An application of reinforcement learning with PyGame
Objective of the game: Avoid obstacles or get penalized Collect coins to increase points
Objective of the project: Demonstrate applicability of reinforcement learning on a simple board game Q-learning (value iteration) of MDP(Markov decision process) has been applied to train agent to take correct action
Play by yourself:
- Install python 2.7 or higher (32 bit) https://www.python.org/download/releases/2.7/
- Install pygame module http://www.pygame.org/download.shtml
- Install numpy https://sourceforge.net/projects/numpy/
- Run module testing_main.py
Watch bot getting trained while it plays the game 4. Run module main.py
Future scope: Add complexity in the game and reward system Apply neural nets to reduce space requirements of virtually infinite state space