This repository contains the code that is used for the training of a TCN as part of the project
Autonomous Docking of Multi-Rotor UAVs on Blimps under the Influence of Wind Gusts
by Pascal Goldschmid and Aamir Ahmad.
The main project repository can be found here (UPDATE LINK).
If you have any questions, comments or suggestions please contact pascal.goldschmid@ifr.uni-stuttgart.de.
When generating the dataset used for the training of the TCN as described here (UPDATE LINK) a file is created that contains the limit values that are used to perform the normalization of the features and labels in the dataset during the training. These values need to be copied to the config file that defines the training parameters. The config file is a .yaml file and should be stored in the directory config. Furthermore, in order to run a training also the parameters in the sections dataset_info and training need to be adapted so that they mirror the values used for the generation of the dataset.
In order to launch a training, run the script python3 time_series_train.py.
The trained model is stored in the folder model.
The scripts TCN_pred_time_series_analysis_multiple.py and TCN_pred_time_series_analysis_plot.py can be used to evaluate the performance of the trained TCN. The former calculates several metrics to compare the performance to a constant velocity model whereas the latter produces a plot illustrating the TCN prediction of the airhsip velocity.