Skip to content

code for DyPS: Dynamic Parameter Sharing in Multi-Agent Reinforcement Learning for Spatio-Temporal Resource Allocation

Notifications You must be signed in to change notification settings

tsinghua-fib-lab/DyPS

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DyPs

This is the code for paper "DyPS: Dynamic Parameter Sharing in Multi-Agent Reinforcement Learning for Spatio-Temporal Resource Allocation "

Installation and Setups

conda create -n DyPs python=3.9
conda activate CoTa
pip install -r requirements.txt

Run Experiments

  • Command to run our method
cd run
python run_lstm_hrl.py
  • Command for baselines
cd run
python run_hrl.py   # Our method without lstm
python run_cvae.py  # Our method withour hierarchical structure
python PS_noid.py   # Naive parameter sharing 

Visualizations

  • You can visualize the learning curves by tensorboard.
tensorboard --logdir logs
  • You can visualize the city map and demand-supply heat map by following jupyters.
plot/grid_map.ipynb
  • training curve of ride-hailing (2) scenario

曲线图

About

code for DyPS: Dynamic Parameter Sharing in Multi-Agent Reinforcement Learning for Spatio-Temporal Resource Allocation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published