Julia package for simulation and validation of the TU Delft V3 leading edge inflatable (LEI) kite. Built on SymbolicAWEModels.jl; ships calibration, model setup, flight replay, and bundled V3 geometry.
git clone https://github.com/OpenSourceAWE/V3Kite.jl
cd V3Kite.jl
./bin/install
./bin/run_julia./bin/run_juliainclude("examples/menu.jl")Pick an example from the menu. See V3SimConfig in src/simulation.jl for
the simulation options; bundled geometry and flight data live at
v3_data_path().
First run is slow (compilation + system build). Subsequent runs are fast — keep the same REPL open across runs.
You can make the first run much faster by using a custom system image. You can create it with the script bin/create_system_image. This requires up to 64GB RAM (for example 16GB physical RAM and a 48 GB swap file) and 30 min of time (on a laptop with a Ryzen 7840U CPU).
All examples share one project — julia --project=examples. Run the
interactive launcher with examples/menu.jl, or pick one from the table:
| # | Script | What it does |
|---|---|---|
| 1 | v3kite.jl |
Hello-world: heading PID + 3D replay. Start here. |
| 2 | reel_out_v3.jl |
Single reel-out maneuver, 2D Makie plots. |
| 3 | realtime.jl |
Keyboard-controlled simulation (arrows steer/depower, ESC stops). |
| 4 | open_loop.jl |
Settle in power zone, then ramped open-loop steering. |
| 5 | flight_replay.jl |
Replays real EKF flight data through the simulator. See below. |
| 6 | batch_run_circles.jl |
Parameter sweep of circular-flight runs. See below. |
| 7 | batch_load_circles.jl |
Loads a batch directory; writes metrics CSV + scatter plot. |
| 8 | batch_run_zenith_then_circles.jl |
Two-phase variant (zenith hold → circles). |
| 9 | load_and_plot.jl |
Post-processes any saved log: timeseries, 3D replay, line stretch. |
Utilities (no simulation): photogrammetry_aoa.jl, plot_wind_sources.jl,
depower_drum_model.jl.
batch_run_circles.jl runs a grid of circular-flight sims (settle → ramp
steering → early-stop on course-rate convergence), the study behind the second
paper (Citation). Logs land in
processed_data/<batch_tag>/; failures get listed in failed_runs.txt.
Edit defaults/sweeps/combine_all at the bottom of the script to
define the grid, then include it from the REPL:
include("examples/batch_run_circles.jl")
include("examples/batch_load_circles.jl")batch_run_circles.jl runs one full sim per grid point, so a large sweep can
take a couple of hours.
Keep working in the same REPL session (the one from Quick Start) — each fresh
julia process re-runs precompilation, so staying in-session is much faster.
batch_load_circles.jl prompts for a batch directory and emits
circles_batch_analysis.csv plus the plot below.
|u_s · v_a| vs |χ̇|, one dot per run, with each dot labelled by its swept
value:
Legend — color marks which parameter was swept away from the baseline:
defaults— the baseline run (all parameters at their default).sweep us— steering fraction.sweep up— depower (power) fraction.sweep vw— wind speed (m/s).sweep lt— tether length (m).gk=… (defaults)— turn-rate gainG_kfitted through the baseline run.
flight_replay.jl slices a maneuver from an EKF H5 by UTC, settles the wing
into the recorded conditions, then steps the simulator while feeding recorded
steering/depower/tether inputs. A second SymbolicAWEModel driven straight
from the EKF state is replayed alongside. Outputs land in processed_data/;
PDFs go to output/ when SAVE_FIGS=true. Toggles for maneuver, year,
feedback gains, and tape reductions are at the top of the script.
include("examples/flight_replay.jl")The default run replays a 9 s straight-to-right-turn segment (V3 kite, Oct.
