Open nlitz88 opened 2 months ago
Another very important question we will have to ask ourselves is: How do we deal with the angles provided by the pose message--as they are provided as quaternions. Maybe the script can just convert these to euler angles simply enough--but that's just one detail I want to make sure we don't overlook.
Also, I suppose my dynamics model could learn to use quaternions in theory--but I haven't messed around with that yet--so I'm not sure it'd be safe to go down that road just yet.
@just-ap Instead of the above script, @CMUBOB97 is going to just create a ROS node that subscribes to each of the above topics that we need for our model training and directly creates CSV files.
However, we WILL STILL need the script(s) to interpolate and synchronize/merge the separate topic CSV files--so that will also be something we have to work on coming up.
Okay, I guess we kind of established that whatever node is publishing the odometry estimate from the F1TENTH gym is kinda whack. I have been trying some different things to see if I can fit a model to the data, but I haven't had any luck so far.
Maybe there's a chance we can debug this in time, but if not, we may just have to can it.
We will then need a script to convert the topic data recorded from these into CSV format. Could maybe get away with just adapting Haoru's script and adding new adapters--that might be the most straightforward way.
Additionally, we will absolutely need the two topics that are recorded to be merged into a single CSV file--so that means we will also need each topic's data to be interpolated and then synced up (merged) somehow.