leekooktao / scrimp

offical code of paper 'SCRIMP: Scalable Communication for Reinforcement- and Imitation-Learning-Based Multi-Agent Pathfinding'

SCRIMP

This is the code for implementing the SCRIMP algorithm :SCRIMP: Scalable Communication for Reinforcement- and Imitation-Learning-Based Multi-Agent Pathfinding

Requirements

Python == 3.7

 pip install -r requirements.txt

Setting up Code

• cd into the od_mstar3 folder.
• python3 setup.py build_ext --inplace.
• Check by going back to the root of the git folder, running python3 and import od_mstar3.cpp_mstar.

Running Code

• Modify the parameters in alg_parameters.py to set the desired training setting and recording methods.
• Call python driver.py.

Key Files

alg_parameters.py - Training parameters.

driver.py - Driver of program. Holds global training network for PPO.

episodic_buffer.py - Defines the episodic buffer used to generate intrinsic rewards.

eval_model.py - Evaluates trained model.

mapf_gym.py - Defines the classical Reinforcement Learning environment of Multi-Agent Pathfinding.

model.py - Defines the neural network-based operation model.

net.py - Defines network architecture.

runner.py - A single process for collecting training data.

Other Links

Fully trained SCRIMP model - https://www.dropbox.com/scl/fo/ekhxyt7gm575kfwaerwb5/h?rlkey=j3cdikwofz0zelj2oci9q97k8&dl=0

Authors

Yutong Wang

Bairan Xiang

Shinan Huang

Guillaume Sartoretti