This is the code repository for the paper "Encoding formulas as deep networks: Reinforcement learning for zero-shot execution of LTL formulas"
- Python 3.7
- OpenAI Baselines
- Spot 2.9.3
We recommend to run in a new virtual environment. Please installed the library listed above before proceeding. You can clone this repo and run pip install -r requirements.txt to get the dependencies.
run_char.sh and run_craft.sh are the training scripts for the experiments. This will run for a long time as it also runs the test set every 100 formula updates.