Distributed-MADDPG

Master Course Graduation Project - Distributed Multi-Agent DDPG using PPO as policy optimizer

Composition introduction

1. Actor network in ddpg

• input: state_dim
• => dense(64) + ReLU + norm
• => dense(64) + ReLU + norm
• => output: dense(action_dim) + softmax

Model(inputs=input_obs,outputs=pred)
model.compile(optimizer='Adam',loss='categorical_crossentropy')

ToDO

A lot of work to do.