We train deep neural networks on a large dataset of state-control pairs (for a spacecraft). The state-control pairs are extracted from trajectories corresponding to the optimal orbital transfer from Earth to Mars.
Code for dataset generation can be found here.
We use Keras with Tensorflow backend for neural network training.
For more details, see paper:
'Machine learning and evolutionary techniques in interplanetary trajectory design'. Izzo D, Sprague C, Tailor D. (2018)
To reproduce environment, execute
pip install -r requirements.txt
.