Skip to content

tairtahar/reverse_curriculum_vicinity_learning

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

python torch gym numpy

RCVL: Reverse Curriculum Hierarchical Vicinity Learning.

This repository was developed for my master's thesis in Artificial Intelligence. It contains a hierarchical algorithm based on two levels in which the higher level draws the high-level path with milestones (subgoals) and the low hierarchy performs the primitive steps in the sub-trajectories between those milestones.

The algorithm is trained and tested in Simple Minigrid environemt: Empty Room 15x15 and FourRooms 15x15.

Dependencies

Usage

Running experiments:

  1. For execution of training please run train_rcvl.py. Make sure to insert job_name and to adjust any of the parameters if needed.
  2. For execution of testing please run test_rcvl.py. Make sure to insert checkpoint_name, to add the relevant files into the checkpoints directory, and to adjust any of the parameters if needed. The files default checkpoint_name is the algorithms that were presented in the thesis for four_rooms environment. In addition, in the checkpoint directory, it is possible to find the checkpoint for empty room.

~

Trained RCVL policy

Simple-MiniGrid-FourRooms-15x15-v0 learned policy

For rendering the test, add --render to the configuration file.

Citation

 @phdthesis{Tahar Tair,
    title={Reverse Curriculum Vicinity Learning},
    url={https://upcommons.upc.edu/handle/2117/371021},
    school={UPC, Computer Science Faculty},
    author={Tahar, Tair},
    year={2022},
    month={July}
}

About

Reverse Curriculum Vicinity Learning: Reinforcement Hierarchical algorithm for solving complex RL tasks.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages