This repository contains the material used for the Pommerman project. The methods used to get a winrate of over 50 percent over 3 simpleagents was PPO together with random network distillation.

Several files can be found including:

• Notebook with results and pre-trained model to run tests
• Notebook with code used for training
• Python file that changes the pommerman enviornment to an input used by our model

Required:

• Exploration by Random Network Distillation
• Proximal Policy Optimization Algorithms