I-ETC 2023 Drone Propeller Rotation Analysis

This repository contains the code used in our paper Preliminary Analysis of Drone Propeller Rotation using Wingbeat-Modulation LiDAR, which has been submitted to the I-ETC 2023 conference.

Path setup

Run the setup.m script to add the feature-extraction folder to your matlab path. Or just add it manually using addpath()

Directory setup

The code has been written to use relative paths. This can be changed if desired.

Here's the directory structure this code is set to use:

i-etc-2023
├── code
│   ├── combineScans.m
│   ├── LICENSE
│   ├── nestedcell2mat.m
│   ├── orientationTrainTestSplit.m
│   ├── randomTrainTestSplit.m
│   └── README.md
└── data
    ├── on-axis
    ├── perpendicular
    ├── raw
    ├── testing
    └── training

Recreating results

Our code is split into multiple parts: data wrangling and preprocessing, training, testing, and figure creation.

Data wrangling

1. Run combineScans.m to combine all raw data into one data structure.
2. Run extractAllFeatures.m to preprocess the data and extract features.
3. Run randomTrainTestSplit.m and orientationTrainTestSplit.m to split the data into training and testing sets.

Training

Run the experiment*.m files to run the training experiments. These scripts launch the Classification Learner with the appropriate data. You are then free to run whatever models you like in the Classification Learner.

Models must be exported and saved manually, unfortunately.

Testing

Once the models have been trained and exported from the Classification Learner application, run the testing scripts:

• testModel.m tests the random split
• perpendicularModelTest.m tests the models trained on perpendicular data
• onAxisModelTest.m tests the models trained on on-axis/parallel data

Figure creation

Figure creation scripts are in the figures directory. These scripts are designed to be run from the figures directory. If you run them from elsewhere, you'll need to update the data directory paths appropriately.