Quadcopter Optimal Landing: Neural Network Controller

WARNING: This is an initial proof-of-concept, not a carefully polished project!

Watch a video of landings on YouTube:

Quick Instructions

# Generate Data
pushd generate_data
uv run quadcopter.py  # Wait 50 minutes
popd

# Train
pushd nn
cargo run --release --bin train  # Wait 40 minutes
popd

# Inference
pushd nn
cargo run --release --bin vis_results  # Pretty quick
# Creates PNG frames from animated trajectory in nn/animation
popd

See here for a detailed walkthrough.

Slides

There are some slides from a presentation to the Sydney Functional Programming Group here: https://lancelet.github.io/nn-landing-poc.