This is the companion repository for the following paper. Trained and tested with Python 3.6 + TensorFlow 1.12.0.
L. Liang, H. Ye, and G. Y. Li, "Spectrum sharing in vehicular networks based on multi-agent reinforcement learning," to appear in IEEE Journal on Selected Areas in Communications, Sep. 2019. Also available at: https://arxiv.org/abs/1905.02910
How to use the code:
- To train the multi-agent RL model: main_marl_train.py + Environment_marl.py + replay_memory.py
- To train the benchmark single-agent RL model: main_sarl_train.py + Environment_marl.py + replay_memory.py
- To test all models in the same environment: main_test.py + Environment_marl_test.py + replay_memory.py + '/model'. Change the V2V payload size using the variable "self.demand_size" in Environment_marl_test.py.
Please send all questions/inquires to lliang@gatech.edu.