, and the reference_trajectory.csv file, and the results showed the following

[sdhar16]

Change the fallback radius to 0 during data collection when the network is initialised from scratch. Its in the dagger settings