The model.rar file contains trained neural network of the model. To run the output.py correctly, the model.rar file should be extracted in the same directory.
Mark is a robot, which plays in the simulated environment. Whenever Mark hits the ball with it's nose it gets a reward point, and the ball reappears in a new random position. The brain of Mark is neural network which was trained using policy gradient algorithm of reinforcement learning for about 12 hours. The complete game environment is created using pygame module in python. Mark has three actions available to make at a time, rotate right, rotate left, move forward.
A representational video is available at https://youtu.be/J6DzNFhJnq0