/PCLaSt

Codebase for the PCLaSt paper

Primary LanguagePython

Hierarchical Planning

Setup

pip install -r req.txt

Data generation

env="room-multi-passage" # opts are "polygon-obs", "room-multi-passage", "room-spiral", "room-multi-passage-large"

python main.py --env ${env} --opr generate-data

Training PCLaSt encoder

env="room-multi-passage" # opts are "polygon-obs", "room-multi-passage", "room-spiral", "room-multi-passage-large"

python main.py --env ${env} --opr train --max_k 10 --contrastive --contrastive_k 2 --ndiscrete 64

Clustering

env="room-multi-passage" # opts are "polygon-obs", "room-multi-passage", "room-spiral", "room-multi-passage-large"

python main.py --env ${env} --opr cluster-latent

Generate Emphirical MDP

env="room-multi-passage" # opts are "polygon-obs", "room-multi-passage", "room-spiral", "room-multi-passage-large"

python main.py --env ${env} --opr generate-mdp

Excecute plan for goal reaching

env="room-multi-passage" # opts are "polygon-obs", "room-multi-passage", "room-spiral", "room-multi-passage-large"

python main.py --env ${env} --opr high-low-plan --from_to 2 15