2025), the validation case from the paper (Citation). Stay in the
REPL afterwards: the script leaves a plots named tuple, so
evaluating a field opens that figure (GLMakie). In every plot the continuous
line is the open-loop simulation and the dotted line is the flight data.
plots.trajectory # flight path, colored by kite speed
plots.panels # u_s, F_t, v_app, χ time series
plots.yaw_fig_course # turn rate vs |u_s · v_a|, with least-squares G_k fit
plots.twist[7182] # spanwise twist vs photogrammetry, keyed by video frame
 |
 |
Add or remove panels. The stacked time-series figure is built by
plot_2d_panels; edit panels_kwargs near the bottom of the script to toggle
rows — e.g. show_course, show_aoa, show_drag_coeff, show_lift_coeff,
show_lift_drag_ratio.
Photogrammetry. The twist overlays load from frame_<N>.csv files in
v3_data_path(), auto-selected by the maneuver's UTC window and read via
load_extra_points. Depower tapes are calibrated on the straight-flight frame
so the spanwise-mean angle of attack matches them.
| Constant | Value | Meaning |
|---|---|---|
V3_STEERING_L0_BASE |
1.6 m | Neutral steering tape length |
V3_DEPOWER_L0_BASE |
0.2 m | Neutral depower tape length |
V3_STEERING_GAIN |
1.4 m | Max differential at 100% steering |
V3_DEPOWER_GAIN |
5.0 m | 0–100% depower stroke |
Tape reductions are applied via V3GeomAdjustConfig and set_steering! /
set_depower! — see their docstrings.
When GLMakie is loaded, extra plotters become available, e.g.:
plot_body_frame_local(sys_struct; dir=:front)julia --project -e 'using Pkg; Pkg.test()'- SymbolicAWEModels.jl — symbolic kite modeling
- VortexStepMethod.jl — aerodynamics
- AtmosphericModels.jl — wind shear and turbulence models
- KiteUtils.jl — shared utilities
If you use this package, please cite both papers. The first covers the flight-test validation (see Flight replay), the second the circular-flight simulations (see Batch sweeps):
B. van de Lint and J. A. W. Poland, "Coupled aerodynamic-structural simulation of a leading-edge inflatable kite: validating against flight test data," Journal of Physics: Conference Series, vol. 3224, no. 9, p. 092025, 2026. doi:10.1088/1742-6596/3224/9/092025
J. A. W. Poland, B. van de Lint and R. Schmehl, "Dynamic aero-structural coupled circular flight simulations of soft kites," Journal of Physics: Conference Series, vol. 3224, no. 9, p. 092026, 2026. doi:10.1088/1742-6596/3224/9/092026
@article{vandelint2026coupled,
author = {van de Lint, Bart and Poland, Jelle A. W.},
title = {Coupled aerodynamic-structural simulation of a leading-edge
inflatable kite: validating against flight test data},
journal = {Journal of Physics: Conference Series},
volume = {3224},
number = {9},
pages = {092025},
year = {2026},
publisher = {IOP Publishing},
doi = {10.1088/1742-6596/3224/9/092025},
}
@article{poland2026circular,
author = {Poland, Jelle A. W. and van de Lint, Bart and Schmehl, Roland},
title = {Dynamic aero-structural coupled circular flight simulations of
soft kites},
journal = {Journal of Physics: Conference Series},
volume = {3224},
number = {9},
pages = {092026},
year = {2026},
publisher = {IOP Publishing},
doi = {10.1088/1742-6596/3224/9/092026},
}See CITATION.cff for machine-readable metadata.
This project is licensed under the MPL-2.0 License. The documentation is licensed under the CC-BY-4.0 License. Please see the below Copyright notice.
Technische Universiteit Delft hereby disclaims all copyright interest in the package “V3Kite.jl” (model of the V3 kite) written by the Author(s).
Prof.dr. H.G.C. (Henri) Werij, Dean of Aerospace Engineering, Technische Universiteit Delft.
See the copyright notices in the source files, and the list of authors in AUTHORS.md.