A reinforcement learning simulator in Python
This is a simulation of a boxy creature whose goal is to pick up the blue 'fruit' and deliver it to the green-colored goal marker. The agent is rewarded for actions such as picking up the object and walking over to the goal with it.
The script saves the agent's progress every 10 episodes to agent.sav. Each episode has a total of 200 steps. When loading the script a second time, it will attempt to load a brain from agent.sav if it exists.
- Numpy
- PyGame
- Pickle