How to use this App

Setup Python Virtual Environnement

On Windows

Open a terminal in the current repo.

Enter the following command:

py -m venv env

In the same terminal, enter the following command:

.\env\Scripts\activate.bat 

You have now activated your python virtual environnement in this terminal.

Now go to https://www.lfd.uci.edu/~gohlke/pythonlibs/ and download the numpy+mkl wheel, as well as the slycot wheel. the file names should look something like this:

numpy-1.22.4+mkl-cp310-cp310-win_amd64.whl
SciPy-1.8.1-cp310-cp310-win_amd64.whl
slycot-0.4.0-cp310-cp310-win_amd64.whl
pandas-1.4.3-cp310-cp310-win_amd64.whl

Dowload the files that correspond to the python version on your system.

Put both these files in the custom_wheels folder in this repo.

In your requirements.txt file in this repo, copy the file names of the wheel files you just downloaded. requirements.txt should look like this:

custom_wheels/numpy-1.22.4+mkl-cp310-cp310-win_amd64.whl
custom_wheels/SciPy-1.8.1-cp310-cp310-win_amd64.whl
control
custom_wheels/slycot-0.4.0-cp310-cp310-win_amd64.whl
scikit-learn
dataclasses-json
custom_wheels/pandas-1.4.3-cp310-cp310-win_amd64.whl

Note Please keep the packages in the same order. It is very important that numpy+mkl is the first package installed.

Now enter the following command in your terminal, with your virtual environnement activated:

pip install -r requirements.txt

On Linux

TODO

How to store data and organise files

When using part1 or part2 files, need to select in which folder are or will be inserted input/output files. This folder needs to contain

• main_folder (folder where all input/output files are to be found, can have any name)

  • DATA (where all data will be found)

    • reference/ref.csv (this is used to compare performance of closed loop plant vs reference)
    • data_file_0001.csv
    • data_file_0002.csv
    • etc

  • good_plants

    • v1 (at least)
      • good_plant_v1.txt
    • v2
    • v3
    • etc

  • try_all_results

  • training_plans

    • plan.json

File Description

data_file_00XX.csv

only needs to be in DATA and end with .csv
should have
t,r,u,y
columns (or more, those wont be used)
with column headers

if using other column names, modify the "sensor_data_column_names" section in plan.json
file explained below

should not change name or add other data files in DATA if you want the
good_plants_vX.txt file to train exactly the plants it is meant to train
(this will be changed in the future)

plan.json

- sensor_data_column_names

Used to determine which columns in csv data files used for training and testing 
correspond to time ("t"), command ("r"), input ("u") and output ("y") data of 
plant we want to characterize

- plants_to_train

tells numerator and denominator order of plants to train, also wheter to use 
normalization or standardization while training


format:

```json
{
    "sensor_data_column_names":
    {
        "t": "Time (s)",
        "r": "Command (Nm)",
        "u": "Motor Voltage (V)",
        "y": "Torque Output (Nm)"
    },
    "plants_to_train":
    [
        {
            "num_order": 5,
            "denum_order": 6,
            "better_cond_method" : "normalizing"
        },
        {
            "num_order": 2,
            "denum_order": 3,
            "better_cond_method" : "standardizing"
        }
    ]
}
```

Warning Do not train on more than 4-6 files, and each file should be no more than 500-600 kb, or else your computer will crash. Respecting the max number of files is more important than respecting the file size, but both are important as else this program will fill up your memory.