M-Parvini/CACC-RRM-Design-ICC-2023

Simulation code for the paper "Joint Resource Allocation and String-Stable CACC Design with Multi-Agent Reinforcement Learning"

CACC-RRM-Design-ICC-2023

Simulation code of the paper: "Joint Resource Allocation and String-Stable CACC Design with Multi-Agent Reinforcement Learning"

If you want to cite:

M. Parvini, A. Gonzalez, A. Villamil, P. Schulz and G. Fettweis, “Joint Resource Allocation and String-Stable CACC Design with Multi-Agent Reinforcement Learning,” in Proceedings of 2023 International Conference on Communications (ICC 2023), Rome, Italy, May 2023.

---

prerequisites:

1) python 3.7 or higher
2) PyTorch 1.7 or higher + CUDA
3) It is recommended that the latest drivers be installed for the GPU.

---

Algorithms that you can evaluate:

1. Federated Multi-Agent Reinforcement Learning
   • Set Algorithm = Hybrid, federated_communication = True, Train = True, Test = True
   • You can also change the activation function to have either a linear or nonlinear function approximation model.
     • Set activation=Relu, elu or leaky_relu for nonlinear function approximation or set activation=linear otherwise.
2. Decentralized Multi-Agent Reinforcement Learning
   • Set Algorithm = Hybrid, federated_communication = False, Train = True, Test = True
   • You can also change the activation function to have either a linear or nonlinear function approximation model.
     • Set activation=Relu, elu or leaky_relu for nonlinear function approximation or set activation=linear otherwise.
3. Sum-capacity optimization
   • Set Algorithm = Opt_non_fair, federated_communication = False, Train=False, Test = True
4. Max-Min optimization
   • Set Algorithm = Opt_fair, federated_communication = False, Train = False, Test = True
5. Random
   • Set Algorithm = random, federated_communication = False, Train = False, Test = True

Good Luck with your simulations!!!