pfheatwole / thesis

Parametric Paraglider Modeling

Parametric Paraglider Modeling

This repository contains the source material for my Master's thesis: Parametric Paraglider Modeling

The project develops a parametric paraglider flight dynamics model suitable for simulating flight trajectories under typical flight conditions. Its motivating purpose is to approximate the flight dynamics of commercial paraglider wings using only basic technical data.

To enable its use in common engineering applications (such as control modeling and statistical filtering) the dynamics are encoded as differential equations (ẋ = f(x, u)). An implementation of the dynamical system (including a rudimentary flight simulator) is available as a Python library under a permissive open source license: https://github.com/pfheatwole/glidersim/

Backstory

The motivating goal was to learn wind patterns from sets of recorded paraglider flights. The primary difficulty is that flight records are position-only time series data, so extracting wind information requires a causal relationship between paraglider motion and the underlying wind field. That causal link is provided by the paraglider dynamics. If the dynamics of the paraglider that created a flight are known, it may be possible to use statistical filtering to perform probabilistic flight reconstruction; specifically, the problem requires generating probability distributions over the paraglider state, control inputs, and wind vectors.

Unfortunately, flight records do not include a dynamics model of the glider that generated the data (in fact, they frequently omit any description of the glider at all), so the dynamics model itself is uncertain. Thus, performing flight reconstruction over a large number of flight tracks requires more than a single dynamics model: it needs an entire distribution of models. Developing models is time consuming and requires extensive construction details that are not readily available, so the new goal was to create a tool that can generate paraglider models from publicly available information. Specifically, I wanted a dynamics model that is parametrized by the most readily available wing data: user manuals, pictures, and physical measurements. This paper develops such a parametric model.

Building

This document was produced using Sphinx.

To rebuild the document from its source:

$ git clone https://github.com/pfheatwole/thesis.git
$ cd thesis
$ python -m venv .venv
$ source .venv/bin/activate
$ pip install -r requirements/build.txt
$ make html

Many of the figures were produced by scripts using the Python library. To run any Python scripts that depend on pfh.glidersim :

$ git clone https://github.com/pfheatwole/thesis.git
$ cd thesis
$ python -m venv .venv
$ source .venv/bin/activate
$ pip install -r requirements/develop.txt
$ python source/figures/paraglider/belloc/belloc.py  # for example