TA-HPC-Scripts

This repository contains codes used for training models of Deeplanding autoland system. Make sure you've already trimmed your data (as written here) before you train your models with it using this codes.

Make sure you're working in TA-HPC-Scripts folder before proceeding.

$ cd TA-HPC-Scripts/

Setting up your data

1. If it didn't already exist, create Data/Processed/Selected/ directory inside TA-HPC-Scripts/. Your directory structure should look like this:

$ deeplanding
$ ├── ...
$ └── TA-HPC-Scripts
$     └── Data
$         └── Processed
$             └── Selected
$ ├── ...
$ ...

2. Now, copy all the trimmed data into Selected folder.
3. If you want to train longitudinal models put all files inside deeplanding/xpcclient/Trimmed.lon into TA-HPC-Scripts/Data/Processed/Selected/. And if you want to train aileron model, replace it with files from deeplanding/xpcclient/Trimmed.ail.

Train your models

First, we will train the longitudinal models, it contains elevator model, throttle model, selected airspeed model, and flap model.

1. Make sure you're working in TA-HPC-Scripts folder.

$ cd TA-HPC-Scripts/

2. Train the models one-by-one. Make sure to remember the order in which you train the models, because we need to rename the folder containing our models later.

## Bellow are the command to train each model

# 1. Elevator model
$ ./hpcs --mid e_simp_1
$ ./hpcs --from trainready --until post

# 2. Throttle model
$ ./hpcs --mid t_simp_1
$ ./hpcs --from trainready --until post

# 3. Selected airspeed model
$ ./hpcs --mid s_simp_1
$ ./hpcs --from trainready --until post

# 4. Flap model
$ ./hpcs --mid f_simp_1
$ ./hpcs --from trainready --until post

3. Rename each respective models' folder into into Elevator, Throttle, SelectedAirspeed, and Flap. And then copy those folder into deeplanding/xpcclient/Models/.

Now that we finished training longitudinal models, next we will train the aileron model.

4. Rename TA-HPC-Scripts/Data/Processed/ into TA-HPC-Scripts/Data/Processed.lon (so we can accessed the data inside it later).
5. Copy all files from deeplanding/xpcclient/Trimmed.ail to TA-HPC-Scripts/Data/Processed/Selected/.
6. Train the aileron model

# Train aileron model with the following command
$ ./hpcs --mid a_simp_1
$ ./hpcs --from trainready --until post

7. Rename the model's folder to Aileron and then copy that folder to deeplanding/xpcclient/Models/.

Check models' performance on training data

You can also test the models's prediction on training data by checking the compare_model.ipynb notebook.