czlwang / grounded_LTL_parser

[CoRL 2020] Learning a natural-language to LTL executable semantic parser for grounded robotics

grounded_LTL_parser

This is the code for Learning a natural-language to LTL executable semantic parser for grounded robotics

Attention mechanism implementation from [1]

Requirements

• Python 3.7
• Torch 1.2.0
• OpenAI Baselines
• Spot 2.9.3

Install

The execution environment and other python libraries are needed. We recommend creating a new virtual environment before proceeding. In the cloned directory, run the following:

1. git clone --recursive https://github.com/czlwang/grounded_LTL_parser.git
2. cd ltl-environment-dev
3. git checkout stable
4. pip install -e ltl-environment-dev
5. pip install -r requirements.txt

Data

The compressed data is in the data directory.

• The 1k_10_env_1_track_no_neg_human data contains 1,000 examples, with 10 separate demonstrations per example. The sentences are generated by humans.
• The 1k_10_env_1_track_no_neg data contains 1,000 examples, with 10 separate demonstrations per example. The sentences are generated by a grammar.

NOTE for most of our experiments, we only use 3 of the 10 demonstrations during training.

Run

./reinforce.py configs/rl_parse_config.yml(This will run for a long time. If you just want to check if everything is working, you set train_split to be smaller in the config.)

Important config flags

• data_dir should be set to wherever you put the data
• train_mode is either iml, rl, or ml for Iterative Maximum Likelihood, Reinforcement Learning, or Maximum Likelihood respectively
• eval_only is True if running evaluation only. In that case, make sure to set load_pretrained and pretrained as well.

Contents

• reinforce.py contains our code for training
  • iter_ml() contains the Iterative Maximum Likelihood procedure
• model.py contains our sequence-to-sequence model
• rewards.py contains our rewards methods
  • compute_ltl_rewards() contains the reward computation as described in Section 4

Citations

[1] J. Bastings. 2018. The Annotated Encoder-Decoder with Attention. https://bastings.github.io/annotated_encoder_decoder/