DecentralizedLearning
Code associated with Daniel Willemsen's Master Thesis in Aerospace Engineering at the Delft University of Technology: "Sample-efficient multi-agent reinforcement learning using learned world models"
To recreate the experiments in the thesis, use run_experiment.py. This file is currently set to recreate the MAMBPO runs in the cooperative navigation domain. To recreate the other experiments, change the config name or the environment name in run_experiment.py.
analyze_log.py is used to plot the data generated from run_experiment.py (which is periodically saved in a .p file)
The full master thesis will soon be available at the TU Delft education repository.