I am using this repository to give an overview of basic reinforcement learning techniques and show case some of the projects I have been doing next to university.
- Project CliffWalking.py is a demonstration of SARSA vs. Q-Learning. It simulates an agent in a grid world of size 6 x 10 that tries to get from one side to the other while avoiding a dangerous cliff. The project aims to contrast the different policy update methods. Conclusion shows that Q-Learning finds the ideal route using the shortest path, while SARSA only approximates the ideal route, maintaining a safe distance to the cliff.
- The project can be easily adapted to grid worlds of varying sizes. All variables are declared in the Constructor and should be easy to interpret. Feel free to use this code or reach out to me if you would like to collaborate further on it!