soraxas/CoRL_flow-based-motion-planner-all

Shell

Populating sub repository

This repository contain 3 nested repos, with some clever dirty tricks.

Before anything would work, run

make

to populate all of the nested repositories, or

make clean

to cleanup and revert the previous operation.

Run

make weight-files

to download and extracts the flow-based distribution trained weights into normalising-flow-distribution/.

Generating runtime stats

Folder: GenerateStats/
Run
```
python main.py
```
which would run 3 random maps, for 3 pairs of $q_\text{init},q_\text{goal}$ and for 3 times.
Use git checkout {2,4,6}d to go to the specific branch for the desire dimensions, eg. git checkout 2d for generating stats for 2d.

Learning $\mathcal{Q}_\theta$ and visualising results

Folder: normalising-flow-distribution/

Generating random environments $\omega$

Folder: path_dataset_generator/