source

This repository contains MATLAB scripts as well as recorded flight data used for system identification purposes on the commercial Parrot Bebop 2 drone.

For support regarding the usage of this repository, please contact one of the authors

Matthijs van der Boon - [email protected]
Pim de Bruin - [email protected]
Jasper Wiersema - [email protected]
Eduard Jongkees - [email protected]
Ewoud van Mourik - [email protected]

A special thank you is extended to H. Zhu for providing a lot of the code base for running flight experiments.

Files used for recording experiment data are located in /data. A guide for experiment setup in the DCSC lab is available from the supervisors. A troubleshoot document is online here. Running run_main starts an experiment. Run save_log to save all recorded data after an experiment.

Roll, pitch and vertical velocity input vectors can be set in modules/CDrone by setting object properties m_ppi_phi, m_ppi_theta, and m_ppi_vz.

Scripts used for processing recorded data are located in /processing. Run batch to batch process all recorded data and save experiment data into separate data .mat files. Other scripts in this folder are used to analyse recorded data.

Files used for system identification are located in /systemidentification. The main workflow to identify new models is as follows.

1. use batchIdentify or CCDInitial to identify new models using either random initial values, or initial values based on central composite design.
2. For first order models FONLIFnc or FOPTDNIFnc are used for identification. GONLIdentification is used to identify models of higher order.
3. Generated models are saved as .mat files.

Files used for system identification are located in /validation. The workflow for validation is as follows.

1. Pick the best identified model for each data set using batchPicker
2. Use batchValidation to validate the best models over validation data sets. Batch validation uses simGOFull to simulate model responses. The results of validation are saved in .mat files.
3. Run [valddata, simdata] = simGOFull([],[]) to simulate the response of a desired model over a data set. UI dialogs are opened to select model and data.
4. Plot the response of the simulation by running plotResultAndSave(valdata, simdata). Figures can be saved if desired.