An extension of the Bachelor Project of Nil Stolt Ansó and Anton Orell Wiehe. We implement different reinforcement learning algorithms and architectures in order to compare them on a self-written agar.io clone and MuJoCo physics simulation environments.
Clone the github repository and install the following set of dependencies: python3.5.2, h5py, pygame, tensorflow, keras, gym, and matplotlib. To run simply launch the aigar.py file in the source folder using python 3.5.2