cqn2219076254 / marl-dmfb

A cooperative multi-agent reinforcement learning framework for droplet routing for droplet routing in DMFB

MARL-DMFB

A cooperative multi-agent reinforcement learning framework for droplet routing for droplet routing in DMFB

Train model

python train.py dmfb --drop_num=4 --fov=9

the training data will be saved in "data-dmfb/TrainResult/vdn/fov9/10by10-4d0b/"

the trained model will be saved in "data-dmfb/model/vdn/fov9/4d0b/

Evaluate model in health mode

python evaluate.py dmfb --drop_num=4 --chip_size=20 --evaluate_task=100 --show

This will evaluate the performance of the model: "data-dmfb/model/vdn/fov9/4d0b/rnn_net_params.pkl" and "data-dmfb/model/vdn/fov9/4d0b/vdn_net_params.pkl"

Evaluate model with electrodes degrade

python evaDegre.py dmfb --evaluate_task = 20 --fov=9 --drop_num=4

This will evaluate the performance of the model: "data-dmfb/model/vdn/fov9/4d0b/rnn_net_params.pkl" and "data-dmfb/model/vdn/fov9/4d0b/vdn_net_params.pkl"

The data will be saved in "data-dmfb/DgreData/10by10-4d0b"

For MEDA

just change 'dmfb' to 'meda' in the commands: i.e., training command is python train.py meda --drop_num=4

Then all data are saved under fold "data-meda/"

Parameters

You can find more usages or change the parameters of the algorithm in the file "common/arguments.py"